Int. J. Applied Pattern Recognition, Vol. 1, No. 2, 2014

Fruit disease recognition using improved sum and difference histogram from images Shiv Ram Dubey* and Anand Singh Jalal Department of Computer Engineering and Applications, GLA University, Mathura, Uttar Pradesh, India E-mail: [email protected] E-mail: [email protected] *Corresponding author Abstract: Diseases in fruit cause devastating problem in production and availability. The classical approach of fruit disease recognition is based on the naked eye observation by experts. Detection of defects is still problematic due to the natural variability of colour in different types of fruits, high variance of defect types, and presence of stem/calyx. In this paper, a framework for the recognition of fruit diseases is proposed. The proposed approach is composed of the following three main steps; defect segmentation, feature extraction, and classification. This paper also introduces an improved sum and difference histogram (ISADH) texture feature based on the intensity values of the neighbouring pixels. The gradient filters are also used with ISADH in this paper to boost the discriminative ability. We have considered apple diseases as a test case and evaluated our program. Experimental results suggest that the proposed method can significantly support automatic recognition of fruit diseases. The classification accuracy has achieved more than 97% using ISADH texture feature. Our method is able to achieve nearly 99.9% of accuracy in conjunction with the gradient filters. Keywords: K-means clustering; sum and difference histogram; multi-class support vector machine; MSVM; texture classification. Reference to this paper should be made as follows: Dubey, S.R. and Jalal, A.S. (2014) ‘Fruit disease recognition using improved sum and difference histogram from images’, Int. J. Applied Pattern Recognition, Vol. 1, No. 2, pp.199–220. Biographical notes: Shiv Ram Dubey was a Research Fellow in Computer Vision lab, CSE, IIT Madras. His research was funded by DST, Government of India. He received his BTech in Computer Science and Engineering in 2010 from the Gurukul Kangri Vishwavidyalaya Haridwar, India and his MTech in Computer Science and Engineering in 2012 from GLA University Mathura, India. His current research interests are image processing, computer vision and machine learning. Anand Singh Jalal received his MTech in Computer Science from Devi Ahilya Vishwavidyalaya, Indore, India. He received his PhD in the area of Computer Vision from Indian Institute of Information Technology (IIIT), Allahabad, India. He has 14 years of teaching and research experience and currently, he is working as Head and Professor in Department of Computer Engineering and Applications, GLA University, Mathura, India. His research interests include image processing, computer vision and pattern recognition.

Copyright © 2014 Inderscience Enterprises Ltd.

199

200

1

S.R. Dubey and A.S. Jalal

Introduction

The classical approach for detection and identification of fruit diseases is based on the naked eye observation by experts. In some developing countries, the availability of fruit disease experts is an important issue due to their distant locations. The various types of diseases in fruits determine the quality, quantity, and stability of yield. The diseases in fruits not only reduce the yield but also deteriorate of the variety and its withdrawal from the cultivation. Fruit diseases appear as spots on the fruits and if not treated on time, cause severe losses. Excessive use of a pesticide for fruit disease treatment increases the danger of toxic residue level of agricultural products and has been identified as a major contributor to the ground water contamination. Pesticides are also among the highest components in the production cost, thus their use must be minimised. Early detection of disease and crop health can facilitate the control of fruit diseases through proper management approaches such as vector control through fungicide applications, disease-specific chemical applications and pesticide applications; and improved productivity. Automatic detection of fruit diseases is essential to automatically detect the symptoms of diseases as soon as they appear on the growing fruits. Monitoring of health and detection of diseases is critical in fruits and trees for sustainable agriculture. To the best of our knowledge, no sensor is available commercially for the real time assessment of trees health conditions. Scouting is the most widely used method for monitoring stress in trees, but it is expensive, time-consuming and labour-intensive process. Polymerase chain reaction which is a molecular technique used for the identification of fruit diseases, but it requires detailed sampling and processing. Fruit diseases can cause significant losses in yield and quality appeared in harvesting. For example, soybean rust (a fungal disease in soybeans) has caused a significant economic loss and just by removing 20% of the infection, the farmers may benefit with an approximately 11 million-dollar profit (Robbins et al., 2006). Some fruit diseases also infect other areas of the tree, causing diseases of twigs, leaves and branches. An early detection of fruit diseases can aid in decreasing such losses and can stop further spread of diseases. To know what control factors to take next year to avoid the same kind of losses, it is crucial to recognise what is being observed. Some common diseases of apple fruits are apple scab, apple rot, and apple blotch (Hartman, 2010). Apple scabs are grey or brown corky spots. Apple rot infections produce slightly sunken, circular brown or black spots that may be surrounded by a red halo. A fungal disease, apple blotch appears on the surface of the fruit as dark, irregular or lobed edges. The precise segmentation is required for the defect detection. Major works performed defect segmentation of fruits are done using a simple threshold approach (Li et al., 2002; Mehl et al., 2002). A globally adaptive threshold method (modified version of Otsu’s approach) to segment fecal contamination defects on apples are presented by Kim et al. (2005). Classification-based methods attempt to partition pixels into different classes using different classification methods. Bayesian classification is the most used method by researchers Kleynen et al. (2005) and Leemans et al. (1999), where pixels are compared with a pre-calculated model and classified as defected or healthy. Unsupervised classification does not benefit any guidance in the learning process due to lack of target values. This type of approach was used by Leemans et al. (1998) for defect segmentation. The recent work related to fruit disease recognition using machine learning is reported in Dubey and Jalal (2014). Adaboost improved fast PCA algorithm is also applied for dimensionality reduction (Kumar et al., 2011a).

Fruit disease recognition using improved sum and difference histogram

201

In this paper, we propose and experimentally validate a framework for the fruit disease recognition problem and also show the significance of clustering technique in the disease segmentation. The presented approach works in three steps; defect segmentation, feature extraction and training and classification. This paper also introduces an efficient texture feature from the intensity values of the neighbouring pixels. Proposed texture feature is also tested with the different gradient filters in this paper. In order to validate the proposed approach, we have considered apple as a test case and evaluated our program on the three types of the apple diseases. The rest of the paper is structured as follows: Section 2 focuses on the related works; Section 3 presents a novel methodology for the recognition of fruit diseases using multi-class support vector machine (MSVM), this section also introduces an efficient texture feature for these types of image categorisation problems; Section 4 shows the used dataset of apple diseases and presents a detailed result analysis; finally Section 5 made the conclusions and highlighted the future work.

2

Literature review

In this section, we focus on the previous work done by several researchers in the area of image categorisation, fruit disease identification and defect segmentation. Fruit disease identification can be seen as an instance of image categorisation. The spectroscopic and imaging techniques are unique disease monitoring approaches that have been used to detect diseases and stress due to various factors, in plants and trees. Current research activities are toward the development of such technologies to create a practical tool for a large-scale real-time disease monitoring under field conditions. Various spectroscopic and imaging techniques have been studied for the detection of symptomatic and asymptomatic plant and fruit diseases. Some the methods are: fluorescence imaging used by Bravo et al. (2004); multispectral or hyperspectral imaging used by Moshou et al. (2006); infrared spectroscopy used by Spinelli et al. (2006); visible/multiband spectroscopy used by Yang et al. (2007) and Chen et al. (2008); and nuclear magnetic resonance (NMR) spectroscopy used by Choi et al. (2004). Hahn (2009) reviewed multiple methods (sensors and algorithms) for pathogen detection, with special emphasis on postharvest diseases. Several techniques for detecting plant diseases is reviewed in Sankarana et al. (2010) such as, molecular techniques, spectroscopic techniques (fluorescence spectroscopy and visible and infrared spectroscopy), and imaging techniques (fluorescence imaging and hyper-spectral imaging). A ground-based real-time remote sensing system for detecting diseases in arable crops under field conditions is developed (Moshou et al., 2005), which considers the early stage of disease development. The authors have used hyper-spectral reflection images of infected and healthy plants with an imaging spectrograph under ambient lighting conditions and field circumstances. They have also used multi-spectral fluorescence images simultaneously using UV-blue excitation on the same plants. They have shown that it is possible to detect the presence of disease through the comparison of the 550 and 690 nm fluorescence images. Large scale plantation of oil palm trees requires on-time detection of diseases as the ganoderma basal stem rot disease was present in more than 50% of the oil palm plantations in Peninsular Malaysia. To deal with this problem, airborne hyper-spectral

202

S.R. Dubey and A.S. Jalal

imagery offers a better solution (Shafri and Hamdan, 2009) in order to detect and map the oil palm trees that were affected by the disease on time. Airborne hyper-spectral has provided data on user requirement and has the capability of acquiring data in narrow and contiguous spectral bands which makes it possible to discriminate between healthy and diseased plants better compared to multispectral imagery. Citrus canker is among the most devastating diseases that affect the marketability of citrus crops. In Qin et al. (2009), a hyper-spectral imaging approach is developed for detecting canker lesions on citrus fruit and hyper-spectral imaging system is developed for acquiring reflectance images from citrus samples in the spectral region from 450 to 930 nm. In Purcell et al. (2009), Purcell et al. have investigated the power of NIR spectroscopy as an alternative to rate clones of sugarcane leaf spectra from direct measurement and examined its potential using a calibration model to successfully predict resistance ratings based on a chemometrics approach such as partial least squares. To populate the nature of the NIR sample, they have undertaken a scanning electron microscopy study of the leaf substrate. Marcassa et al. (2006) have applied laser-induced fluorescence spectroscopy to investigate biological processes in orange trees. They have investigated water stress and Citrus Canker, which is a disease produced by the Xanthomonas axonopodis pv. citri bacteria. They have discriminated the Citrus Canker’s contaminated leaves from the healthy leaves using a more complex analysis of the fluorescence spectra. However, they were unable to discriminate it from another disease. Some recent works to identify the fruit disease from images using image processing techniques is incorporated by Dubey et al. (2013) and Dubey and Jalal (2012b, 2012c), where they first segmented the infected area of the image then extracted some features from that infected area and finally classified it with trained classification approaches. Some other approaches such as local similarity decisions (James, 2013), contourlet transform (Kumar and Sheeba, 2014), priority curve (Kumar et al., 2010), DCT features (Varshney et al., 2014), feature combination (Singh et al., 1012), Hu moments (Gupta et al., 2013), 3D model (Kumar et al., 2011b) and SIFT features (Li, 2014) are also being used in several applications of recognition and classifications. KNN is also being used by several researchers as classifier due to its simplicity and efficiency (James and Dimitrijev, 2010; Fukunaga and Hummels, 1987; Cover and Hart, 1967).

3

Fruit disease recognition

The steps of the proposed approach are shown in Figure 1. For the fruit disease recognition problem, precise defect segmentation is required; otherwise the features of the non-infected region will dominate over the features of the infected region. In this approach K-means-based defect segmentation is performed to detect the region of interest, i.e., infected part only. In the second step, features are extracted from the segmented image of the fruit. Finally, training and classification is done by a MSVM.

Fruit disease recognition using improved sum and difference histogram Figure 1

203

Framework of the proposed approach

Training images

Defect segmentation and Feature extraction

Feature database

Image for testing

Defect segmentation and Feature extraction

Trained MSVM Classification Kind of disease in the fruit

3.1 Defect segmentation K-means clustering technique is used for defect segmentation. Images are partitioned into four clusters in which one cluster contains the majority of the diseased part of the image. K-means clustering algorithm was developed by MacQueen (1967) and later enhanced by Hartigan and Wong (1979). The K-means clustering algorithms classify the objects into K number of classes based on a set of features. Algorithm for the K-means defect segmentation: Step 1

Read input image.

Step 2

Transform image from RGB to L*a*b* colour space.

Step 3

Classify colours using K-means clustering in ‘a*b*’ space.

Step 4

Label each pixel in the image from the results of K-means.

Step 5

Generate images that segment the image by colour.

Step 6

Select the segment containing the disease.

In this experiment, squared Euclidean distance is used for the K-means clustering. We have used L*a*b* colour space because the colour information in the L*a*b* colour space is stored in only two channels (i.e., a* and b* components), and it causes reduced processing time for the defect segmentation. In this experiment input images are partitioned into four segments. From the empirical observations, it is found that using three or four cluster yields good segmentation results. The process of selection of a defective cluster from among n clusters identified is based on the mean ‘a*’ and ‘b*’ value of each cluster. We experimentally determined the range ‘a*’ and ‘b*’ between which most of the defected pixels have its intensity values. The cluster having the mean ‘a*’ and ‘b*’ in this range is selected as the defected cluster. Figure 2 demonstrates the output of K-means clustering for an apple fruit infected with apple rot disease. Figure 3 also depicts some more defect segmentation results using K-means clustering technique.

204 Figure 2

Figure 3

S.R. Dubey and A.S. Jalal K-means clustering for an apple fruit that is infected with apple rot disease, (a) the infected fruit image (b) first cluster (c) second cluster (d) third cluster (e) fourth cluster (f) single grey-scale image coloured based on their cluster index (see online version for colours)

(a)

(b)

(c)

(d)

(e)

(f)

Some defect segmentation results using K-means clustering technique (see online version for colours)

Fruit disease recognition using improved sum and difference histogram

205

3.2 Feature extraction In this sub-section, we are presenting a texture feature for the image categorisation problems. Unser (1986) has defined sum and difference histogram of an image which are calculated from the sum and difference of two intensity values with a displacement of (d1, d2). Unser has considered the displacement in x- and y-directions simultaneously, but it lacks some information which is present in the x- and y-directions separately. We improved the Unser’s descriptor by considering the information present in x- and y-direction separately. To use the information presented in both x- and y-directions, first we calculate the sum and difference in x-direction and then simulate this result in the y-direction. Simulation is carried out by taking the sum and difference in y-direction on outcome of x-direction.

3.2.1 Improved sum and difference histogram texture feature Improved sum and difference histogram (ISADH) texture feature is an improvement in the Unser’s feature. It is shown that ISADH feature outperformed other state-of-art features in fruit and vegetable classification scenario (Dubey and Jalal, 2012a, 2013; Dubey, 2012). We analyse the accuracy of ISADH texture feature and compare with other colour and texture features for the fruit disease recognition and classification problem in the experimental result section. ISADH feature algorithm is given as, 1

2

3

4

Find the sum S and difference D for the first channel of an image I with a displacement of (1, 0) as: S ( x, y ) = I ( x, y ) + I ( x + 1, y )

(1)

D( x, y ) = I ( x, y ) − I ( x + 1, y )

(2)

Find the sum S1 and the difference D1 of S with a displacement of (0, 1) as: S1( x, y ) = S ( x, y ) + S ( x, y + 1)

(3)

D1( x, y ) = S ( x, y ) − S ( x, y + 1)

(4)

Find the sum S2 and difference D2 of D with a displacement of (0, 1) as: S 2( x, y ) = D( x, y ) + D( x, y + 1)

(5)

D 2( x, y ) = D( x, y ) − D( x, y + 1)

(6)

Find the histogram for the first channel by concatenating the histograms of S1, D1, S2, and D2 as H 1 = [ S1 D1 S 2 D 2]

(7)

5

Repeat Step 1 to Step 4 for the second and third channel of the colour image.

6

Concatenate the histograms of all three channels in order to find the ISADH texture feature of the input image I as ISADH = [ H 1 H 2 H 3]

(8)

206

S.R. Dubey and A.S. Jalal where H1, H2, and H3 are the histograms obtained in Step 4 for red, green and blue channels respectively).

ISADH texture feature relies upon the intensity values of neighbouring pixels. The histogram of two images of the same class may vary significantly. On the other hand, the ISADH feature has less difference for these images. If the difference in feature of two images is less, then images are more likely belongs to the same class. But if the difference is significant, then images are more likely belongs to the different class. This can be illustrated by an example of two 5 × 5 matrix having intensity values in the range of 0 to 7. Let us have two matrices ‘A’ and ‘B’ having little difference in their values as follows: Matrix ‘A’

Matrix ‘B’

3

5

3

4

6

3

7

3

4

6

2

4

2

6

1

2

3

2

4

1

2

6

4

1

7

2

7

4

1

7

2

4

5

2

6

2

4

3

2

6

2

5

4

4

5

2

3

4

4

3

Calculate three features 1

simple histogram

2

Unser’s feature

3

ISADH feature.

The length of each feature calculated is 8-bin. Table 1 shows the simple histogram for ‘A’ and ‘B’, Table 2 shows the Unser’s feature for ‘A’ and ‘B’, and Table 3 shows the ISADH feature for ‘A’ and ‘B’. Table 1

Simple histogram for both matrices (8-bin)

Simple histogram

I0

I1

I2

I3

I4

I5

I6

I7

Matrix ‘A’

0

2

6

2

6

4

4

1

Matrix ‘B’

0

2

6

6

6

0

2

3

Table 2

Unser feature for both matrices (8-bin)

Unser’s feature

I0

I1

I2

I3

I4

I5

I6

I7

Matrix ‘A’

0.5

2.5

7

2.5

0.5

3.5

6

2.5

Matrix ‘B’

0.5

5

5

2

0.5

3.5

5.5

3

Table 3

ISADH feature for both matrices (8-bin)

ISADH feature

I0

I1

I2

I3

I4

I5

I6

I7

Matrix ‘A’

0

6.25

2

4.25

0.5

5.75

1.25

5

Matrix ‘B’

1.25

5

1.75

4.5

1.25

5

1.5

4.75

Fruit disease recognition using improved sum and difference histogram

207

In Tables 1, 2, and 3, I0 to I7 represents the intensity levels (i.e., 0 to 7 for 8-bin). Let the difference between the feature of matrix ‘A’ and the feature of matrix ‘B’ are defined as the sum of square of difference of the values for each intensity level and can be calculated using the following equation 7

Diff =

∑ ( FA(i) − FB(i) )

2

(9)

i =0

where FA is a feature of matrix ‘A’, FB is a feature of matrix ‘B’, and Diff is the difference between FA and FB. Table 4 shows the values of difference for three features simple histogram, Unser’s feature, and ISADH feature for the ‘A’ and ‘B’. From Table 4, it is clear that ISADH feature has the lowest value of difference between ‘A’ and ‘B’. The value of Diff will be minimal if ‘A’ and ‘B’ are more likely belongs to the same class and that is achieved in the case of ISADH feature. Table 4

Difference in feature of matrix ‘A’ and matrix ‘B’

Feature

Difference

Simple histogram

40

Unser’s feature

11

ISADH feature

4.5

3.3 Training and classification Recently, a unified approach was presented in Rocha et al. (2010) that can combine many features and classifiers. The author approached the multi-class classification problem as a set of binary classification problem in such a way one can assemble together diverse features and classifier approaches custom-tailored to parts of the problem. They define a class binarisation as a mapping of a multi-class problem onto two-class problems (divide-and-conquer) and referred binary classifier as a base learner. For N-class problem N × (N – 1) / 2 binary classifiers will be needed where N is the number of different classes. According to the author, the ijth binary classifier uses the patterns of class i as positive and the patterns of class j as negative. They calculate the minimum distance of the generated vector (binary outcomes) to the binary pattern (ID) representing each class, in order to find the final outcome. The test case will belong to that class for which the distance between ID of that class and binary outcomes will be minimised. Table 5

Unique ID of each class x×y

x×z

y×z

x

+1

+1

0

y

–1

0

+1

z

0

–1

–1

Their approach can be understood by a simple three class problem. Let three classes are x, y, and z. Three binary classifiers consisting of two classes each (i.e., x × y, x × z, and y × z) will be used as base learners, and each binary classifier will be trained with training images. Each class will receive a unique ID as shown in Table 5. To populate the table is

208

S.R. Dubey and A.S. Jalal

straightforward. First, we perform the binary comparison x × y and tag the class x with the outcome +1, the class y with −1 and set the remaining entries in that column to 0. Thereafter, we repeat the procedure comparing x × z, tag the class x with +1, the class z with −1, and the remaining entries in that column with 0. In the last, we repeat this procedure for binary classifier y × z, and tag the class y with +1, the class z with –1, and set the remaining entries with 0 in that column, where the entry 0 means a ‘do not care’ value. Finally, each row represents the unique ID of that class (e.g., y = [–1, 0, +1]). Each binary classifier results a binary response for any input example. Let us say if the outcomes for the binary classifier x (y, x (z, and y (z are +1, –1, and +1 respectively, then the input example will belong to that class which have the minimum distance from the vector [+1, –1, +1]. So the final answer will be given by the minimum distance of min dist ({+1, − 1, + 1}, ({+1, + 1, 0}, {−1, 0, + 1}, {0, − 1, − 1}) )

Recently, Singh et al. (2014) have made a ‘Statistical comparison of classifiers for script identification from multi-script handwritten documents’. Bird and Ko (2013) have combined the diverse classifiers using precision index functions to enhance its discriminative ability. Effectiveness of various similarity measures in classification scenario is tested by Sengupta et al. (2014) and Anzar and Sathidevi (2014) have performed optimal score level fusion by combining multi-normalisation and separability measures and found promising results. We have used MSVM as a set of binary support vector machines (SVMs) for the training and classification. We also used KNN classifier (James and Dimitrijev, 2012; Du and Chen, 2007; Kohavi et al., 1997) for the training and classification purpose in this paper.

4

Experimental result

In this section, we describe the dataset of apple fruit diseases and discuss various issues regarding the performance and efficiency of the system. In Section 4.1, we describe the dataset used in this experiment and highlight several difficulties present in the dataset. In Section 4.2, the performance of proposed ISADH texture feature is presented and compared with other colour and texture feature. In order to show the efficiency of the proposed texture feature, we have compared it with three state-of-the-art features. We also consider and compare the performance of the system in two colour-spaces (i.e., RGB and HSV colour-space). We also compared the performance of ISADH feature for SVM and KNN classifiers.

4.1 Dataset preparation To demonstrate the performance of the proposed approach, we have used a dataset of normal and diseased apple fruits, which comprises four different categories: apple blotch (104), apple rot (107), apple scab (100), and normal apple (80). This dataset is also used in Dubey and Jalal (2012b, 2012c). The total number of apple fruit images (N) is 391. Figure 4 depicts the classes of the dataset. The presence of a lot of variations in the type and colour of apple makes the dataset more realistic.

Fruit disease recognition using improved sum and difference histogram Figure 4

209

Sample images from the dataset of type, (a) apple scab (b) apple rot (c) apple blotch (d) normal apple (see online version for colours)

(a)

(b)

(c)

(d)

4.2 Result discussion In the quest for finding the best categorisation procedure and feature to fruit disease recognition, we have compared the result of ISADH feature with some colour and texture-based image descriptors considering MSVM as classifier. If we use M images per class for training, then remaining N – 4 * M are used for testing and repeat this process ten times for different set of training images. The accuracy of the proposed approach is defined as,

210

S.R. Dubey and A.S. Jalal Accuracy (%) =

Total number of images correctly classified *100 Total number of images used for testing

To evaluate the accuracy of the proposed approach quantitatively, we compare our results with the results of global colour histograms (GCHs), colour coherence vectors (CCV), and Unser’s descriptor. GCH is a set of ordered values, one for each distinct colour, representing the probability of a pixel being of that colour. Uniform quantisation and normalisation are used to reduce the number of distinct colours and to avoid scaling bias (Gonzalez and Woods, 2007). In order to compute the CCV, the method finds the connected components in the image and classify the pixels within a given colour bucket as either coherent or incoherent (Pass et al., 1997). After classifying the image pixels, CCV computes two colour histograms: one for coherent pixels and another for incoherent pixels. The two histograms are stored as a single histogram. (a) Comparison of GCH, CCV, Unser, and proposed ISADH features using SVM as a base learner in RGB colour space (b) Area under curve for each feature in RGB colour space (see online version for colours)

Figure 5

Average Accuracy (%)

100 90 80 70

GCH_RGB CCV_RGB UNSER_RGB ISADH_RGB

60 50

26

31

36

41

46

51

56

Number of Training Images per Disease (%)

61

67

(a) 3700

Area Under Curve

3600 3500 3400 3300 3200 3100 3000

GCH_RGB

CCV_RGB

UNSER_RGB

Features in RGB color Space (b)

ISADH_RGB

Fruit disease recognition using improved sum and difference histogram

211

(a) Comparison of GCH, CCV, Unser, and proposed ISADH features using SVM as a base learner (b) Area under curve for each feature in HSV colour space (see online version for colours)

Figure 6

Average Accracy (%)

100 90 80 70

GCH_HSV CCV_HSV UNSER_HSV ISADH_HSV

60 50

26

31

36

41

46

51

56

Number of Training Images per Disease (%)

61

67

(a) 3700

Area Under Curve

3600 3500 3400 3300 3200 3100 3000

GCH_HSV

CCV_HSV

UNSER_HSV

Features in HSV Color Space

ISADH_HSV

(b)

In order to extract the Unser feature, first the method finds the sum and difference of intensity values over a displacement of (d1, d2) of an image then it calculates two histograms sum and difference histograms and stores both the histograms as a single histogram. In this experiment, we have used a different number of images per class for the training. Figures 5 and 6 show the results for different features in the RGB and HSV colour spaces respectively. The x-axis represents the images per class in the training set (%) and the y-axis represents the accuracy in the testing set (%). This experiment shows that the GCH does not perform well and accuracy for it is lowest in both the colour spaces. One possible explanation is that, GCH feature has only colour information, it does not utilise any neighbouring information in computing the future. CCV and Unser both are performing nearly same in both the colour spaces. Both CCV and Unser features perform better than GCH feature because CCV and Unser features employed the neighbouring information. Since CCV uses frequency of each colour in coherent and incoherent regions separately, so it shows good results as compared to the GCH and Unser. The accuracy for each feature in RGB colour space is shown in Figure 5(a) and its

212

S.R. Dubey and A.S. Jalal

area under the curve (AUC) is plotted in Figure 5(b). AUC for ISADH texture feature is highest with a large margin among all the features which interprets that ISADH is far better than other features. Figure 6(a) depicts the average classification accuracy for the features derived from HSV colour images and AUC for Figure 6(a) is plotted in Figure 6(b). Average classification accuracy and AUC for ISADH is better than other features in HSV colour space also. Figure 7

Performance of proposed ISADH texture feature in RGB and HSV colour space considering SVM and KNN classifier (see online version for colours)

Average Accuracy (%)

100

80

60

20

Table 6

S. no.

ISADH_SVM_RGB ISADH_KNN_RGB ISADH_SVM_HSV ISADH_KNN_HSV

40

26

31

36

41

46

51

56

61

Number of Training Examples per Class (%)

67

Accuracy (%) in RGB colour space when the system is trained with 60 images per class

Type of disease

Average accuracy (%) in RGB colour space when system is trained with 60 images per class GCH

CCV

UNSER

ISADH

1

Apple blotch

82.82

78.88

80.88

85.53

2

Apple scab

76.78

86.89

85.71

93.41

3

Apple rot

80.67

79.67

83.67

94.00

4

Normal apple

98.00

98.00

98.00

99.00

80.82

84.86

82.07

92.99

Average accuracy (%) Table 7

S. no.

Accuracy (%) in HSV colour space when system is trained with 60 images per class

Type of disease

Average accuracy (%) in HSV colour space when system is trained with 60 images per class GCH

CCV

UNSER

ISADH

Apple blotch

83.82

79.41

86.82

93.95

2

Apple scab

71.43

84.13

78.02

97.86

3

Apple rot

70.00

81.67

77.33

96.19

4

Normal apple

100

100

100

100

81.31

86.30

85.54

97.00

1

Average accuracy (%)

Fruit disease recognition using improved sum and difference histogram

213

Figures 5 and 6 illustrate that proposed ISADH texture feature outperforms the other features because ISADH feature uses the neighbouring information of x-direction with the neighbouring information of y-direction. When we use the information of x-direction to calculate the sum and difference in y-direction, then the approach preserves much information in the resulting output feature. For instance, the reported accuracy is 81.34% for GCH feature, 85.88% for Unser feature, 86.36% for CCV feature, and 97.35% for ISADH feature, when MSVM is trained with the 67% images per disease in the HSV colour space. We have also evaluated our proposed ISADH texture feature using k-nearest neighbour (KNN) classifier, as shown in Figure 7. The classification accuracy of ISADH feature is better for SVM classifier as compared to KNN classifier. Tables 6 and 7 show the accuracy (%) of fruit disease recognition problem when features are extracted from RGB and HSV colour images respectively and training is done with 61% images per class. From these tables, it is clear that ISADH feature shows a better result for RGB and HSV colour images nearly for each disease where other features fail to produce good results. Comparison of the accuracy (%) in RGB and HSV colour space for the GCH, CCV, Unser, and ISADH features (see online version for colours)

100

Accuracy (%)

90 80 70 GCH_RGB GCH_HSV

60 50

26

31 36 41 46 51 56 61 Number of Training Images per Disease (%)

67

100 90 Accuracy (%)

Figure 8

80 70 CCV_RGB CCV_HSV

60 50

26

31 36 41 46 51 56 61 Number of Training Images per Disease (%)

67

214 Figure 8

S.R. Dubey and A.S. Jalal Comparison of the accuracy (%) in RGB and HSV colour space for the GCH, CCV, Unser, and ISADH features (continued) (see online version for colours)

100

Accuracy (%)

90 80 70 Unser_RGB Unser_HSV

60 50

26

31 36 41 46 51 56 61 Number of Training Images per Disease (%)

67

100

Accuracy (%)

90 80 70 ISADH_RGB ISADH_HSV

60 50

26

31 36 41 46 51 56 61 Number of Training Images per Disease (%)

67

We have also observed across the plots that each feature performs better in the HSV colour space than the RGB colour space as shown in the Figure 8. For ISADH feature with 61% training examples per disease, the reported classification accuracy is 92.98% in RGB and 97.00% in HSV colour space. One important aspect when dealing with fruit disease classification is the accuracy per class. This information points out the classes that need more attention when solving the confusions. Figure 9 depicts the accuracy for each one of four classes of apple fruit using ISADH texture feature in RGB and HSV colour spaces. The graph in Figure 9 clearly illustrates that the class Apple Blotch needs more attention as it yields the lowest accuracy when compared to other classes in both RGB and HSV colour spaces because the defected area in the apple blotch disease is very small and so it is very difficult to segment these areas from the image precisely. Figure 9 also depicts that normal apples are very easily distinguishable with diseased apples and a very good classification result is achieved for the normal apples in both colour spaces. For ISADH feature in HSV colour space, for instance, reported classification accuracy are 94.06%, 98.00%, 97.33%, and 100% for the apple blotch, apple rot, apple scab, and normal apple respectively, resulting an average accuracy nearly 97.35% when training is done with 67% images per class using the MSVM as a classifier. Among apple rot and apple scab, the accuracy is better for the apple scab generally.

Fruit disease recognition using improved sum and difference histogram Figure 9

215

Accuracy per class (i.e., apple blotch, apple rot, apple scab and normal apples) for the ISADH feature in RGB and HSV colour space, (a) ISADH in RGB colour space (b) ISADH in HSV colour space (see online version for colours)

Accuracy per Disease (%)

100 90 80 Apple Blotch Apple Rot Apple Scab Normal Apple

70 60 50

26

31 36 41 46 51 56 61 Number of Training Images per Disease (%)

67

(a)

Accuracy per Disease (%)

100 90 80 70

Apple Blotch Apple Rot Apple Scab Normal Apple

60 50

26

31 36 41 46 51 56 61 Number of Training Images per Disease (%)

67

(b)

4.3 Effect of gradient filters James (2013) has designed a gradient filter-based image description framework. We also used our method in conjunction with the gradient filters. All-pass, average, horizontal-vertical difference, diagonal difference, horizontal sobel edge operator, and vertical sobel edge operator are the six gradient filters suggested by James are also used in this paper and the effect of these filters over the defected cluster of Figure 10(b) are shown in Figures 10(c)–10(f) respectively. We extracted the ISADH feature over each filtered images and finally added them to find a single ISADH with gradient filters (GFs_ISADH) feature description. We extracted the GFs_ISADH in both RGB (GFs_ISADH_RGB) and HSV (GFs_ISADH_HSV) colour spaces and compared its performance with the ISADH_RGB and ISADH_HSV feature description in Figure 11. It is evident from Figure 11 that gradient filters boost the discriminative ability of the

216

S.R. Dubey and A.S. Jalal

ISADH feature as we are able to gain the classification accuracy upto 99.9% using GFs_ISADH_RGB feature when 67% images are used for the training purpose. Figure 10 Effect of gradient filters, (a) original image (b) detected diseased area, filtered diseased area with gradient filters (James, 2013) (c) all-pass (d) average (e) horizontal-vertical difference (f) diagonal difference (g) horizontal sobel edge operator (h) vertical sobel edge operator (see online version for colours)

(a)

(c)

(d)

(b)

(e)

(f)

(g)

(h)

Figure 11 Accuracy per class (i.e., apple blotch, apple rot, apple scab and normal apples) for the ISADH feature in RGB and HSV colour space (see online version for colours)

Classification Accuracy (%)

100 95 90 85 80 75 70

5

ISADH_RGB GFs_ISADH_RGB ISADH_HSV GFs_ISADH_HSV 26

31

36

41

46

51

56

61

Number of Training Examples per Class (%)

67

Conclusions and future work

An image processing-based solution is proposed for the automatic detection and recognition of fruit diseases from images using colour and texture features. The proposed approach is composed of three steps: in the first step defect segmentation is carried out using K-means clustering-based image segmentation method, in the second step features are extracted from the segmented image, and finally in the third step images are classified

Fruit disease recognition using improved sum and difference histogram

217

into one of the diseases using MSVM as a classifier which is trained with the same features of training images of each type of diseases. This paper also presented an efficient ISADH texture feature from the intensity values of the neighbouring pixels. We also incorporated our method with the gradient filters to enhance ISADH. We have taken apple diseases as a case study and evaluated our approach. Three types of apple diseases, namely: blotch, rot, and scab are used for the experiments. Based on the experiments, we have reported more than 97% average classification accuracy and concluded that proposed ISADH texture feature outperforms other colour and texture features. It is also observed that ISADH performs better when it is applied with several gradient filters and achieved nearly 99.9% classification accuracy. Further work includes consideration of multiple features at a time and to develop a segmentation approach that can segment disease from the image more precisely to improve the final outcome of the introduced approach.

References Anzar, S.M. and Sathidevi, P.S. (2014) ‘Optimal score level fusion combining multi-normalization and separability measures’, International Journal of Applied Pattern Recognition, under press. Bird, J. and Ko, D. (2013) ‘Combining diverse classifiers using precision index functions’, International Journal of Applied Pattern Recognition, Vol. 1, No. 1, pp.3–26. Bravo, C., Moshou, D., Oberti, R., West, J., McCartney, A., Bodria, L. and Ramon, H. (2004) ‘Foliar disease detection in the field using optical sensor fusion. Agricultural engineering’, International: The CIGR Journal of Scientific Research and Development, Vol. 6, No. 1, pp.1–14. Chen, B., Wang, K., Li, S., Wang, J., Bai, J., Xiao, C. and Lai, J. (2008) ‘Spectrum characteristics of cotton canopy infected with verticillium wilt and inversion of severity level’, Proceedings of the Computer and Computing Technologies in Agriculture, Vol. 2, No. 1, pp.1169–1180. Choi, Y.H., Tapias, E.C., Kim, H.K., Lefeber, A.W.M., Erkelens, C., Verhoeven, J.T.J., Brzin, J., Zel, J. and Verpoorte, R. (2004) ‘Metabolic discrimination of catharanthus roseus leaves infected by phytoplasma using 1H-NMR spectroscopy and multivariate data analysis’, Plant Physiology, Vol. 135, No. 1, pp.2398–2410. Cover, T. and Hart, P. (1967) ‘Nearest neighbor pattern classification’, IEEE Transactions on Information Theory, Vol. 13, No. 1, pp.21–27. Du, H. and Chen, Y.Q. (2007) ‘Rectified nearest feature line segment for pattern classification’, Pattern Recognition, Vol. 40, No. 5, pp.1486–1497. Dubey, S.R. (2012) Automatic Recognition of Fruits and Vegetables and Detection of Fruit Diseases, MTech Thesis. Dubey, S.R. and Jalal, A.S. (2012a) ‘Robust approach for fruit and vegetable classification’, Elsevier Procedia Engineering, Vol. 38, No. 1, pp.3449–3453. Dubey, S.R. and Jalal, A.S. (2012b) ‘Adapted approach for fruit disease identification using images’, International Journal of Computer Vision and Image Processing, Vol. 2, No. 3, pp.44–58. Dubey, S.R. and Jalal, A.S. (2012c) ‘Detection and classification of apple fruit diseases using complete local binary patterns’, Proceedings of the Third IEEE International Conference on Computer and Communication Technology, pp.346–351. Dubey, S.R. and Jalal, A.S. (2013) ‘Species and variety detection of fruits and vegetables from images’, International Journal of Applied Pattern Recognition, Vol. 1, No. 1, pp.108–126. Dubey, S.R. and Jalal, A.S. (2014) ‘Automatic fruit disease classification using images’, Computer Vision and Image Processing in Intelligent Systems and Multimedia Technologies, pp.82–100, IGI-Global.

218

S.R. Dubey and A.S. Jalal

Dubey, S.R., Dixit, P., Singh, N. and Gupta, J.P. (2013) ‘Infected fruit part detection using K-means clustering segmentation technique’, International Journal of Artificial Intelligence and Interactive Multimedia, Vol. 2, No. 2, pp.65–72. Fukunaga, K. and Hummels, D.M. (1987) ‘Bayes error estimation using Parzen and k-NN procedures’, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 9, No. 5, pp.634–643. Gonzalez, R. and Woods, R. (2007) Digital Image Processing, 3rd ed., Prentice-Hall, Upper Saddle River, NJ, USA. Gupta, J.P., Singh, N., Dixit, P., Semwal, V.B. and Dubey, S.R. (2013) ‘Human activity recognition using gait pattern’, International Journal of Computer Vision and Image Processing (IJCVIP), Vol. 3, No. 3, pp.31–53. Hahn, F. (2009) ‘Actual pathogen detection: sensors and algorithms – a review’, Algorithms, Vol. 2, No. 1, pp.301–338. Hartigan, J.A. and Wong, M.A. (1979) ‘Algorithm AS 136: a K-means clustering algorithm’, Journal of the Royal Statistical Society, Series C (Applied Statistics), Vol. 28, No. 1, pp.100–108. Hartman, J. (2010) ‘Apple fruit diseases appearing at harvest’, Plant Pathology Fact Sheet, College of Agriculture, University of Kentucky [online] http://www. ca. uky. edu/agcollege/plantpathology/ext_files/PP FShtml/PPFS-FR. James, A.P. (2013) ‘One-sample face recognition with local similarity decisions’, International Journal of Applied Pattern Recognition, Vol. 1, No. 1, pp.61–80. James, A.P. and Dimitrijev, S. (2010) ‘Inter-image outliers and their application to image classification’, Pattern Recognition, Vol. 43, No. 12, pp.4101–4112. James, A.P. and Dimitrijev, S. (2012) ‘Nearest neighbor classifier based on nearest feature decisions’, The Computer Journal, Vol. 55, No. 9, pp.1072–1087. Kim, M.S., Lefcourt, A.M., Chen, Y.R. and Tao, Y. (2005) ‘Automated detection of fecal contamination of apples based on multispectral fluorescence image fusion’, Journal of Food Engineering, Vol. 71, No. 1, pp.85–91. Kleynen, O., Leemans, V. and Destain, M.F. (2005) ‘Development of a multi-spectral vision system for the detection of defects on apples’, Journal of Food Engineering, Vol. 69, No. 1, pp.41–49. Kohavi, R., Langley, P. and Yun, Y. (1997) ‘The utility of feature weighting in nearest-neighbor algorithms’, Proceedings of the Ninth European Conference on Machine Learning, pp.85–92. Kumar, K. and Sheeba, V.S. (2014) ‘A blind watermarking algorithm for fingerprint images based on contourlet transform’, International Journal of Applied Pattern Recognition, under press. Kumar, K.S., Prasad, S., Barnwal, S. and Semwal, V.B. (2010) ‘Sports video summarization using priority curve algorithm’, International Journal on Computer Science & Engineering, Vol. 2, No. 9, pp.2996–3002. Kumar, K.S., Prasad, S., Semwal, V.B. and Tripathi, R.C. (2011a) ‘Real time face recognition using adaboost improved fast PCA algorithm’, International Journal of Artificial Intelligence & Applications, Vol. 2, No. 3, pp.45–58. Kumar, K.S., Semwal, V.B., Prasad, S. and Tripathi, R.C. (2011b) ‘Generating 3D model using 2D images of an object’, International Journal of Engineering Science and Technology, Vol. 3, No. 1, pp.406–415. Leemans, V., Magein, H. and Destain, M.F. (1998) ‘Defect segmentation on ‘golden delicious’ apples by using colour machine vision’, Computers and Electronics in Agriculture, Vol. 20, No. 1, pp.117–130. Leemans, V., Magein, H. and Destain, M.F. (1999) ‘Defect segmentation on ‘jonagold’ apples using colour vision and a bayesian classification method’, Computers and Electronics in Agriculture, Vol. 23, No. 1, pp.43–53. Li, Q., Wang, M. and Gu, W. (2002) ‘Computer vision based system for apple surface defect detection’, Computers and Electronics in Agriculture, Vol. 36, Nos. 2–3, pp.215–223.

Fruit disease recognition using improved sum and difference histogram

219

Li, X. (2014) ‘Comparative analysis on SIFT features in visible and infrared aerial imaging’, International Journal of Applied Pattern Recognition, under press. MacQueen, J. (1967) ‘Some methods for classification and analysis of multivariate observations’, Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability, pp.281–297. Marcassa, L.G., Gasparoto, M.C.G., Belasque, J., Lins, E.C., Nunes, F.D. and Bagnato, V.S. (2006) ‘Fluorescence spectroscopy applied to orange trees’, Laser Physics, Vol. 16, No. 5, pp.884–888. Mehl, P.M., Chao, K., Kim, M. and Chen, Y.R. (2002) ‘Detection of defects on selected apple cultivars using hyperspectral and multispectral image analysis’, Applied Engineering in Agriculture, Vol. 18, No. 2, pp.219–226. Moshou, D., Bravo, C., Oberti, R., West, J., Bodria, L., McCartney, A. and Ramon, H. (2005) ‘Plant disease detection based on data fusion of hyper-spectral and multi-spectral fluorescence imaging using Kohonen maps’, Real-Time Imaging, Vol. 11, No. 2, pp.75–83. Moshou, D., Bravo, C., Wahlen, S., West, J., McCartney, A., De, J., Baerdemaeker, J.D. and Ramon, H. (2006) ‘Simultaneous identification of plant stresses and diseases in arable crops using proximal optical sensing and self-organising maps’, Precision Agriculture, Vol. 7, No. 3, pp.149–164. Pass, G., Zabih, R. and Miller, J. (1997) ‘Comparing images using color coherence vectors’, Proceedings of the Fourth ACM International Conference on Multimedia, New York, USA, pp.65–73. Purcell, D.E., O’Shea, M.G., Johnson, R.A. and Kokot, S. (2009) ‘Near-infrared spectroscopy for the prediction of disease rating for Fiji leaf gall in sugarcane clones’, Applied Spectroscopy, Vol. 63, No. 4, pp.450–457. Qin, J., Burks, F., Ritenour, M.A. and Bonn, W.G. (2009) ‘Detection of citrus canker using hyper-spectral reflectance imaging with spectral information divergence’, Journal of Food Engineering, Vol. 93, No. 2, pp.183–191. Robbins, M., Schimmelpfennig, D., Ashley, E., Livingston, M., Ash, M. and Vasavada, U. (2006) The Value of Plant Disease Early-Warning Systems: A Case Study of USDA’s Soybean Rust Coordinated Framework, Economic Research Report No. (ERR-18). Rocha, A., Hauagge, C., Wainer, J. and Siome, D. (2010) ‘Automatic fruit and vegetable classification from images’, Computers and Electronics in Agriculture, Vol. 70, No. 1, pp.96–104. Sankarana, S., Mishraa, A., Ehsania, R. and Davisb, C. (2010) ‘A review of advanced techniques for detecting plant diseases’, Computers and Electronics in Agriculture, Vol. 72, No. 1, pp.1–13. Sengupta, S., Ojha, P., Wang, H. and Blackburn, W. (2014) ‘Effectiveness of similarity measures in classification of time series data with intrinsic and extrinsic variability’, International Journal of Applied Pattern Recognition, under press. Shafri, H.Z.M. and Hamdan, N. (2009) ‘Hyperspectral imagery for mapping disease infection in oil palm plantation using vegetation indices and red edge techniques’, American Journal of Applied Sciences, Vol. 6, No. 6, pp.1031–1035. Singh, N., Dubey, S.R., Dixit, P. and Gupta, J.P. (2012) ‘Semantic image retrieval by combining color, texture and shape features’, Proceedings of the International Conference on Computing Sciences (ICCS), pp.116–120. Singh, P., Sarkar, R., Das, N., Basu, S. and Nasipuri, M. (2014) ‘Statistical comparison of classifiers for script identification from multi-script handwritten documents’, International Journal of Applied Pattern Recognition, under press. Spinelli, F., Noferini, M. and Costa, G. (2006) ‘Near infrared spectroscopy (NIRs): perspective of fire blight detection in asymptomatic plant material’, Proceedings of the 10th International Workshop on Fire Blight, Acta Horticulturae, pp.87–90.

220

S.R. Dubey and A.S. Jalal

Unser, M. (1986) ‘Sum and difference histograms for texture classification’, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 8, No. 1, pp.118–125. Varshney, P., Farooq, O. and Upadhyaya, P. (2014) ‘Hindi Viseme recognition using subspace DCT features’, International Journal of Applied Pattern Recognition, under press. Yang, C.M., Cheng, C.H. and Chen, R.K. (2007) ‘Changes in spectral characteristics of rice canopy infested with brown planthopper and leaffolder’, Crop Science, Vol. 47, No. 1, pp.329–335.

Fruit disease recognition using improved sum and ...

A fungal disease, apple blotch appears on the surface of the fruit as dark, irregular or lobed edges. The precise segmentation is required for the defect detection.

597KB Sizes 7 Downloads 194 Views

Recommend Documents

moving object recognition using improved rmi method - CiteSeerX
e-mail: [email protected], [email protected] ... framework for use in development of automated video surveillance systems. RMI is a specific ...

moving object recognition using improved rmi method - CiteSeerX
plays a major role in advanced security systems and video surveillance applications. With the aim to recognize moving objects through video monitoring system, ...

Fruit and vegetable recognition by fusing color and ...
background subtraction, (2) feature extraction and (3) training and classification. K-means ... Anand Singh Jalal received his MTech Degree in Computer Science from Devi ... Computer Vision from Indian Institute of Information Technology (IIIT),. All

High yielding and moderately resistant to fruit rot disease ... - CiteSeerX
Among them the hybrid combination of Sln 1 x CA 97 was identified as best .... (3.27 per cent) and PDI of the top two best .... Host resistance to anthracnose of.

extraction and characterization of vegetable oil using bread fruit seed ...
etc.) visit our website for more details. Page 1 of 1. extraction and characterization of vegetable oil using bread fruit seed.pdf. extraction and characterization of ...

Speaker Recognition using Kernel-PCA and ...
Modeling in common speaker subspace (CSS). The purpose of the projection of the common-speaker subspace into Rm using the distance preserving ...

SPEAKER-TRAINED RECOGNITION USING ... - Vincent Vanhoucke
advantages of this approach include improved performance and portability of the ... tion rate of both clash and consistency testing has to be minimized, while ensuring that .... practical application using STR in a speaker-independent context,.

Efficient Speaker Recognition Using Approximated ...
metric model (a GMM) to the target training data and computing the average .... using maximum a posteriori (MAP) adaptation with a universal background ...

IC_55.Dysarthric Speech Recognition Using Kullback-Leibler ...
IC_55.Dysarthric Speech Recognition Using Kullback-Leibler Divergence-based Hidden Markov Model.pdf. IC_55.Dysarthric Speech Recognition Using Kullback-Leibler Divergence-based Hidden Markov Model.pdf. Open. Extract. Open with. Sign In. Main menu.

Rapid Face Recognition Using Hashing
cal analysis on the recognition rate of the proposed hashing approach. Experiments ... of the images as well as the large number of training data. Typically, face ...

Activity Recognition Using a Combination of ... - ee.washington.edu
Aug 29, 2008 - work was supported in part by the Army Research Office under PECASE Grant. W911NF-05-1-0491 and MURI Grant W 911 NF 0710287. This paper was ... Z. Zhang is with Microsoft Research, Microsoft Corporation, Redmond, WA. 98052 USA (e-mail:

Customized Cognitive State Recognition Using ... - Semantic Scholar
training examples that allow the algorithms to be tailored for each user. We propose a ..... to the user-specific training example database. The TL module is used ...

Face Recognition in Surgically Altered Faces Using Optimization and ...
translation and scale invariance [3]. Russell C.Eberhart (2006) emphasis that the ... Each particle in the search space evolves its candidate solution over time, making use of its individual memory and knowledge gained by the swarm as a ... exchange

Speech Recognition Using FPGA Technology
Figure 1: Two-line I2C bus protocol for the Wolfson ... Speech recognition is becoming increasingly popular and can be found in luxury cars, mobile phones,.

SPEAKER-TRAINED RECOGNITION USING ... - Vincent Vanhoucke
approach has been evaluated on an over-the-telephone, voice-ac- tivated dialing task and ... ments over techniques based on context-independent phone mod-.

Texture recognition by using GLCM and various ...
Apr 4, 2010 - 3125, Australia (phone: +613 9251 74754; email: [email protected], ... calculating, for each cell, the occurrences of each grey-level combination ..... [7] T. Calvo, G. Mayor, and R. Mesiar, Eds., Aggregation Operators. New.

Automatic speaker recognition using dynamic Bayesian network ...
This paper presents a novel approach to automatic speaker recognition using dynamic Bayesian network (DBN). DBNs have a precise and well-understand ...

Rapid Face Recognition Using Hashing
cal analysis on the recognition rate of the proposed hashing approach. Experiments ... of the images as well as the large number of training data. Typically, face ...

offline handwritten word recognition using a hybrid neural network and ...
network (NN) and Hidden Markov models (HMM) for solving handwritten word recognition problem. The pre- processing involves generating a segmentation ...

Speaker Recognition using Kernel-PCA and ... - Semantic Scholar
[11] D. A. Reynolds, T. F. Quatieri and R. B. Dunn, "Speaker verification using adapted Gaussian mixture models,". Digital Signal Processing, Vol. 10, No.1-3, pp.

Speech Recognition Using FPGA Technology
Department of Electrical Computer and Software Engineering ..... FSM is created to implement the bus interface between the FPGA and the Wolfson. Note that ...