Outdoor Video Surveillance System

Members:

Harish Guruprasad.R Kishore Kumar.S Manikandan.R.R Lokesh Prabhu Allwin Amalraj.A Krishna.V Narendhiran.C.D Vijaya Kumar.S

College:

College of Engineering, Guindy, Anna University, Chennai

Guide:

Dr.P.V.Ramakrishna, Professor, Dept of ECE, CEG, Anna University

CONTENTS

1. Introduction 2. Air Vehicle Detailed Description 3. Image Sensor 4. Video RF Transmitter 5. Onboard Payload DC Power Estimation 6. Ground Based RF Video Receiver 7. Data processing 1. Image stabilization 2. Mosaicing 3. Independent motion detection 4. Human detection 8. Cost Details 9. Additional options to be attempted 1. Digital video and compression 2. Error control coding 3. Link Budget Calculations 4. Hardware for Digital Video Transmission/Reception 10. Proposed Work Plan 11. References 12. Figures

1. Introduction Our main objective here is to build a working prototype system a capable of providing a sense of situational awareness for the end user using video surveillance. The main issues for the construction of such a system are airworthiness, ease control, quality of video, cost and ease of construction. The proposed solution consists of starting with a commercially available mini RC aircraft build/assemble/attach the necessary electronics and video camera. The basic system just transmits the acquired video to a base station where most of the processing is carried out. While the basic system to build will meet the requirements spelt out for the project, we also propose to attempt a number of improvements over this basic system to enhance the control and video quality. The payload part of the system and the ground segments briefly given in the form of block diagrams in Fig.1. and Fig.2 below.

The main components shown in Fig.1 for the onboard system are all easily available low cost components (COTS components), designed especially for RC planes. The camera is a low power, light weight, analog output system capable of either NTSC or PAL output. The output of the camera is fed to the transmitter which is also a readily available module capable of directly interfacing to the camera.

The ground station carries the task of recovering the video signal and provides the display. The ground based video receiver shown in Fig.2. receives the FM modulated analog video, demodulates it and gives its output in the NTSC/PAL format. The analog video converter block shown in Fig.2. converts the standard NTSC/PAL into USB for viewing in the laptop. The 1|P age

laptop is used for implementing all the video signal processing algorithms and provides the display in real time. 2. Air Vehicle Detailed Description: We have looked at various airborne platforms like balloons, aircrafts and helicopters. We have chosen a RC plane because of its simplicity, availability and maneuverability. The main constraints in selecting an RC plane are the weight and dimensions of the payload to be carried. The proposed basic system just adds a camera, a video transmitter and batteries for the camera and transmitter to the RC aircraft. The engine throttle is controlled by a servo motor. An independently operated four channel receiver set with 4 servo motors and connectors is used for the control of the aircraft. The Futaba make R/C unit weighs about 560gms. The pitch, roll and yaw of the craft are controlled by means of the rudder, elevator and aileron. A throttle control will also be provided. Table 1. gives all the relevant technical details of the air vehicle system chosen. Table 1: Selected Aircraft Details Aircraft Weight Estimation Servos Engine and battery Electronic Systems (Payload) Other linkages Total Weight Aircraft Dimension Details Wing Span Fuselage Length Biplane effective air chord Wing loading Aircraft Propeller Details RPM Work done Aircraft Battery Specifications Battery Type Voltage No of Cells Time of flight

60 g 300 g 100 g 100 g 560 g (5.6N) 825 mm 670 mm 200 mm 22.5 g/dmxdm 15000 30 W NiMh 13.5V, each 1.7V 8 7.2 min at a max

3. Image Sensor: We have selected Panasonic make KX141 High Resolution Color CCD Camera

because of its light weight, small size, wide usage and many compatible transceivers. This camera is also available easily and also reported to have been used in many low cost MAV missions/examples. 2|P age

Table 2: Camera Details Camera Details Resolution Sensitivity Standard lens Weight Supply Output format

480 lines 2lux 2.8mm (approx 90o) 12.9 g 5V DC NTSC/PAL

At the resolution indicated above, the maximum area that can be covered clearly is about 30m*30m. So at a height of 30 metres this requires an angle of view of 45 degrees. If necessary, additional optics will be designed and retrofitted onto this camera. 4. Video RF Transmitter: The video transmitter to be used by us is the FlexWAV® RF

transmitter from Black Widow AV. This model has again been chosen based on the size, ease of availability and its reported history of use in similar applications. It takes the direct NTSC/PAL video from the camera and transmits it using frequency modulation an RF carrier. The carrier frequency is about 2.4 GHz and takes a bandwidth of nearly 5 MHz. It operates out of a standard 5V supply which will be provided by an onboard regulator. The relevant technical details for the transmitter re provided in the Table 3. given below Table 3: Onboard Transmitter details Onboard Transmitter details Output Power 200mW Modulation FM Channels 4 (2.41GHz, 2.43GHz, 2.45GHz, 2.47GHz) Frequency used 2.4GHz ISM Band Maximum range achievable 1.5 km Output format PAL or NTSC video Antenna 8dBi Circular polarized patch antenna Transmitter Weight 12.5 grams bare (25g including antenna) Transmitter size 27mm x 24mm x 9mm

5. Onboard Payload DC Power Estimation: Since the battery for air vehicle is a stand- alone and proven part specifically for air vehicle propulsion, a separate battery is used for the payload and the sizing of this battery is explained in this section. We have made an estimate of 3|P age

the payload components power consumption and calculated the energy required from the battery. We have also chosen an appropriate easily available, low cost battery for our needs. We propose to use two 3.7 V Lithium ion polymer batteries for our purpose. In Table .3. given below, we first give the power budget of the payload components and then provide the details of the battery chosen Table 4: Power consumption of onboard payload components and specs Power consumption of onboard payload components and specs Camera power 2W Tx DC power 1.2 W On time 5 minutes Energy required 0.1 + 0.2 WH Battery voltage 5V Current rating required 70mAH (For Transmitter and camera) Battery Weight 10*2 =20 g Battery Energy 240 mAH

6. Ground Based RF Video Receiver: The receiver is the video receiver from Black Widow AV Solutions, is commonly encountered in the MAV (Micro Aerial Vehicle) literature. A circular patch antenna from Black Widow AV is connected to the receiver. The output from the RF Video Receiver is in standard NTSC/PAL format. To view the video on a Laptop we use Pollin R1 USB converter which interfaces directly to the Black Widow receiver on the one side and directly the USB port of a Laptop on the other side. The receiver operates from a DC voltage of 12V and consumes a power of 350mA. 7. Data Processing: Once the Laptop receives the input through the USB port as described above, it carries out all the image processing tasks and also serves as a display unit. The image processing to be implemented in the ground station is described systematically in the sections below: 7.1 Image stabilization: The video is taken from an onboard camera without any damping mechanism. So an image stabilization module is required to remove the high frequency components in the motion vector of the pixels caused due to vibration. 7.2 Mosaicing: As the view from the camera is usually narrow, the viewer at the ground station is unable to make sense of the video stream. So to solve this problem, the present frame can be concatenated with the images received from previous frames to form a mosaic. The method specified in [4] will be used for this purpose. The viewer will then be able to see the events along with the bigger picture. Assuming a flat surface the image can be ortho-rectified to form an approximate map.

4|P age

7.3 Independent motion detection: The independent motion estimation technique to be used by us is detailed in [1]. The objective of this technique is to identify the independently moving objects and separate them from camera motion. It is a useful step for screening purposes. The motion detection algorithm to be implemented will be based on the flow chart shown below. Capture 3 frames

Find domain plane homography using RANSAC and mark inlier points

If inlier points occupy large area

Mark all outliers as motion

Compute epipole using trilinear constraint

Separate 3D from motion by parallax direct estimation

Mark motion pixels Fig.3 Flowchart for Independent motion detection algorithm

7.4 Human detection: There are many well established procedures that are described in the literature. We propose to use the method described in [3] and can be briefly described as follows. One uses a cascade of Haar features and a cascade of covariance features for detecting humans in static images. The objective of finding the location of humans in the image is carried out by giving sub images at various scales and locations to the classifier cascade. Each subimage is classified as either human or non human based on features learnt during the training phase. In the training phase several samples of human images at different poses are given. A cascade of classifiers is found using AdaBoost. The procedure to be followed is given in the form of a flow chart in the fig. below

5|P age

Grab frame

Scan window across image

Give each sub-image as input to classifier stage

If output is not true

If output is true mark region as human

Fig.4 Flowchart for testing phase

8. Cost Details: The basic system proposed by us uses only commercially available low cost components. Their approximate cost is given below. Table 5: Approximate Cost Estimation Approximate Cost Estimation Camera RC plane with controls Video Transmitter & Receiver Modules

Rs. 10000 Rs. 20000 Rs. 7500

9. Additional options to be attempted: While the main goal to be pursued is to demonstrate the operation of a minimum system as described above, once this minimum system is operational, many enhancements which will be treated as optional will be attempted. Each of these options to be attempted is described below and is expected to enhance the capabilities of the minimum system described above. 9.1. Digital video and compression: The usage of a digital camera with a trigger input gives us the option of much higher resolution and greater flexibility. We have chosen to use the Flea2® camera by Point Grey due to its small size. It measures just 3*3*3 cms and weighs 60 6|P age

gms. To enable compression, a high speed and low power onboard processor is required and will be chosen from one of the state of the art low power multimedia DSPs. A high speed processor capable of taking a stream of IEEE 1394 video as input is required for MPEG compression of the video. If possible the independent motion detection can also be implemented here so that the processor can preferentially encode pixels with motion. 9.2. Error control coding Normal convolutional coding will be implemented on the data and will further be extended to include turbo coding. While convolutional code alone can reduce the transmit power by about 3dB, Turbo coding can provide a reduction of about 8dB minimum in the transmit power. Error control coding does not increase the onboard complexity and hence can be easily attempted. 9.3. Link Budget Calculations: Noise Floor

= = Receiver Noise Floor = = SNR =

kT* BW -109.2 dBm -109.2 dBm + 15 dB guard -94.2 dBm

Receiver Sensitivity

= = =

7.44 dB Receiver Noise Floor + SNR -86.76 dBm

Propagation Loss

=

20

Required Tx Power

= = =

112.1 dB Rx Sensitivity –Tx Antenna Gain –Rx Antenna Gain +Prop loss 342 mW

Assumptions: (Worst case parameters) Bandwidth=3MHz; Freq of operation=2.4GHz; Temperature=20 0C ; Range=1 Km ; Eb/N0 =9.2 dB for BPSK modulation Thus, we need a Transmitter whose transmit power is above 25.34 dBm (342 mW). So the transmitter of about 500mW transmit power is used. 9.4. Hardware for Digital Video Transmission/Reception: The output of the Digital camera is compressed and BPSK modulated using FPGA/Processor and a mixer followed by up conversion to RF frequency 2.4GHz ISM Band using IC TRF1121 and TRF1122. The RF signal is then power amplified using IC TRF1123 and fed to an antenna. The sizes of these devices are all within the limits imposed by on the payload.

7|P age

The received signal at the ground station is down-converted, digitized and then dumped onto a laptop where all the digital demodulation and further processing like motion and human detection are carried out.

10. Proposed Work Plan:

8|P age

11. References: The present proposal is partly based on the material presented on the references given below. [1] H. S. Sawhney, Y. Guo, and R. Kumar, “Independent motion detection in 3D scenes,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 22, pp. 1191–1199, Oct. 2000. [2] B. D. Lucas and T. Kanade, “An iterative technique of image registration and its application to stereo,” in Proc. 7th Int. Joint Conf. Artificial Intelligence, 1981, pp. 674–679 [3] Sakrapee Paisitkriangkrai, Chunhua Shen, and Jian ZhangFast Human Detection Using a Cascade of BoostedCovariance Features. IEEE transactions on circuits and systems, Aug 2008 [4]Rakesh Kumar et al, Aerial video surveillance and exploitation, IEEE invited paper 2001. [5] http://www.ptgrey.com/products/compare.asp [6] Small UAV Command, Control and Communication Issues, Dr Joseph Barnard Barnard Communications, IEEE conference on UAVs, 2007 Product references:

1. Black widow AV kit. 2. Texas Instruments.

9|P age

12. Figures Airplane Designs

10 | P a g e

11 | P a g e