Detection And Tracking Of The Moving Object Based On The SoC System

With the development of technology of imaging and image processing, people can obtain richer and richer information from image, which makes technology of image processing widely used in various areas of daily life and manufacture. However, image contains a great amount of information. The image processing algorithm that used to analyze the imaging data is more and more complex. So the technology of image processing has to face many challenging issues. The detection and tracking of moving object is one of the important research a reas in computer vision. Due to the highly requirements of real-time application of imaging system, the image processing speed is the most critical factor. FPGA-based programmable SOC has many advantages such as small size, low power consumption, high spee d, etc. It has broad application prospects in image processing.In this thesis, the image processing base on SoC can achieve the same result comparing to the processing on personal computer, which validates the feasibility of image processing base on hardware. This thesis uses product of Zynq-7000 series Introduced by Xilinx as an image processor, building image processing system by So C, the system is mainly composed by modules of image acquisition, image processing, image buffering, and image displaying. In the design of image processing module, we use pipeline architecture to optimize the image processing algorithm to improve the performance of the image processing module, package the imaging processing module into IP core which conforms with AXI protocol, pre-set corresponding register about the rows and columns of image data, data width and other parameters, then the complete design of image processing module can be easily reused.On the basis of complete system of image acquisition and processing, the system can detect and track the moving object with specific color, and keep the moving objects at the center of the capture frame. The algorithms of tracking a moving object are mainly completed in image processing module. We choose color as the main feature of the moving object, then using a threshold value to extract the moving object in the HSV color space, keeping updating the detecting color template, and determining the moving object by morphological processing. In our system, we see the centroid as the representation of a moving object. In order to reduce the interference of mistaken identified pixels for centroid calculation, we give weight to the recognized value of the pixel; the farther from the current centroid pixel the less weight will be given. In order to keep the moving object always stays at the center of the capture frame, we calculate the difference between the target’s centroid and center of the screen coordinates. With the location difference and position feeding back from servo motors, we use the PD controller to generate the PWM wave to control the pan and tilt angle to make sure the difference is always minimized.