Research And Design Of Photoelectric Imaging Target Tracking System

Study of moving small target detection and tracking is one of the hot researchareas in the domain of computer vision. Captured small target image in a complexbackground has the characteristics of low signal-to-noise ratio and low contrast,therefore preprocessing before detection is needed. Target detection is the process offinding and extracting target from the image sequence. After target detection,efficiently tracking moving target in each frame is needed in the image sequence. Onthe basis of reading relevant literature, image preprocessing, target detection andtarget tracking algorithms have been researched and compared, image pre-processingand target detection are the research focuses, in addition, the software part ofphotoelectric imaging target tracking system has been completed, it is performanceverification platform of different algorithms.Effective background suppression is the premise of target detection and relates tothe performance of the system. At the beginning, several typical backgroundsuppressing algorithms is recommended, the background suppressing method basedon correlation between adjacent pixel block is discussed in detail and deeply. In theexperiment real images are processed by the proposed algorithm in this paper andother methods, image processing results are compared, and the performance ofdifferent algorithms are evaluated by signal-to-noise ratio and signal-to-noise ratiogain, experiment results indicate that the proposed algorithm in this paper canefficiently suppress the complex background.Several frequently-used methods of target detection based on single frame imageare researched and compared, such as local threshold method, Otsu method andmaximum entropy method, experiment results indicate that two-dimensional(2-D)maximum entropy method has better segmentation results. The traditional2-Dmaximum entropy method has the problem of large amount of calculation and poorreal-time performance, on the basis of fast2-D maximum entropy method, integralimage technique and features of gray distribution based on pixel gray value and theaverage gray level of eight-neighborhood of the pixel are used to improve thealgorithm’s operating efficiency. According to the continuity of the target trajectory, aneighborhood judging method is used to select the targets in the image sequence,target point is specially marked by the gray value in the superposition image in orderto distinguish target point’s frame number and position, so the complexity ofalgorithm and data storage capacity are reduced. Mean Shift tracking algorithmcombined with Kalman filter is used in this paper, and experiment results indicate thatthe algorithm can effectively and accurately track targets.Optoelectronic tracking system based on two-axis turntable is used as the performance verification platform of different algorithms, the system adopts thethought of modular design, and uses MFC to write the human-computer interactioninterface, the simplicity of operator is convenient, and it is easy to modify and addalgorithm. The linear camera model and camera imaging principle are used to adjustthe angle of rotation of turntable, in order to achieve the conversion between visualinformation and control information. And then this paper introduces the function ofeach module in detail, and tests system performance by experiment.