Detecting And Tracking Of Small Infrared Target Under The Background Of Sea Level

In resent years, the application of intelligent surveillance system is becoming one of the most important research topics in many countries:Moving targets detecting and tracking are considered as the basic and one of the most important problems in this research. It is also an active field that had attracted many researchers.Detecting and tracking of small infrared target under the background of sea level are mainly discussed in this thesis. Due to dynamic complex background and defects of infrared image, which are low contrast between object and background, blurred edge of object and high amount of noise, the infrared images need to be done some preprocessing, such as noise reduction, background suppression and so on. The major task is to simulate the detection algorithm via Matlab and realize it on hardware development platform based on TMS320DM642 DSP, and program under circumstance of Visual C++ to realize tracking.In the infrared image preprocessing, sea and sky background is partitioned by the algorithm based on sea-sky-line, which can reduce the computation time. What's more, infrared image is coped with by the median filter and high pass filter to achieve infrared image denoising and edge enhancement as well as the background suppression.In the processing of target detecting, the moving object detecting algorithm is proposed in this thesis, which combines the method of Background Subtraction with the Top-Hat based on morphology. The detecting algorithm is transplanted and optimized on hardware development platform based on TMS320DM642 DSP. The method of Background Subtraction can generate most of foreground image, while Top-Hat can greatly suppress background noise and can't detect all the information of moving target. So the algorithm combining the two methods can complete detection of the target.In the processing of target tracking, the algorithm which combines Mean shift to Kalman filter is used. Kalman filter can predict the possible position in the current frame, which is used as the starting position of Mean Shift. Then the algorithm of Mean Shift will search in this neighboring range to find the true position of the target. This method is effective for fast moving target, and it also makes sense to the object under the circumstance of low signal-noise rate.