Study On The Algorithmic Implementation Technology Of Photoelectric Target Detection And Tracking System

The background of this paper is a scientific research project: "the photoelectric shipborne scout system building on a naval reconnaissance ship". In this thesis, the algorithm of the photoelectric target detection and tracking system is mainly studied. Following the step of photoelectric data processing, the technologies of image pre-processing and photoelectric target detection are introduced in detail.In exordium, the application background of the article is introduced firstly. The framework and working outline of the photoelectric target detection and tracking system are introduced too. At last, research actuality and developing trend of the photoelectric shipborne scout system are given.Secondly, the bad pixel replacement and nonuniformity correction technologies of photoelectric image pre-processing are introduced. Existing scene statistical algorithms for nonuniformity correction are analyzed and compared. Aim at solving the intrinsic problem of kalman filtering algorithm in building state model, two state models are used to simulate the non-drifting case and the severe drifting case. IMM algorithm is used to combine the estimations of the response parameters of detectors in two cases, in this way, the bad impact of the indetermination of response parameter is gotten rid of. The experiment and simulation analysis reveal that the IMM algorithm compares favorably to the kalman filtering algorithm in convergence performance.In chapter three, a simple and effective process is proposed for the detection of photoelectric target in sky background. First, background component is suppressed using temporal differencing method to find the position of target in image. Then, only in the small potential region of target, original image is segmented to extract target. In this way, the bad impact of background on segmentation is removed, and the computation complexity decrease. Image segmentation is very important in target detection algorithm, so in this chapter, some segmentation algorithms are introduced. These algorithms are compared at computation complexity and memory requirement. OSTU and Moment Preserving Thresholding method are fitter for real-time photoelectric image segmentation.In chapter four, a process of small target detection is proposed. The process consists of suppression of background and multistage hypothesis testing (MHT) algorithm for searching target. Designing method of MHT is introduced in detail. The performance of MHT is analyzed and the relationship between the SNR of image and the algorithm computation complexity and memory requirement is acquired, which provide theoretical guide for the selection and designing of hardware.Finally, the main works of the thesis are summarized. The defect of system designing is analyzed. Also the future research areas are pointed out.