Research On Target Detection Of Panoramic Infrared Search System

This dissertation, in the background of development of real-time signal processing system of panoramic infrared searching, focus on study of the automatic dim small target detection in the low altitude sky complex background and the realization of the real-time signal processing system in depth. In order to overcome the difficulties to detect the small targets in complex background and process large volumes of data in real time, an effective and real-time detection algorithm which is applied in panorama image and satisfies the engineering needs is proposed. The proposed algorithm has been applied to high-performance long linear detectors panoramic infrared search alert system and performed very well . The dissertation is comprised of three parts: In the first part, based on the analysis of the imaging of the long linea panoramic infrared search alert system and the characteristics of panorama image, aiming at the phenomenon of shift and scaling for panorama image caused by unsteadiness of servo, a multiple region matching calibration algorithm is proposed. Firstly, optimal positions are obtained by a weighted correlation tracking algorithm based on absolute difference which fully utilizes the similarity and difference information; Secondly, the acquired optimal positions are applied to two adjacent frames to calibrate panorama image linearly. The experiment carried on the true data proves the algorithm effective.In the second part , aiming at real-time dim small target detection under panoramic airspace, on the base of analysing the key technologies about background suppression , target detection and confirmation, and fully considering detection performance, complexity and realtime , two target detection algorithms are proposed, which are rough detection based on spatial domain combined with precise detection using space-time information and target detection algorithm based on extracting target potential areas.The first algorithm obtains suspect targets under high probability target detection by quick real-time morphological filtering; then focusing on the image slice of detected suspect target, it has further delicate detection and recognition based on the integrated processing of morphological filtering and reformative interframe differences. The second algorithm divides the whitening filtered panorama image into mulit-blocks, and distinguishes potential target areas according to the variation of block parameters, and then fuses the potential target areas to form small view-field chips; the last step is process every small view field image conventionality. The detection results operated on the test images present the efficiency of proposed algorithms.In the third part, the paper studied the engineering realization technology of panoramic infrared searching and real-time signal processing system. For the application in mulit-targets processing and image acquisition, it designs a real-time image processor based on dual-DSP and FPGA, on the base of which the DSP software is developped combining rough detection based on spatial domain and precise detection using space-time information; For the application in system display and control, it designs a comprehensive software integrating display and control. The joined debugging results of the system show that all kinds of indexes have satisfied the designed specifications, and realize stable detection under the condition of low altitude sky complex background of panoramic airspace.