Small Target Detection And Tracking In Deep Space

The technique of small target detection and tracking under the background of deep space has been a hot research topic in the field of digital image processing, the main purpose of this Master dissertation is to develop the detection and tracking of weak target in low SNR(signal noise ratio) and low contrast image sequence.there are usually numerous stellar targets and noises that interfere with the image of deep space.The dim target has similar image characteristics with the stellar targets,which are often submerged in noise interference.The dim target is lack of clear shape and texture characteristic, the traditional single-frame detection technology is difficult to effectively detect and identify the target.This paper mainly develops the target detection technology based on motion information of multiple frames.Firstly, this paper studies the preprocess technology of dim target detection. The method to suppress the highlight center and edge of stellar targets based on dilative difference method domain and mathematical morphological reconstruction. Supressing the noise with spatial domain filtering algorithm such as median filter and morphological filtering.In the aspect of image segmentation,a modified m+kδ segmentation method based on the threshold of constraint points is adapted to extract the target and obtain the binary image. In view of the slow moving dim target, improve the traditional way of energy accumulation of the original high correlated method.9 frames after four related calculation can get all the information with a relevant unit of three frames in the sequence, at the same time, the azimuth Angle and curvature constraints are adapted to reduce the calculation and improve the efficiency of calculation, which reduces the probability of target residual and suppress the noise interference. In the view of moving target,The paper developed the moving target detection algorithm which combined the the projection algorithm and hough transform, Reducing the number of target with the target clustering for the projection image,at the same time,Improving the original classic linear hough transform algorithm with a series of constraints.adopt a principle of “many-to-one mapping ” to detect the trajectory of target.The theory and experiments show that the improved hough transform algorithm has a good performance of target detection, which got a correct test results.