Design And Implementation Of Small Target Detection And Tracking Based On Mathematical Morphology And STC

Object detection and tracking is a key problem in computer vision, and it is the basis of follow up treatment such as visual scene analysis and semantic analysis. The technique of object detection and tracking has a wide range of application in intelligent video surveillance, robot navigation, intelligent transportation, human computer interaction, industrial inspection, medical image processing, national defense, etc.With the widely application of high-tech in military, kinds of precision guided weapons(such as guided artillery, precision guided missiles, guided mines, etc.) play an important role in the modern war, which leading to a serious threat to the ship, command center and key areas such as security. Therefore, it is important to design an effective target detection and tracking system to detect the incoming target timely for our country's military operations and defense system.This thesis focuses on the three important steps of preprocessing, detection and tracking in target detection and tracking system, so the main contributions of this thesis are as follows:1. The image preprocessing algorithm based on the combination of NWTH transform and ellipse paraboloid fitting is researched, and the algorithm is verified by using C++ and Opencv. Aiming at the strong interference between noise and complex background clutter in the image, the structure elements of NWTH transform are optimized. Moreover the target is fitted by the ellipse parabolic to eliminate the noise further. Finally, a global threshold segmentation technique is easily used to detect small targets. Results show that the proposed algorithm's accuracy is more than 90% in complex background, strong noise and low SNR compared with the Top-hat algorithm.2. The algorithm of target tracking based on spatial-temporal context is improved. Aimming at the problem that the space model can not be updated when the context area is outside the image boundary, the boundary problem of STC algorithm is optimized in this paper and the stability and success rate of tracking is improved. The STC algorithm is implemented on C++ platform in this paper, and the general target and the small target were tracked successfully. The STC algorithm and Struck algorithm are evaluated by single frame processing time and tracking accuracy. Results show that the improved STC algorithm has good performance in accuracy.3. A small target detection and tracking system is designed. The system combined with the small target detection algorithm and target tracking algorithm can detect and track the image sequences automatically. This system is implemented on the C++ and Opencv platform. Results show that the algorithm has good performance.