Research On Dynamic Target Detection And Tracking Algorithm Based On Python+OpenCV

Dynamic target detection and tracking technology is one of the hot topics in the field of computer vision.Combining the research results of visual navigation,artificial intelligence and pattern recognition,it plays a major role in video surveillance,video retrieval,medical image analysis and many other fields.So the research topic has important theoretical significance and practical use value.This thesis is based on OpenCV computer vision library,using Python programming,taking the moving target in video as the research carrier,using digital image processing technology to process the image,improving the target detection algorithm to improve the real-time and effectiveness of the detection process,and optimizing the tracking processing algorithm.Achieve fast and accurate tracking of your goals.The main research contents of this thesis are as follows:1)Moving object detection algorithm based on deep learning:First,based on the traditional SSD(Single Shot MultiBox Detector)algorithm,ResNet50 is used to replace the traditional VGG16 network model for improving real-time performance.Secondly,for improving effectiveness,Based on the traditional SSD algorithm of the cost function,the Softmax activation function introduces the central loss as a cost function of the improved SSD algorithm.Finally,experiments are performed in the Pascal VOC 2007 dataset and the actual scene,and compared with YOLO and traditional SSD algorithms.The experimental results show that the improved SSD algorithm has enhanced effectiveness,improved real-time performance and higher precision.2)Research on Robust Tracking Algorithm Based on Mean Shift and Kalman Filter:In this paper,the Mean Shift algorithm is improved and fused with Kalman filtering to track moving targets.First,the video consists of three bytes of R?G and B for each pixel in each frame,and shifted every byte of the pixel in the tracking window to the right for four bits,in another word,reduce each byte to 1/16;Secondly,joint the three half bytes to an integer range from 0 to 4095.With the improved Mean Shift algorithm,it is not necessary to calculate the Bhattacharyya coefficient,reduce the computational complexity of the tracking process,and track the moving target in real time,but find that the tracking robustness is not ideal,so the Kalman filter is used to improve its tracking window.The purpose of optimization.The experimental results show that the method can track moving targets quickly and the robustness is enhanced.