Research on Multipoint Collaborative Target Recognition and Tracking Algorithm

Frame differencing method can adapt well to dynamic environments with changes in lighting and complex background transformations. However, its characteristic of determining whether an object is in motion by extracting the object's contour leads to misdetection and missed detection when detecting moving objects. Meanwhile, frame difference operation also increases the calculation processing time and thus generates a higher delay. To address the mentioned issues, the article proposes a multi-point collaborative object recognition algorithm. Firstly, perform grayscale processing on the image. Secondly, it employs an improved BM3D (Block Matching 3D) algorithm for denoising. Then the image data after noise removal is identified and tracked by an improved adaptive interframe difference algorithm. Finally, a multi-point collaborative FPGA-based object recognition and tracking system is designed to validate the algorithm. Experimental results show that, compared to traditional frame differencing methods, the improved algorithm reduces latency by approximately 200 microseconds and significantly enhances the accuracy of object recognition.