Research On Moving Target Detection Technology Based On ZYNQ

Moving object detection is one of the research hotspots in the field of computer vision.This technology can locate and detect objects in video images and it is the basis for tasks such as target tracking and behavior analysis.But in the process of target detection,there will be various kinds of problems,such as complex background images,obstacles partial occlusion and sunlight exposure.So how to detect moving objects better is of great significance.At the same time,the hardware and software co design is based on ZYNQ which is developed by FPGA and ARM to explore new ideas for the problem of computer vision in complex mode.In this paper,the research on moving object detection is based on ZYNQ,and it is based on the traditional AdaBoost algorithm,combined with the structural characteristics of Haar,presenting a new target detection algorithm,and the software and hardware co design is implemented on ZYNQ.This paper is divided into four parts:The first part introduces the research background and the significance of moving target detection,and also the current research status.The second part begins with the introduction of the hardware structure of this system,and the system mainly consists of front-end camera acquisition subsystem,image processing subsystem and image display subsystem.After that,the specific hardware and its parameters of each subsystem are introduced in detail.The third part introduces some commonly used moving object detection algorithms,and then proposes a target detection based on structural Haar feature and AdaBoost algorithm,and the traditional AdaBoost learning algorithm is improved,which not only impproves the training speed of the samples,but also improves the accuracy of the detection.The fourth part introduces the hardware and software collaborative for moving target detection implementation based on structural Haar features and AdaBoost algorithm.First,this part introduces the system used development tools,AXI protocol,AXI VDMA basic configuration and the image preprocessing framework,and then sets up an embedded development framework,transplants OpenCV,Qt and other required libraries,and finally in the ZedBoard development board hardware implementation,the final experiment shows that the system can detect the moving target correctly and have high detection rate under the circumstances of complex background,obstructions and sunlight.At the same time,the way of using hardware and software than the way of working simply using the software,you can save nearly 4.5 times as long,so the system can achieve real-time detection.