Hardware And Software Codesign Of Real-time Motion Detection System Based On FPGA

In the last decade vision-based intelligent video surveillance systems have been widely used as an important tool to protect the safety of society. Intelligent video surveillance systems are an important part of the security system. It is an integrated system with strong prevention capabilities. Motion detection is the first essential processing component for many video applications such as video surveillance, robot navigation, path planning, military reconnaissance, collision avoidance and so on. According to a survey of relevant research, motion detection methods can be grouped into three main categories: temporal differencing, optical flow, and background subtraction. As the most widely used motion detection method, background subtraction has always been a popular research area. In recent years, in order to build a good adaptive background model, researchers have proposed many kinds of background modeling methods.In this research, we present an improved motion detection algorithm, and implement it in the computer system and as well as in an embedded system.In the first instance, we present an efficient motion detection algorithm based on a background subtraction method to extract moving objects without complex computing. It comprises of three modules: a background generating module, a block-based filter module and an object extraction module.Secondly, based on our algorithm, we use computer programming to simulate and test. As we can see from the testing results, our algorithm can be used with complex backgrounds as it has a good ability to deal with complex background noise. Our algorithm is able to generate reliable reference background and get high accuracy results with a low computational complexity.Finally, we finish the hardware and software co-design of the motion detection system. In the hardware implementation, we use C++ as the programming language to implement the system using the Vivado HLS tool. The hardware part provides the morphological erosion and dilation operations in the block-based filter module. The software part runs in the porting system. It uses OpenCV to simplify the video stream processing.After we finish the hardware and software co-design, our system has a high accuracy, robustness and a high real-time performance.