Research on Segmentation and Recognition Algorithm of Moving Human Image Based on Deep Learning

With the improvement of living standards, people have higher and higher requirements for the quality of life and continue to pursue a green and healthy life. Video surveillance system has more and more security requirements in the fields of transportation, banking, security, power and military facilities. Therefore, video surveillance has been widely used in all aspects of people's production and life. The image segmentation algorithm predicts a discrete semantic label for each pixel of the image, which can be used to extract the target object with hard boundary in the image and synthesize it into a new background. But this does not satisfy all needs. Most of the current methods of human image segmentation can only guarantee one of the indicators, which makes the practical application of human image segmentation greatly limited. According to the difference between different areas in the video image, the video image is divided into motion area, background area and occlusion area, and the edge detection operator is used to detect the edge of the moving object, and finally the moving object in the video image is segmented. The algorithm is characterized by simplicity and high efficiency, and is suitable for monitoring systems with limited resources such as embedded systems. Experiments show that this algorithm can segment moving human images in different scenes, and it has good robustness and accuracy, reaching a practical level. This algorithm has practical significance for automatic control of lens aiming at moving human in surveillance system.