Autonomous Monitoring Technology Of Moving Target Based On Machine Vision

With the rapid development of artificial intelligence technology and machine vision technology,the need to improve the quality of life of the people and promote the construction of national security has become increasingly urgent.Whether civil or military,visual image information has a huge application value,and the intelligent processing of large data volume image information has also become a research hotspot in recent years.As the core technology in the field of machine vision,target detection and identification technologies based on artificial intelligence algorithms and resource architectures play an important role in autonomous monitoring systems such as ground and air.This paper mainly studies the machine vision-based self-monitoring technology of moving targets,and explores applications under various monitoring platforms.This paper analyzes the software algorithms related to the detection,tracking,and recognition of moving targets involved in the target autonomous monitoring technology.Firstly,with regard to the background motion,a motion target detection and tracking method based on affine transformation and frame difference method is used to compensate the background motion to a relatively stationary background.In ground and drone experiments,detection errors caused by dynamic background can be eliminated,and moving targets can be detected and tracked.Secondly,based on the HOG feature vector and SVM classifier,a specific target detection and recognition algorithm is proposed.Based on the deep learning neural network,a multi-objective autonomous detection and identification algorithm was proposed and a ground verification test was conducted.In the visible light image data,94.64% detection accuracy of the aircraft target and 84.85% detection accuracy of the ship target are achieved.In the infrared image data,95.31% detection accuracy of the automotive target and 92.79% detection accuracy of the pedestrian target were achieved.It is verified that the detection and recognition method based on deep learning neural network algorithm has better target detection and recognition ability in the visible and infrared images with better target shape information.At the same time,under the ground test hardware platform conditions,the single-frame image detection time is 0.7s,and the average detection and recognition accuracy rate can reach 91.89%.This topic is based on the algorithm research of autonomous monitoring system of moving targets.Using Python language OpenCV library,we design and develop a self-monitoring technology platform for moving targets and verify the feasibility of the algorithm.Experiments were carried out on a drone platform,and experimental images were validated using infrared aircraft images.Finally,the constraints of ground intelligence servers and on-board resource environment were analyzed and compared.The intelligent ground processing server integrates a high-density GPU module to increase the processing power of single-frame images to 0.12 s or even 0.06 s.Hardware architecture,weight,power consumption,and operating temperature all have the possibility of working on the star,which can provide support for the use of artificial intelligence algorithms and hardware resource architectures for space-based platforms to complete intelligent processing of big data under constrained conditions on the satellite.