Design Of ARM-based Bionic Ammunition Target Detection System

With the rapid development of machine vision,video image moving object detection technology has become the research focus in this field.In recent years,great achievements have been made in the detection of moving objects in the scene with fixed background,such as video surveillance.However,in some complex scenes with moving objects in the background,the detection effect is not ideal,such as birds,fish,insects,snakes and other bionic scenes Ammunition is affected by its own motion in the process of moving target detection,and the background motion information is the interference item,which increases the difficulty of target detection.The characteristics of bionic ammunition are small size,high requirements of embedded image real-time processing system,the need for appropriate algorithms and lowpower embedded platform to complete the real-time detection of moving targets.In this paper,the above problems are studied deeply,the real-time moving target detection algorithm is optimized,and a bionic ammunition target detection system based on arm is designed.The main research is as follows:(1)The basic theory and algorithm principle of three common moving target detection algorithms,optical flow field method,frame difference method and background subtraction method,are compared.The detection results of three algorithms in static background and dynamic background are compared through simulation experiments.Its application scope,advantages and disadvantages are evaluated.(2)Aiming at the problem that biomimetic ammunition is affected by its own motion and the difficulty of target detection is increased after the background motion information is integrated,an improved optical flow algorithm based on multi method fusion is proposed,which uses the optical flow gradient to solve the approximate boundary of moving target,and introduces the maximum class square difference method to replace the traditional artificial experience threshold selection method to select the best threshold.The ray projection algorithm is introduced to fill the interior pixels of the moving object.In view of the problem that a large number of isolated noise points appear in the threshold segmentation and pixel filling of the target,which affects the extraction of the moving target,the black noise points in the foreground moving target and the white noise points in the background are removed by the open and close operations in the morphological filter.The simulation experiment results show that the detection accuracy of the improved algorithm is significantly improved.(3)In view of the small size of bionic ammunition and the need to effectively combine the embedded platform with the moving target detection technology,the arm based bionic ammunition moving target detection system is constructed,the hardware platform design with S3C2440 as the main control chip is completed,the software development environment is built,and the embedded Linux operating system is transplanted,on this basis,the embedded movement is completed The application software of target detection system is designed,and the test platform of bionic ammunition target detection system based on arm is developed.Based on the test platform,a comparative test is carried out under the conditions of simple background and complex background,strong light and dark light,single target and multi-target.The test results show that the arm based biomimetic ammunition target detection system designed in this paper has strong feasibility.