Application Of Machine Vision In Equipment Display Information Detection

With the continuous development of industrial inspection technology,the speed and quality of traditional manual inspection cannot meet the current operation needs.In order to improve the efficiency and quality of inspection,many intelligent systems have been widely used in inspection process.In this paper,an intelligent software inspection system based on machine vision is designed and implemented to improve the intelligence of industrial inspection.A machine vision-based detection system is designed for the detection environment with low intelligence,many equipment points and wide range.The system can realize the automatic detection and recognition of equipment data with the functions of target tracking,target detection and target recognition.Target tracking: Firstly,YOLOv3 depth learning algorithm is used to locate the target in the initial frame,and then ECO-HC target tracking algorithm is used to predict the target position.The ECO-HC algorithm predicts the target position to adjust the camera platform position,so that the robot does not lose the target in the process of approaching the target,and ensures that the target is in the center of the machine vision when tracking is completed,so as to improve the accuracy of information recognition.Target detection: YOLOv3 deep learning algorithm is used for multi-target detection.The algorithm can locate the target location quickly and accurately,and classify it accurately,then capture the image of each target area.Target recognition: Firstly,according to the characteristics of the target area,such as color and brightness,the corresponding image preprocessing method is selected.Then,the digital area is divided into several independent digital images,and the LeNet deep learning network is adopted for digital recognition.The system is built on the platform of the Ubuntu system and ROS(robot operating system).The performance test of the vision detection system is completed with NVIDIA Geforce GTX 1080 GPU as the computing resource.The results show that the system can achieve the effect that the detection accuracy is higher than 95%,the detection time is less than 150ms/frame,so the target location and target recognition can be accomplished quickly and accurately.