Target Detection And Ranging System Based On Binocular Vision

With the rapid development of computer vision,binocular vision technology has become indispensable in industrial inspection,scientific research and other fields.It can detect the target and obtain the distance information of the target without touching the target,so it has a very high research value.However,the actual situation is complex and changeable,only relying on binocular vision technology can not detect the target well,and accurately calculate the target distance information,and the speed is relatively slow.In view of the above problems,this paper combines the target detection algorithm and binocular ranging technology to build a target detection and ranging system based on binocular vision.The main research work is as follows:(1)In the aspect of target detection,this paper adopts a lightweight YOLOv5 s target detection algorithm.Firstly,the target to be identified in this paper is self-made by data enhancement and Label Img labeling.Then the transfer learning method is used to train and test the network model.Finally,the detection performance of YOLOv5 s model is evaluated by four indicators: accuracy,recall rate,average precision and transmission frame rate per second of images,and the results are compared with other typical target detection models.The experimental results show that YOLOv5 s model not only ensures the accuracy of target detection,but also meets the real-time requirements.(2)In the aspect of binocular ranging,this paper adopts the target ranging method based on binocular vision.Firstly,Zhang Zhengyou calibration method and Bouguet algorithm are used to preprocess the binocular images.Then,three stereo matching algorithms,GC,BM and SGBM,are elaborated.The advantages and disadvantages of BM and SGBM are analyzed in the design experiment.The stereo matching algorithm suitable for this topic is selected to match the pixels of the corrected left and right images,and the corresponding disparity map is generated.Finally,the distance of the object is estimated according to the binocular distance measurement formula to obtain the distance information of the object in the real scene.The experimental results show that the measurement accuracy of the target ranging method based on binocular vision is high,and the error is within 6%.(3)The target detection and ranging system based on binocular vision is established by combining the target detection algorithm and binocular ranging method.Firstly,the target in the scene is detected by the trained YOLOv5 s model,and the position information and category information of the target are obtained.Then,the disparity map of the scene image is obtained by the calibrated binocular camera,and the distance between the detected target and the binocular camera is calculated according to the principle of binocular distance measurement.In order to verify the effectiveness of the results,this paper conducts verification experiments at different distances for single target and multiple target situations.The experimental results show that the system can obtain the target category information in real time and calculate the real distance of the target accurately.