3D Information Reconstruction Based On Binocular Stereo Vision

With the development of artificial intelligence technology in recent years,computer vision has also become one of the current research hotspots.Since the basic principle of binocular stereo vision is basically the same as that of the human visual system,binocular stereo vision is the focus of current research.In order to provide trajectory correction information for the indoor mobile robot avoiding obstacles,it is necessary to use a binocular stereo vision system to obtain the spatial information of obstacles in the external scene.Binocular stereo vision mainly uses two cameras to collect images of the same scene from different angles.Through the processing of image pairs,the geometric data between the corresponding points of the objects in the scene is obtained,and then according to the geometric model of binocular stereo vision,the scene is finally obtained.The three-dimensional information of the object can be used for obstacle avoidance of robots.Due to the simplicity and practicability of this method,it has a good application prospect.The three-dimensional reconstruction of environmental information is divided into four parts: camera calibration,image preprocessing,image stereo matching and three-dimensional information reconstruction.First,the camera and lens were selected.Based on the parallel binocular stereo vision model,the two cameras were placed in parallel to complete the construction of the binocular stereo vision platform.In the camera calibration stage,the self-made calibration board was taken from different angles,and the MATLAB calibration toolbox was used to manually complete the camera calibration to obtain the internal and external parameters related to the two cameras.In the image preprocessing stage,the acquired color image was denoised using Gaussian filtering,the histogram matching completes the brightness difference balance,the next grayscale processing obtains the grayscale image,the LOG operator sharpens the four-step processing,and finally obtains the noise Image pairs with few,small differences in brightness,and prominent edge details.In the image stereo matching stage,the region matching algorithm was selected to obtain the disparity map.Aiming at the problem that the traditional Census algorithm has single window information and is susceptible to noise interference,the traditional Census algorithm is improved,and new pixel evaluation criteria were introduced to enrich the acquisition of window information.Experiments show that the improved algorithm in this paper can obtain dense disparity maps,and has improved matching accuracy and noise resistance,which can be used to complete the final 3D information reconstruction work.In the end,the complete 3D information reconstruction process is sorted out,using the disparity map that has been obtained,the 3D space information of the object is completed through the three-dimensional coordinate calculation method,and the three-dimensional reconstruction is completed.Accuracy experiments were performed on the obtained 3D information reconstruction results to verify the accuracy and practicability of the reconstruction results.