Research On 3D Reconstruction Of Hazardous Chemical Warehouse Based On Stereo Vision

At present,video monitoring technology has been widely used in hazardous chemicals storage industry,but there is a problem that video monitoring lacks data monitoring,and there is a lack of real-time,effective and automatic analysis of storage status of video monitoring mathematical model,which makes video monitoring technology not play its due role.Using binocular vision 3D reconstruction technology to monitor the safe distance of dangerous chemicals storage goods stacking,to ensure the safety of storage has important practical significance.In this paper,a double vision technology is used to establish a safe distance detection system for the stacking of dangerous chemicals.The system analyzes binocular images,calculates parallax,and reverses the 3D information of the object.In order to avoid the influence of complex background on disparity,this paper proposes a binocular stereo matching method based on image segmentation.Firstly,the image is classified by pixel by pixel using semantic segmentation model,and binocular images are divided into multiple pairs according to the category of items.The improved sgbm stereo matching algorithm is used to calculate parallax for each pair of images,and then the disparity image is mask processed once.Each image is fused to obtain the final version of disparity map.In the image segmentation,the deep learning based semantic segmentation model of deeplbv3 is used to realize the migration training under the condition of small sample.Moreover,disparity map distinguishes different types of objects,so that the distance between stacks can be calculated directly by the subsequent 3D reconstruction.In the three-dimensional reconstruction,based on the disparity map obtained by stereo matching,the 3D point cloud information in the world coordinate is obtained,and the outliers are removed.In the calculation of stacking distance information,a new algorithm is proposed to calculate the minimum distance between objects,and the experimental comparison is made with the traditional algorithm of finding the parallel distance between objects.The results show that the new algorithm can obtain more accurate measurement results under the condition of known semantic segmentation information.The experimental results show that the relative error of measurement distance is less than 1%,which can meet the requirements of safety distance detection.It can provide effective technical support for the establishment of safety status monitoring system of hazardous chemicals storage,and is of great significance to improve the safety production level of hazardous chemicals storage.