Research On Key Techniques And Algorithms For Binocular Vision Monitoring Of Safe Distance In Hazardous Chemical Stores

In recent years,there have been frequent explosions which have caused a large number of casualties and property losses in hazardous chemical warehouses.The safety "five-distance" overrun in the warehouse is an important factor.Research on the measurement algorithm for safe storage distance in the binocular vision warehouse to realize the "five-distance" warning is a promising solution.The paper adopts the triangle similarity theorem and the algorithm of matching pixel differences to simplify the measurement steps of binocular vision;for the problem of low accuracy of the cargo boundary extraction under the night vision experimental environment,the average gray histogram algorithm is improved based on the two-dimensional Otsu to improve the accuracy of threshold calculation enhances the robustness in different environments.In public data sets and experimental scenarios,all the pictures to be tested have a threshold value of 77.3% of the pictures whose threshold values are distributed on the negative diagonal,which basically meets the requirements of distance measurement.The accuracy of corner detection directly determines the accuracy of matching and parallax calculation in binocular ranging.Aiming at the problem that the traditional harris corner detection algorithm is sensitive to interferences such as light irradiation,shadow and other non-palletized goods,the non-maximum suppression of edge features and K-means clustering algorithm are used to improve the noise resistance of the traditional harris algorithm.The experimental results show that the accuracy rate increases by 10.62% under different light conditions;the accuracy rate increases by 3.95% under the influence of shadows;and the accuracy rate increases by 52.53% under the interference of non-palletized goods.Based on the three states of storing dangerous goods in the warehouse,full goods,non-full goods,and exceeding the standard,combined with the judgment of the position of the corner point,it is determined whether the "five distances" of the palletizing in the warehouse are exceeding the standard.In addition,the binocular corner matching algorithm based on palletizing edge features is used to improve the efficiency of the traditional binocular matching algorithm,for the problem of matching errors,the algorithm is optimized using strong feature point matching and window and neighborhood matching.The accuracy of the traditional binocular matching algorithm has been improved.At the same time,the paper also uses the generative adversarial network GAN to generate samples for the problem of insufficient number of palletized storage image samples in the hazardous chemicals warehouse,and explored the cloud platform transmission and management of data.