Research On Automatic Detection Of The State Of Hydraulic Support Guard Plate

The hydraulic support guard plate is a kind of protection device for the fully mechanized coal mining face.It is used to protect the exposed coal wall and the roof plate,and prevent the accident of the top sheet from being injured or killed.Therefore,it is particularly important to detect the status of the guard plate,but the traditional guard plate state detection method has the problems of high cost and inaccurate detection results.This paper adopts target detection technology and aims at the dust and fog phenomenon of fully mechanized mining face,and proposes an algorithm that combines image defogging and Tiny-YOLOv3 to solve the problems of traditional guard plate status detection.1.In order to improve the running speed of the image defogging algorithm,the CUDA parallel optimization scheme of the image defogging algorithm was designed.First,replace the dark channel map with any single channel map of the original image to shorten the calculation time of the dark channel map;Secondly,a calculation method for obtaining atmospheric light values in columns is proposed to accelerate the acquisition of atmospheric light values;Third,the minimum value filtering kernel is split into rows and columns to optimize the calculation process of the initial transmittance.Finally,the mean value filtering is realized by matrix transposition to improve the efficiency of refining the transmittance.The experimental results show that the image defogging algorithm optimized by the CUDA platform can meet the requirements of real-time defogging under the premise of ensuring the defogging quality.2.In order to improve the detection accuracy of Tiny-YOLOv3 algorithm,an improved Tiny-YOLOv3 algorithm is proposed.First,the residual structure and pyramid pooling module are introduced into the network structure to prevent the loss of feature information and obtain rich semantic information;secondly,the idea of gradient equalization is used in the classification loss function to solve the detection accuracy caused by sample imbalance and discrete samples.loss.The experimental results on the VOC2007 data set show that the improved TinyYOLOv3 algorithm has improved detection accuracy compared to other target detection algorithms.3.In order to reduce the impact of dust and fog on the fully mechanized mining face on the detection task,an algorithm combining image dehazing and Tiny-YOLOv3 is proposed.The algorithm first performs real-time defogging on the video image,secondly detects the state of the guard board on the defogging image,and finally outputs the detection result in real time.The experimental results in the data set of this paper show that the detection accuracy of the fusion algorithm is significantly higher than that of other target algorithms,which meets the requirements of real-time detection and verifies the effectiveness of the fusion algorithm.4.Based on the fusion algorithm,a guard plate status detection system is designed.This system realizes the automatic recognition of the guard board state,and can alarm the abnormal state of the guard board in real time,remind the hydraulic supporter and the safety officer,and further ensure the operation life safety of personnel.