Research On Inspection Behavior Identification Of Underground Power Distribution Room

Stable power supply is a prerequisite for ensuring safe production in coal mines.The underground power distribution room is an important part of the coal mine power supply system.Therefore,the underground power distribution room needs to be regularly inspected,and hidden troubles are dealt with in time.At present,it is judged whether the personnel complete the inspection task according to the regulations,mainly through the inspection of paper records and manual monitoring in the monitoring room.There are problems such as cheating by the inspection personnel and heavy manual monitoring workload.Therefore,it is necessary to study the intelligent detection method of the inspection behavior of the distribution room to ensure that the inspection personnel complete the inspection according to the regulations and ensure the safety of the coal mine power system.This paper divides the inspection behavior inspection of the power distribution room into two parts: inspection behavior recognition and positioning,which are based on the two-stream convolutional network(Two-Stream CNN)human behavior recognition algorithm and the boundary sensitive network(BSN)human behavior positioning algorithm is researched and improved,and a patrol behavior recognition system for power distribution room is designed to intelligently recognize patrol behavior of patrol personnel in power distribution room.The main research contents include:(1)Propose a method for identifying the inspection behavior of underground power distribution rooms based on the improved two-stream method.Through scene analysis,the inspection behaviors are divided into five types: standing detection,squat detection,walking,standing records,and sitting down records.An inspection behavior data set(IBDS5)was produced.Divide each inspection behavior video into 3 parts,corresponding to the start of inspection,middle inspection and end of inspection;randomly sample the three parts of the video to obtain RGB images representing spatial features and continuous motion features Optical flow images,and input into spatial flow network and time flow network respectively for feature extraction;weighted fusion of the predicted features of the two networks,and the results of inspection behavior recognition are obtained after classification.Experimental results show that the accuracy of Top1 recognition on the IBDS5 dataset and the public dataset UCF101 is better than the existing methods such as 3D-CNN and traditional two-stream CNN.(2)A method for locating the timing sequence of underground power distribution room inspection based on improved BSN network is proposed.For the timing behavior positioning of underground power distribution room inspection,it is necessary to accurately locate the location of the inspection behavior in the video.First,the features of patrol inspection are extracted.The improved two-stream convolutional network can extract the features of longer videos well.Therefore,the improved two-stream convolutional network is used for feature extraction to obtain feature sequences.Secondly,the extracted video features are input into the BSN network to locate the start and end times of the inspection to form candidate time regions,and then evaluate whether each candidate time region contains the confidence of the inspection behavior,and obtain the candidate combination containing the confidence score.Finally,the NMS algorithm is used to remove the redundancy,and the location of the inspection behavior is obtained.The experiments compared the impact of different Io U thresholds and the number of different candidates generated by each video on the location recall rate of inspection behavior.The results show that,compared with other methods,this method achieves a higher average recall rate.(3)Based on the improved Two-Stream convolutional neural network algorithm,a C/S mode power distribution room inspection behavior recognition system is designed and implemented.By inputting the monitoring video of the power distribution room,the system detects and analyzes whether the personnel in the power distribution room have completed the inspection,which reduces the cost of human resources and increases the reliability of monitoring.