| Coal is one of the main energy sources in our country,and ensuring the safe production of coalliery is one of the main goals of the coal industry.The conveyor belt area under the mine is a dangerous area in the production of coal.Workers often operate illegally,which can easily lead to safety accidents.Due to the particularity of the environment in the coal-mining production,the development of video surveillance technology under the coal mine is not mature at present,and it is impossible to control the illegal behaviors and accidents that are prone to occur in the conveyor belt area.Therefore,this paper studies the existing problems of the condition monitoring of miner and conveyor belts in the coal-mining underground conveyor belt area,and proposes a collaborative analysis method based on the monitoring video,and uses this method to realize the illegal detection software of the underground conveyor belt area.The main contents are as follows:1.Aiming at the problem that it is difficult to supervise the state of people and conveyor belts in the underground of conveyor belt area in the coal-mining production,a video-based collaborative analysis for the state of people and conveyor belts in underground mines is proposed by us.Based on the multi-task learning strategy,the method uses fixed intervals to read video frames,shares the features maps of multi-frame video images and optical flow information between video frames,and then uses different tasks to further achieve two subtasks which include the detection of miner and the analysis of conveyor belt status.In the task of personnel detection,we have improved the SSD method to solve the problem of motions blur that may occur in videos.After the stage of the feature extraction,we wrapped the image features and optical flow information of the target frame and adjacent frames to improve the accuracy of the personnel detecetion model.In the task of the state analysis in conveyor belt area,we construct a two-stream Convolutional Networks that fuses the spatial semantic features of multi-frame images and optical flow to accurately determine the state of the conveyor belt.The method in this paper implements the collaborative analysis of the status of the conveyor belt and the detection of people,and conducts ablation experiments and comparison experiments on the video data of the actual coal-mining underground conveyor belt area.The experiment shows that,at the task of the detection,the collaborative analysis method based on the multi-task learning model we proposed in this paper can more accurately detect the position of people in the mine,and can more accurately determine the running state of the conveyor belt in the task of judging the state of the conveyor belt.2.We designed and implemented a software for detecting violation behaviors in it.For the inputted coal mine monitoring video,firstly,this software uses the data processing module to read and pre-process the inputted coal mine monitoring video.Then in the video analysis module,it uses the method which is proposed in this paper,to analyze the status of conveyor belt and person around the conveyor belt in the nearby surveillance.Finally,the analyzed video images are been displayed on the software interface,and if there is any illegal behavior,it will mark the detection results and give a prompt at the same time.The software can monitor the person in the conveyor belt area and the status of the conveyor belt,and can respond and prompt relevant violations in a timely manner,which plays a great role in ensuring the safety of workers working near the conveyor belt.Meanwhile,it provided a foundation for the method that proposed in this paper to be further applied to the large-scale coalmine intelligent monitoring system,which can effectively promote and improve the safety of coalmine-production level. |
PDF Full Text Request