Research On Detection Of Protective Equipment For Unloading Oil In Gas Station Based On Deep Learning

Gas stations are dangerous places to store petroleum fuels.Once a safety accident occurs,it will cause inestimable property losses.Therefore,in order to ensure the safety of oil unloading personnel,oil unloading work needs to be equipped with protective equipment.At present,the equipment status of protective equipment often depends on personnel inspection,but this method is heavy workload and low efficiency.With the rapid development of artificial intelligence,the target detection method based on deep learning is also gradually applied to the oil unloading scene of gas stations.Relevant theories and technologies are used to carry out safety inspection of oil unloading protective equipment of gas stations,which not only saves manpower but also improves efficiency.YOLOv3 is a relatively classical detection algorithm in deep learning.However,when it is used for the detection of oil unloading protection equipment in gas stations,YOLOv3 network has some problems such as poor accuracy and complex model.Therefore,how to avoid the detection of small targets,improve the identification accuracy of targets,and reduce the complexity of model calculation.It is an important research direction of the detection method of protective equipment for unloading oil in gas stations.In this paper,the mainstream target detection algorithm YOLOv3 in industry is studied in depth,and a set of detection system based on YOLOv3 network is proposed to complete the following aspects of work:(1)Aiming at the problem of low accuracy caused by omission and misdetection of small targets in YOLOv3 network,this paper proposes a detection method of YOLOv3 protective equipment based on cross-stage receptive field.Firstly,by introducing the RFBS receptive field module,the model can adjust the receptive field adaptively,and achieve the matching of targets at different scales,so as to improve the detection accuracy of targets.Secondly,the CSP network is used in the feature extraction network of YOLOv3 to optimize the repeated gradient information,reduce the number of parameters of the network model,and reduce the computational complexity.K-means++ is used to re-conduct cluster analysis on the 9 types of targets in the picture to determine the appropriate network anchor parameters.The experimental results show that the accuracy of YOLOv3 network based on the cross-stage receptive field is greatly improved,and the detection effect and recognition accuracy are obviously improved.(2)In view of the problem that the detection rate is too slow due to the complex network model in the detection method of oil unloading protection equipment proposed above,a detection method of oil unloading protection equipment based on YOLOv3-MCRS is proposed.The lightweight convolutional neural network Mobile Netv3 is introduced into the backbone network of YOLOv3,which effectively reduces the number of network parameters and improves the detection speed of the network.To solve the problem of inaccurate frame loss regression,CIOU regression loss function is used to replace GIOU regression loss function in the network,so that the coincidence degree of prediction frame and real frame can be improved.The comparison experiment shows that the YOLOv3 network based on lightweight backbone not only has a faster detection rate,but also consumes less device memory.(3)Design and complete a detection system for oil unloading protection equipment of gas stations based on YOLOV3-MCRS network,put the proposed Detection algorithm of YOLOV3-MCRS into practical application,and build a detection system for oil unloading protection equipment of gas stations combined with Py Qt5 design software.Through real-time detection of video streaming images,Master the equipment of oil unloading protection equipment and give early warning in time,so that the personnel can make effective protection measures in oil unloading work.