Design And Research Of Oil Leakage Detection Method Based On Image Processing

Target detection is mainly based on the research of the inherent characteristics of the object.Now it is widely used in many fields such as intelligent security,mobile payment platform,space recognition,navigation and reconnaissance.At present,there is a problem of oil leakage in the pumping process of the pumping unit in the oil field operation area,but the existing scheme can not carry out effective oil leakage detection,which will lead to oil leakage and economic loss.In view of this problem,this paper studies the oil leakage detection in the oil field operation area.The main work and innovative contributions are as follows:First of all,the research content of this paper involves many factors,such as safety,so it is difficult to obtain the scene video data,so the scene simulation method is used to make the data set to meet the oil spill scene.Secondly,the oil leakage detection is carried out by using YOLOV3 model and Faster RCNN model.The detection frame rates are 16.7fps and 20 fps respectively,and the detection accuracy is 84.22% and 81.13%.Because the detection accuracy does not meet the needs of oil field enterprises,the oil leakage detection accuracy of YOLOV3 is higher than that of Faster RCNN’s oil leakage detection accuracy,and then the principle analysis of YOLOV3’s oil leakage detection,it is concluded that the "misclassification" caused by shadow interference is the cause of low detection accuracy.Then,aiming at the problem of "misclassification",DSC-Net is introduced into the YOLOV3 detection model to eliminate the interference of shadow.After the introduction of DSC-Net,the detection frame rate and detection accuracy are 1.47 fps and 96.22% respectively,which are higher than the detection frame rate of 1 fps and detection accuracy of 95% required by oil field enterprises,meeting the needs of oil field enterprises;the detection accuracy is increased by 15.09% compared with the original scheme.Finally,according to the actual needs of the oil field,combined with the existing excellent algorithm,this paper designs a new oil leakage detection method,DSC Net + YOLOV3 oil leakage detection method,through the way of scene simulation to verify its performance on the self-made data set,and provides a set of real-time and high-precision oil leakage detection scheme for oil field enterprises by using remote video monitoring system.It has potential engineering application value.