Detection Of Intrusion Target In Oilfield Operation Area Based On Adaptive Mixture Gaussian Model

With the rapid development of intelligent surveillance video technology and the informatization reform and construction of oil field operation area,most of the monitoring system of oil field operation area still adopts traditional manual monitoring.However,the traditional monitoring technology has some shortcomings which bring about the hidden danger and the waste of manpower and financial resources to the operation area of oil field In this paper,several common detection algorithms used in previous researches on intrusion targets are analyzed,combined with the complex background in the focus range of one kilometer captured by the telephoto camera in the oil field operation area.Adaptive selection of K values based on GMM to reduce computational complexity can improve the detection efficiency of real-time intrusion targets and update the background learning rate? in the case of sudden illumination changes.To ensure the robustness and accuracy of intrusion targets and solve the problems existed in traditional monitoring technology for a long time,the security and efficiency of video surveillance in oilfield operation area are improvedAiming at the two disadvantages of GMM in detecting targets in oil field operation area,which are easy to be affected by illumination and large amount of calculation,this paper presents an improvement on adaptive value selection K and learning rate a in the case of light change or illumination mutation based on GMM.The basic idea of GMM is to adopt a fixed value K Gaussian distribution and a fixed background update rate a for each pixel point in the video image.In the actual oil field application scene,the traditional GMM cannot meet the monitoring needs of the oil field operation area because of the different weights of each pixel in the video image and the variation of the all-weather illumination.In view of the problems existing in the actual operation areas mentioned above,this paper puts forward some solutions:First of all,when the invading target passes through the operation area of the oil field,it will make each pixel point in the video image of the region change,and make a larger value K Gaussian distribution for each changed pixel point.For the small value K Gaussian distribution of pixel points without change,the adaptive selection value K can reduce the computational complexity of Gaussian algorithm and improve the real-time eff-iciency of detection.Then,the field operation area needs all-weather monitoring,because of the different illumination intensity at each time,so the background update rate a needs to be updated adaptively according to the different illumination changes.The updating of the background model is mainly based on the updating rate ?.When the operation area of the oil field encounters a sudden change of illumination,the gray scale of the video image will change significantly,and a threshold value should be set,when the change of the gray value is greater than the threshold,re-modeling the background model.When the threshold is less than the threshold,the background modeling can be updated adaptively according to the update rate.In this way,the illumination change problem in the operation area can be solved.The updating of the background model is mainly due to the difference of the learning rate ?.Because of the sudden change of illumination in the operating area of the oil field,the gray scale of the video image will be changed significantly,and a change threshold parameter will be set,when the gray value changes too much,The background model is modeled again,and the gray value updates adaptively according to the learning rate when the gray value does not change.Finally,in order to verify the monitoring effect of the improved GMM in the oilfield operation area,a simulation experiment is carried out using MATLAB2016a.The experiments show that the improved GMM meets the requirements of oilfield operation area in real-time and accuracy of intrusion target detection.