Research On Application Of Image Enhancement And Recognition Technology Based On Deep Learning In Oilfield Work Area

The safety of the production and construction of the oil field operation area plays a decisive role.Intelligent monitoring and management is a big step in production,which not only saves labor costs,but also significantly improves work efficiency.Based on the background of such a subject,this paper starts with the face verification of the staff in the oilfield work area,an image enhancement network and a face recognition network based on deep learning are investigated and designed,and finally builds up a complete face recognition system to be applied in the oil field.On the video surveillance platform,The main content of this article includes:First,the research of deep learning was carried out.Based on the research of outdoor image enhancement algorithms DSLR-Quality Photos on Mobile Devices with Deep Convolutional Networks and WESPE:Weakly Supervised Photo Enhancer for Digital Cameras,this paper improves and designs the image enhancement algorithm model based on deep learning in this paper.First of all,in the design of the generative confrontation network,in order to solve the problem of insufficient brightness,the brightness attention model is added to solve the problem of insufficient image brightness;second,the combination of local residual and global residual learning is used to improve the performance of the network;Finally,the multi-scale convolution kernel is used to extract image feature values to obtain richer and more delicate features of the image at different scales.Then this paper compares the HE,Dong and WESPE algorithms from a qualitative and quantitative perspective,and finally obtains an image enhancement model with a peak signal-to-noise ratio of 29.68,a structural similarity of 97.8%,and an overall enhanced visual effect.Secondly,according to the process of face recognition,this paper conducts the research of face detection and face recognition,including the classic MTCNN face detection model and Inception-Res Net-v2 convolutional neural network.And on the basis of the Resnet50 network combined with the excellent points of Inception-Res Net-v2 to improve and design the face recognition convolutional neural network of this article.At the same time,a variety of model evaluation programs were designed to verify the network model from a qualitative perspective.From the experimental results,the recognition rate of the Resnet 50＿3 inception＿dropout network can reach 98.44%,which is 6.64% higher than Resnet50 and 5.74% higher than Mobilenet.Finally,in order to adapt to actual oilfield applications,the author self-photographed thousands of on-site pictures of oilfield operation areas and thousands of staff face pictures on the oilfield site to create a unique Karamay data set in this article.During the experimental verification of face recognition in the oilfield work area,I built a face monitoring platform similar to oilfield video surveillance in the school laboratory,and used the Karamay data set to conduct field tests,and finally established an effective,recognition rate as high as The 96.4% face verification system has laid a solid foundation for the intelligent management of oilfield operation areas.