Face Recognition And Behavior Warning System In Industrial Monitoring

With the advent of technologies such as cloud computing,big data and artificial intelligence,these new technologies have had a huge impact on the patterns of industrial production.As an important factor in the process of industrial production,human activities are highly autonomous and uncertain.Therefore,the construction of digital twin models around human location information,identity information and behavior information has become the key to the intelligent transformation of traditional factory management.In this paper,taking the factory monitoring video as data-driven,combined with the existing deep learning technology and the needs of factory management,the author proposes a design scheme for the system of face recognition and behavior early warning of personnel in the industrial monitoring.Firstly,this paper focuses on the quality of data sources in industrial monitoring from the aspects of hardware and algorithm.On the hardware level,this paper designs the resolution selection of monitoring equipment,lens selection,installation point position design,supplementary light measures and so on.At the level of image algorithm,this paper studies and verifies the de-noising effect of the Lucy-Richard algorithm for monitoring video noise.In order to solve the problem of low illumination of monitoring images,this paper proposes an improved fusion algorithm based on adaptive parameters combined with the advantages of histogram equalization algorithm and MSRCR algorithm,which can effectively improve the quality of images to be enhanced.Secondly,this paper designed the algorithm of face recognition and behavior warning.In the face detection module,this paper analyzes the factors affecting the computational efficiency in MTCNN.In this paper,an improved k-mtcnn face detection algorithm based on kalman filter is proposed in combination with the characteristics of correlation between video frames in surveillance video.Compared with the original algorithm,this algorithm has a significant improvement in real-time detection with little loss of precision.In the face recognition module,Facenet algorithm is used to extract face features,and a set of face recognition algorithm solution is proposed.In the behavioral early warning module,this paper summarizes several behaviors with obvious behavior characteristics in the specific scene of the factory.Combined with the YOLOV3 target detection algorithm,the paper gives the design of relevant early warning algorithm from the three aspects such as worker number detection,area intrusion and safety helmet detection.At the end of the paper,the architecture of the whole system is designed,and the data storage and interface specifications are designed.