Research And Implementation Of Intelligent Monitoring System In Construction Site Based On Image Recognition

The electric tower construction sites, construction personnel safety issues must pay close attention. Because when you build the tower, there will be a variety of security risks, such as the height of the hanging pieces, rotating cutter mill and so on. Construction personnel must strictly observe the safety regulations, wear safety helmets, high operating personnel must be safety rope. If the command staff specialize in construction site surveillance, due to environmental occlusion, such as the movement of persons to appear not monitor the area. Therefore, the introduction of video surveillance electrical tower construction sites, intelligent monitoring of the construction personnel, and the effort is more scientific approach.Face system is a subsystem of intelligent monitoring system, it took independent role,and can greatly Make the management more convenient. in the process of face detection and recognition, the complexity of the light and the scene is the key problem for study. the pedestrian area is the premise when the monitoring of constructor..after we extract the pedestrian area, we can detect the helmet,So technologies of the system, At the same time, in order to make the supervision personnel immediately get the information and timely deal with the danger, It is very important to develop alarm software.It is a hot spot to develop software Based on the Android platform,the diversity of Android devices and the lack of communication in the construction site, are the problem for developing the alarm software.Based on the intelligent monitoring system on the construction site, in view of the related technology research, this article mainly working content is as follows:(1) Skin color segmentation between video frames and Haar-like Ada Boost for face detectionTraditional Haar- like Ada Boost face detection algorithm under the Mul-light condition, has a high miss rate,and also can appear mistakenly identified under complex scene face. the skin color segmentation method can simplify the complicated background, we can weed out the unqualified video frames through The relationship between frames, and speed up the detection.Template matching is used for secondary authentication,so we can obtain clear positive face.(2) The weighted LBP histogram combined with the nearest neighbor classificationLBP operator is based on the local characteristi C/S, with rotation invariance, and can overcome the influence of illumination in face recognition, this article proposed adaptive weighted LBP histogram based on the traditional LBP. According to the three division five eyes and the helmet shade in ctual project,we can divide face into 2 * 5 blocks, coarse positioning the important characteristi C/S such as eyes, nose and mouth, and find the weighted coefficient of various regions, then get the weighted feature histogram, finally use the nearest neighbor classification algorithm to classfy the face.(3)The development of face recognition systemAccording to the algorithm proposed in this paper, the design face recognition system, realize the core function such as face recognition and the connection with mobile terminal equipment for transporting the result of check on work attendance(4)HOG feature fusion LBP feature of nonlinear SVM pedestrian detectionBased on the current classic algorithm HOG + SVM human detection, combined with the current human shelter in actual project, the recognition effect has certain influence.in the SVM classification, this paper puts forward using nonlinear classifier, samples are divided into three categories- complete human sample, partial shade, non humanoid,then collect characteristics from the samples. And keep out that the effect for human classification is better when the samples are partial shade and detected by LBP feature(5)Android platform for intelligent alarm software development based on C/S(Client/Server) structureBased on the practical application of the project, this paper design the Android alarm software within the wireless local area network(LAN), under the C/S architecture, guarantee the stability communications of mobile with monitoring information platform, and achieve the goal of real-time alarm.In this paper, based on the research of intelligent monitoring system related technology to the construction site, test material is field of real-time video. the experimental results than other method is improved, and the project of field experiment, basic meet the requirements.