Elevator Passengers Face Detection Method Based On The HOG

With the development of society, science and technology level continuously improve.Urban complex appear constantly. In this case, the traffic flow in buildings changes increased, it is impact for elevator dispatch decision in a large building, and it is not conducive to achieve these index of energy saving, fast and comfortable service. This requires a real time data acquisition system of elevator traffic flow, this system can accurately calculate the elevator traffic flow characteristics of the building, and provides data basis for accurate elevator dispatch decision.In view of the above problem, this paper use of machine learning, through the depth of face detection and tracking technology, traffic statistics out of the elevator traffic flow effectively and for computer vision technology in the application of elevator traffic flow data collection provides a new train of thought, which has some reference value and application prospect.This main work of this paper is as follows:(1)By reviewing the relevant technical documents, the research object of the thesis is confirmed. This paper summarize the development of face detection technology in the area of computer vision technology and future trends, determine the research methods, Through HOG and improved gaussian process and made to human face detection and recognition, and using the depth tracking technology of elevator passenger data statistics.(2)Discuss the application of traffic flow data in various fields, summarize the characteristics of elevator traffic data, this explain that the traffic flow data is time sequence. On the base of reviewing the relevant technical documents, summarize several common methods of traffic flow data statistics,Finally introduced the weighing device, photoelectric detection and infrared detection device, purpose floor hall devices and four kinds of traffic flow data collection methods, such as computer vision, this lays the foundation for the further study of the content for later on.(3)The traditional spectrum algorithms are limited in face recognition issue.A novel face recognition method based on modified Gaussian Process Latent Variable Mode(GP-LVM) is proposed.Firstly, the probabilistic model of face manifold is established with the Gaussian Process;Secondly, the shared information and private information can be gotten by analyzing the GP-LVM,Finally, the face recognition can be achieved.(4)Analysis of the actual situation of inside of elevator and passenger in or out of elevator,target is tracked by average depth of target with no feature. This resolves the problem that the target is back to the camera and can not be captured the facial features, and provides a basis forthe correct data collection.Using the computer platform and Kinect collect the elevator traffic flow data, makes use of the traffic pattern recognition system to identify elevator traffic pattern,verifies the effectiveness of the algorithm.(5)Summarize the study of this paper, put forward some questions about when we reserach in the process and make a prospect to the next research.