Research On Energy Efficient Lighting Systems Of Classroom Based On Image Recognition

With the development of economy,the overexploitation of the energy brought lots of problems in the accept of global environment and climate, which made more and more attention be paid to conserving resources and reducing emissions all over the world. The vast expense used in architectural lighting every year means not only a waste of money for each unit but also waste of much electricity and energy. There have actually been many studies on building efficiency, but the energy efficient system that applies sensor technology has not produced good results and is not widely used. Along with the development of the computer vision, image recognition technology and some other relevant technologies, this article presents the image recognition method to economy energy problems of illuminating of classroom.The method in the paper produces good results,which has important meaning to building efficiency.This paper first analyzes the functional requirements of the economy energy system of illuminating of classroom, and designs the implementation scheme. According to this scheme,it studies seat section detection, measurement of illumination, human object segmentation and human detection in classroom. In this study, considering that the video image of camera in classroom including other non-seat section, not only enlarged the detecting range of human, but also increased the computation, so the detection of seat section is presented in the paper. Set sections firstly are roughly marked based on the histogram of much pixels in classroom image,next,are precisely marked using level projection and vertical projection.The paper mainly uses the principle that gray scale image includes the information of luminance to measure the illumination. For the objects segmentation, it uses background subtraction method to achieve the segmentation of foreground from background,and different background model is used in different periods.Then improved erosion and expansion techniques are used for denoising colour image,which has preserved colour information.Human detection detection is the focus of research, so the paper uses the head feature to detect human based on the morphological posture of human in classroom. Based on classic Haar classifier algorithm to detect human, achieved an improved human detection based on connected labeling. Skin and hair are marked in different colours based on skin model and hair colour information firstly in the paper.Then,an improved connected labeling algorithm are presented to realize statistics on the number of people. The experiment results show that the improved connected labeling algorithm to human detection in the paper has a higher detection rate.