Research On Posture Estimation And Micro-variable Magnification Algorithm For Human Thermal Comfort Visual Perception

With the rapid development of the global economy,the world’s energy consumption has increased rapidly.Building energy consumption has accounted for 21% of global energy consumption,while central air conditioning system(HVAC)energy consumption has accounted for 50% of building energy consumption.The current energy supply in a constant room temperature range brings energy waste to a certain extent,while ignoring human feelings.Real-time non-contact human thermal comfort perception is one of the effective ways to alleviate energy waste and realize "people-oriented" intelligent buildings.Based on the method of visual sensing,this paper studies the key technology of non-contact human thermal comfort perception from the perspective of human posture and skin texture.The main contents are as follows:This article starts from the perspective of human posture and conducts a questionnaire survey based on Fanger theory.Through a large amount of subjective data,the correlation between the posture of the human body and the thermal comfort of the human body is verified.On this basis,the key points of human bones are captured based on the key point detection library Open Pose,and then a pose estimation algorithm for human thermal comfort visual perception is constructed.Finally,16 subjects are invited to participate in the experiment,the pictures containing the subject’s current pose information are obtained by a common camera.These pictures are imported in the algorithm and analyzed,the test results are output in real time to verify the effectiveness and robustness of the algorithm Greatness.This article explores the visual perception of human thermal comfort from the perspective of skin texture.The architecture of the human thermal comfort perception method is designed based on micro-variation amplification,which is mainly divided into two modules: 1)The micro-variable magnification algorithm is optimized to realize the amplification and enhancement of small changes;2)Based on the micro-variation amplification algorithm,the skin texture(such as pores)is amplified,and then the thermal comfort of the human body is estimated.In view of the complexity and subjectivity of human thermal comfort estimation,this paper mainly studies module 1,which is to design an unsupervised micro-variable amplification algorithm for human thermal comfort estimation.This paper uses the continuous characteristics of video in time to build a deep learning network to separate the background information from the motion information.After the motion amplification is completed,it is merged with the background features to regenerate the amplified video.This article uses three types of data,synthetic data,dynamic data(network video data),and static data(actual scene collection)to train the network.Data verification shows that the algorithm can amplify small changes in the video to a certain extent.