办公建筑室内人员在室情况识别方式的研究与应用*
Research and application of identification method of indoor occupancy in office buildings
摘要：
有关建筑人员行为的研究主要分为人员位移与人员动作2个层面，研究与应用该领域成果的前提为相关行为数据的获取。人员位移为实时的室内人员在室情况。针对该研究领域的相关需求，认为人员在室情况的数据获取应具有动态地、自动地识别室内人员数量及位置的功能，并有效保障室内人员的隐私。基于图像分类技术及骨骼追踪技术，开发了连接Kinect体感传感器的动态室内人员在室情况识别程序（DIOIK），并通过4 h的视频验证，得到其平均识别准确率为93.57%。
Abstract：
The research on occupant behavior is mainly including two aspects: occupancy and occupant’s action. The premise of the research and application is the acquisition of behavior data, and regarding the occupancy is the real-time indoor occupancy. In the view of related needs of this research domain, proposes that the acquisition of occupancy data should realize the dynamic and automatic identification of the number and location of occupants, and effectively protect the privacy consideration. Furthermore, based on the image classification and skeletal tracking technology, develops the dynamic indoor occupancy identification program with Kinect connection (DIOIK), and the average identification accuracy is 93.57% through a 4 hours of video verification.
Keywords:occupantbehavior,occupancy,imageclassification,skeletaltracking,occupants’privacy