Research On Indoor Occupant Counting Method Based On Active Infrared Intrusion Detector

The number of occupants in each area of the building is an important factor for intelligent operation decision of building equipment and energy saving decision of building operation in intelligent building.The counting method based on infrared detection technology are concerned by academic research and industrial application for its advantages of low cost,fast response,easy installation and privacy protection.Faced with the needs of the realization of the group intelligent building system platform,for the indoor personnel counting requirements based on infrared technology,This thesis aims at whether the entrance and exit are separated,the design and implementation of the device,the design and performance verification of the occupant counting method,and the design and implementation of the application system are carried out.First,considering that in typical infrared occupant detection devices,the host computer and the front-end detector mostly use wire communication,with poor flexibility and wiring difficulty,this thesis designed an addressable active infrared intrusion detector identified by IP.The addressable active infrared intrusion detector gives the active infrared intrusion detector the ability to obtain an IP address.Based on a wireless network,a host can be connected to multiple detector devices through a network configuration which achieved the physical separation of the host computer and the detector.The flexibility of device deployment is significantly improved by the use of wireless data transmission without consideration of wiring issues.The experiment is based on the typical method of counting the number of occupants with 1-0 transition.The experimental results show that not only the host computer can accurately identify the detector position according to the IP address,the device is flexible in deployment,but also the detector can accurately perceive the situation of occupant entering and leaving the detection area and send the signal to the host computer accurately.Secondly,considering the errors in the process of signal acquisition or transmission,the counting results based on the idea of 1-0 transition of a single signal will be greatly deviated from the actual situation,based on sequence signal,this thesis proposed an counting method based on PCCS(partial clustering and classification system)for one-way traffic scenarios.This method used addressable active infrared intrusion detectors to collection the perception data.For the acquired data,through clustering to obtain the indoor occupant entry and exit pattern set contained in the sample data,then training the identification classifier based on the acquired occupant entry and exit pattern set and labels.Finally,the indoor occupants were counted according to the identification results.In the experiment,the combination of three clustering methods of K-means,SOM(Self Organizing Maps),DBSCAN(Density-Based Spatial Clustering of Applications with Noise)and two classification methods of support vector machine and decision tree are selected.Experimental results show that compared with the method based on 1-0 transition,the PCCS method can effectively eliminate the interference of signal errors,and the accuracy can reach more than 90% in the acceptable range of signal errors.Finally,different from the one-way traffic scenario,the two-way traffic scenario also needs to consider the direction of people entering and exiting.To meet the needs of occupant counting in two-way traffic scenarios,this thesis designed a dual-channel multi-beam active infrared intrusion detector device based on the addressable active infrared intrusion detector,and counting method based on LSTM(Long Short Term Memory)neural network is proposed and implemented.The dual-channel multi-beam active infrared intrusion detector realizes the perception of occupant's entry and exit behavior.The LSTM neural network completes the learning of the entry and exit behavior data to obtains the LSTM identification model.Finally,the number of occupants is calculated based on the identification results.The experimental results show that the counting method based on the LSTM neural network can accurately achieve the statistics of people in the two-way traffic scene,the accuracy rate is not less than 90%.The indoor occupant counting method based on active infrared intrusion detector implemented in this thesis can accurately obtain the number of occupant through the analysis of the occupant 's entry and exit behavior,and provides a new support for the intelligent operation of the automation system of building equipment in the intelligent system platform of insect intelligent building,so that whether it is the operation of building equipment or the decision of building energy efficiency,the number of occupant in the area can be used as a new constraint.Figure[50] Table[11] Reference[84]...