Research On Key Techniques For Video Surveillance System Based On BACnet

One of remarkable development trends of building automation and control technology isthe application of uniform data communication network. Based on combining fields of buildingautomation and control network BACnet, image processing and artificial intelligence, somekey technologies about the building automation and control system (BACS) and its sub-system,Video Surveillance System (VSS) are studied here. The work of this paper is supported by thefund of "Building Device Automation System" from the Ministry of Construction of China (No.03-2-043) and the fund of "Research on Intelligent Control Technologies Based on Agent"from the Key Laboratory of Education Ministry for Image Processing and Intelligent Control(No. TKLJ0116). Simultaneously, comprehensive cooperation has been made with the BACnetStandard Project Committee and its related organizations, BMA (BACnet ManufacturersAssociation) and BTL (BACnet Test Laboratory). The major work of this paper is listed as following: Firstly, to supply the lack of the specification about VSS in BACnet standards, theproposal of the extension of BACnet about VSS is put forward here. The complexities of thecontrol of the video sources and the transportation of the video streams are taken into account.By designing new video related objects and services, a new schema of video surveillancesystem based on BACnet standard is implemented. By this schema, VSS is bound with othersub-systems based on BACnet more tightly and the ability of inter-operation is enhanced. Theproposal is also a meaningful supplement for BACnet. Secondly, attention is centered on the video analysis and processing in VSS. To break thespeed bottleneck in video processing, a novel fast block motion estimation (BME) algorithmSSC-APDS is presented here, which combines subsamplings on search candidates and pixels.Compared with APDS, one of the latest BME algorithms, SSC-APDS can maintain its MSE(mean square error) performance very close to that of APDS and improve the search speedremarkably (about 30% increase). And it can perform better than many other algorithms both inMSE and speed. SSC-APDS can be applied in many areas including video analysis,videocompression etc. Afterward, two difficulties about the object detection and tracking withmoving background in VSS are studied. One is the complex moving background registration. Anew image registration approach is presented based on multiple line correspondences, which isespecially efficient for the registration with rotation. A modified local Hough transform (MLHT)is proposed to look for line correspondences and get transform parameters. This approach canmake background registered and moving object detected efficiently and overcomes the III华 中 科 技 大 学 博 士 学 位 论 文uncertainty caused by some other ways such as point-point, point-line correspondences andimproves the registration speed rapidly. The other is the object tracking in complex movingbackground. Aiming at the speed bottleneck of real-time object tracking based on PC, a newobject tracking algorithm based on BME is presented here. Some blocks are selected from thearea of the object, and SSC-APDS is used to attain the displacement probability matrixes ofthem. Then through a time filtering and an elimination algorithm, the displacement of theobject is obtained. This approach can achieve robust tracking with high speed. Thirdly, to introduce the agent technology into the uniform building automation controlnetwork BACnet, a new agent communication channel is proposed and implemented byemploying the File object and related service AtomicWriteFile of BACnet. Every agentplatform (AP) has one specified File object and agents in another AP can communicate withagents in this AP by using AtomicWriteFile service. Most existing agent communicationprotocols such as HTTP, RMI and CORBA use TCP/IP as their communication bedrock, whichrestricts the development...