| Intelligent video analysis technology has been widely used in nowadayssociety.Intelligent surveillance,target detection,human feature recognition belong to itsapplication area. Target Detection Technology, Video-based Positioning Technologyand HumanBehavior Recognition Technology are three important technologies ofintelligent video analysis. In this paper, the three technologies above are taken to thefollowing research work specifically with practical application scenarios:1Target detection technology designed for metal components surface defects. Firstlist the technical difficulties in the technologyimplementation process.With the researchcase battery surface defect detection, theimplementation method of each step of thistechnology is discussed. In the research case,a camera is used to capture the image ofnegative electrodesurface of batteries.Video analysis method is used to determine theexistence of defects on surface of the batteriesand extract the defective region. In thevideo analysis process, Harris corner detector is used to mark surface defects, thenintensive degrees discriminating method is used to filter off pseudo-defective marksThenumber of defective marks is taken account to determine whether the measured batteryis defective at last. According to the results of the research case, the technical essentialsof the target detection technology designed for metal components surface defectsaresummarized.2Video-based positioning technologyfor small targets inclose-range:First list twomajor technical difficulties in its implementation process. Taking the positioning ofcontact pointin video-based whiteboard application development as the research case,the conversion method fromtarget positionin image to its spatial positionis focused on.In the research case, Dual cameras placed in specific location are used as the imagecapture device.Color threshold segmentation method and frame difference method areused to recognize the contact point. Polynomial fitting methodis proposed toapproximate the error in projection process.The spatial position of contact point isacquired by seeking intersection point of two light lines. The hand-write track is shownon the screen at last. According to the results of the research case, the technical essentials of Video-based positioning technology for small targets inclose-rangearesummarized.3. The developmentof the natural person-machine interactive basedondepth-basedhuman behavior recognition technology. Kinect camera is used as thevideo capture device to provide the depth of information of video with OpenNI SDK.By determining the positional relationship of the human joints point,three kind humanactions (Beck, Left Arm Lift up, Right Arm Hem) are designed toexpress four kindcontrol orders (Stop, Show Last Page, Show Next Page and Recovery). A screen showcontrol system to let people control the slideshow of the big screen by their humanactions is developed using these control orders. |
PDF Full Text Request