Research And Realization Of Target Detection Methods Based On Cooperative Targets

Target detection technology is one of the pillar techniques to realise the measurement of targets' position in omputer vision region.Target detection technology involves many arts,including wisdom city monitoring,intellectualized household,signal processing,industrial production,spacecraft TT&C,UAV test&track,etc.Whether the pixel coordinates of target dection in the image is precise makes a great effect on the measurement accuracy of the whole mono-vision measuring system.This paper has carried out the following research work:(1)Developing status of mono-vision measuring system,target detection and identification technology,dectection based on cooperative targets and non-cooperative targets are reviewed and summarized in the beginning of this thesis.(2)A target based on luminos point is designed and detected.The pixel coordinates of the single luminos point target in the image is found based on the algorithm of threshold processing and statistical way;As for three luminos point target,the algorithm of threshold processing,statistical way and pseudo-color technology are used.What is more,the algorithm of back projection is used for targetting luminos as well.In addition,a compare is made between this two algorithm.(3)A chessboard target is designed and detected.The chessboard which is used during camera calibration is adopted as a cooperative target.Image denoising,histogram equalization,self-adaptive threshold,imaging contour extraction,approx-poly of rectangle,seed filling algorithm based on 4-neighbor,harris corner detection are used to dectect the corner coordinates of the chessboard target and a flow chart is drawed based on said algorithm.It is proved through experimental result that the algorithm to dectect the pixel coordinates of chessboard inner corners satisfied the real-time need.What is more,some improvements of the algorithm of target detection based on chessboard is put forward.A sub-pixel extraction of the corners' coordinates in the image is adopted to improve precision of coners' coordinates to make sure the precision of the whole vision measuring system;Robustness of the algorithm is improved by the adoption of continous frame images processing in a video;An algorithm based on the chessboard's rough location together with the method of perspective transformation is used to enlarge the range of the said algorithm's testing angle.(4)Comparisons is made between the luminos point target and chessboard target in terms of the robustness,rel-time performance,precision.Targets' detection algorithms are not only realised in the develoment environment of Microsoft Visual Studio 2010 together with OpenCV2.4.10,but also the testing results are showed by the graphic interface MFC(Microsoft Foundation Classes).