Study On The Paper Detection And Localization Technology

With the consequent popularity of cameras, the interaction based on the cameras hasbecome the research hotspot on human-computer interaction field. The paper detection is thebasis of the pen-and-paper interaction system, and only the accurate paper location canimprove the recognition quality of the information on papers. This thesis be emphasis onstudying the paper detection and location methods about interaction system based on thecameras.Through studying the skin-detection algorithm of the Open CV, this thesis found that theparameters in the algorithm can be modified to detect papers, and it can detect papersaccurately under the condition of stable light.In the paper detection of Line segment detection algorithm based on LSD, wecontrasted and analyzed the Canny Operator, the Hough Transform and the Line segmentdetection algorithm based on LSD, and found that using the LSD line segment detectionalgorithm to extract paper images can achieve much better effect and faster efficiency. On thebasis of LSD line segment detection, computing the each line direction and normalization,calculating the biggest outline of paper image to gain the paper outline, and moreoverutilizing the paper metric information to detect four vertexes of paper, and rectify the papers.Owing to the paper lines extract from the paper detection based on LSD line segmentdetection algorithm are mostly tiny and interlaced short lines, this thesis puts forward a newcluster algorithm to revise these lines. The algorithm firstly utilizes the Hough Transformalgorithm or LSD algorithm to make lines detection of original image, then classifies theselines according to the line direction, and on the basis of the classification, we can make moredetailed division to each constituted segment according to the distance difference, and clusterthe far-away lines to different segment. Finally merging and separating the lines based on therelationship of lines, and forming the finally clustering effect. On the basis of this algorithm,it proved that the paper detection effect is much better, and the detection efficiency is muchhigher.