CAJ | 학술논문

针对交通标志检测中标志互相遮挡导致检测性能下降的问题，提出一种基于分水岭变换的互相遮挡标志自适应分离算法．基于RGB归一化阈值分割算法对标志图像进行二值化处理，然后构造区域轮廓特征矢量对二值图像中各个兴趣区域进行匹配，确定并提取互相遮挡标志候选区域Blob ．对提取的低维数Blob进行形态学膨胀处理，使不连续的边缘趋于连续，然后利用欧氏距离变换和分水岭变换寻求标志间分水岭脊线，利用脊线实现标志的自适应分离．实验结果表明算法取得较好的分离效果，在整个标志检测应用中，与现有算法相比，检测率提高了6．1％，处理速度提升了近3倍．
침대교통표지검측중표지호상차당도치검측성능하강적문제，제출일충기우분수령변환적호상차당표지자괄응분리산법．기우RGB귀일화역치분할산법대표지도상진행이치화처리，연후구조구역륜곽특정시량대이치도상중각개흥취구역진행필배，학정병제취호상차당표지후선구역Blob ．대제취적저유수Blob진행형태학팽창처리，사불련속적변연추우련속，연후이용구씨거리변환화분수령변환심구표지간분수령척선，이용척선실현표지적자괄응분리．실험결과표명산법취득교호적분리효과，재정개표지검측응용중，여현유산법상비，검측솔제고료6．1％，처리속도제승료근3배．
In view of the performance degradation of traffic sign detection due to mutually occluding signs ,an adaptive separation algorithm is proposed based on watershed transformation .Traffic image is segmented based on RGB-normalized thresholding algorithm and binary image is generated .Then ,a regional contour feature vector is constructed and used to match every interested region in binary image .Thus ,a Blob containing candidate mutually occluding traffic signs is determined and extracted from binary image .In order to make the discontinuous edges of traffic sign possess continuity ,the Blob with lesser dimensions is processed using morphological dilation operator . For the obtained Blobs ,watershed ridge line between the traffic signs is achieved by adopting Euclidean distance transform and watershed transformation ,which is used to separate the traffic signs .The experimental results show that the proposed method provides superior results to the existing algorithm ,improving 6 .1% in detection rate and about 3 times in processing speed .