基于合作目标的图像分割和提取方法研究  

Research on Image Segmentation and Extraction Method Based on Cooperation Target

作　　者：任明 吴川[1] 高云龙 彭佳琦 赵春蕾[1] Ren Ming;Wu Chuan;GaoYunlong;Peng Jiaqi;Zhao Chunlei(Changchun Institute of Optics,Fine Mechanics and Physics,Chinese Academy of Sciences,Changchun 130033,China;First Military Representative Office in Changchun,Changchun 130000,China)

机构地区：[1]中国科学院长春光学精密机械与物理研究所,长春130033 [2]驻长春地区第一军事代表室,长春130000

出　　处：《机电工程技术》2022年第6期87-91,共5页Mechanical & Electrical Engineering Technology

摘　　要：为准确快速从全局图像中提取合作目标图像,提出了将图像区域分割筛选后对各个区域进行模板匹配的方法。针对传统Canny边缘检测算法边缘连续性差的问题,引入形态学知识使边缘曲线更平滑。针对全局图像中存在干扰轮廓边缘而影响处理速度的问题,通过设置一系列的轮廓约束条件,筛选轮廓特征。基于模板匹配法设计目标识别提取方式,并针对模板匹配法是进行全局图像处理而影响检测速度的问题,提出了一种模板匹配法与目标位置信息预测相结合的检测方式。实验结果对比表明,所设计的边缘检测方法在连续性上更平滑,所提出的轮廓筛选方式具有明显剔除噪声轮廓的效果,设计的模板匹配法能够高效提取合作目标图像,正确率高达99%,识别时间为0.048 s,引入目标跟踪预测后的目标检测时间为0.03 s,进一步提高了合作目标的提取效率。In order to accurately and quickly extract the cooperative target image from the global image, a method of template matching for each region after segmentation and screening of the image region was proposed. Aiming at the problem of poor edge continuity in the traditional Canny edge detection algorithm, morphological knowledge was introduced to make the edge curve smoother. Aiming at the problem of disturbing contour edges in the global image and affecting the processing speed, a series of contour constraints were set to screen contour features. The target recognition and extraction method was designed based on the template matching method. Aiming at the problem that the template matching method performs global image processing and affects the detection speed, a detection method combining the template matching method and the target position information prediction was proposed. The comparison of experimental results shows that the designed edge detection method is smoother in continuity, the proposed contour screening method has obvious effect of eliminating noise contours, and the designed template matching method can efficiently extract cooperative target images, with a correct rate of up to 99%, the recognition time is 0.048 s, and the target detection time after introducing target tracking prediction is 0.03 s, which further improves the extraction efficiency of cooperative targets.

关 键 词：模板匹配 合作目标 边缘检测 运动跟踪 

分 类 号：TP29[自动化与计算机技术—检测技术与自动化装置]