机构地区: 华中理工大学图像识别与人工智能研究所
出 处: 《华中理工大学学报》 1994年第5期15-19,共5页
摘 要: 提出一种局部遮挡目标的识别方法。该方法利用多边形近似方法将目标轮廓分段,抽取目标轮廓每一小段的不变矩作特征,利用动态规划技术在段间距离表中寻找最佳路径,确定所有的匹配对,这些匹配对序列构成了目标与模型的局部匹配,可从局部匹配中计算匹配可信度。该方法简单、有效、抗噪声能力较强,是一种较好的局部目标识别方法。 A method for recognizing partially-occluded objects is described .The object contour issplit into curve segments by a coarse-to-fine polygon approximation algorithm and the mo-ment invariants of each segment are extracted as the feature vector。The best path in the in-tersegment distance table is then found using the dynamic programming technique. All thematching pairs that constitute the partial match between a model and an object are deter-mined.The confidence level of matching is calculated from the partial matches.
领 域: [自动化与计算机技术] [自动化与计算机技术]