摘要： 图像中不规则物体长宽属性较难统一,且传统的检测算法不够精确。针对该问题,提出一种自动定位检测算法。该算法通过目标主轴方向估计求出珠宝最小外接矩形,根据主轴确定目标的初始矩形,以初始矩形的几何中心为旋转中心对其进行旋转,从而找到最优矩形姿态,并平移优化旋转后的矩形,实现对珠宝轮廓的最终定位。实验结果表明,与现有检测不规则物体最小外接矩形的算法MBR相比,该算法对于珠宝轮廓的定位更加准确。

关键词: 图像处理, 最小外接矩形, 主轴, 形状定位, 珠宝轮廓, 目标识别

Abstract: The length and width of the irregular object is difficult to be unified in an image,and the traditional detection algorithm is not accurate enough.For this problem,this paper states an automatic location detection algorithm.In order to obtain the minimum bounding rectangle of the jewelry,firstly it estimates the target direction of the principal axis and the initial target rectangle related to the principal axis,following by finding out the optimal rectangle profile with the center of rotation formatted by the initial rectangle’s geometric center.It performs translational optimization on the rotated shape of rectangle to achieve the final positioning of the jewelry outline.Experimental results show that this algorithm is more accurate than the existing minimum rectangle algorithm MBR for the detection of irregular object in capable to locate the exact jewelry outline.

Key words: image processing, minimum bounding rectangle, main shaft, shape positioning, jewelry outline, object recognition

中图分类号: 

• TP391