Abstract: Combining with the convex characteristics of each object in the world, this paper proposes a single object detection method in the static image based on the best optimization convex grouping. The basic principle of the best optimization convex grouping is discussed systematically. Then the implementation process is designed in detail. The whole process steps include the parameter setting of the Canny edge detector, linear fitting based on the edge points, how to structure the convex polygons and the best optimization decision method for convex polygon. The experimental results are shown that this method can detect the single object from the static image at any environment, and it has higher detection rate and detection accuracy. It is not affected by the quantity and quality of sample in the machine learning, so it has a better general application for any single object and any environment.

Key words: convex attribute, convex grouping, best optimization decision, object detection

摘要： 针对不同场景下静态图像中单目标的检测问题，结合自然界各个目标特有的凸属性特点，提出了一种基于最优化凸分组的目标检测方法。比较系统地论述了最优化凸分组的基本原理，介绍了详细的实现过程，主要包括Canny边缘检测参数的设置、基于边缘点的线段拟合、凸分组中凸多边形的构造以及最优化凸多边形的判定。实验结果表明，该方法对任意场景下的单目标检出率和检测准确性良好，结合目标凸属性的最优化判定方式具有检出速度快，且不受机器学习中的样本数据影响的特点，具有很好的普遍适应性。

关键词: 凸性, 凸分组, 最优化判定, 目标检测