基于阈值分割与边缘检测的对象轮廓提取方法研究

Approach to Object Contour Extraction Based on the Threshold Segmentation and Edge Detection

为了提高智能视频监控系统中对象检测算法的检测准确性,实现对检测对象轮廓的准确提取,在分析目前常用于获取对象轮廓形态的对象检测方法不足的基础上,提出了基于阈值分割与边缘检测的对象轮廓提取方法。该方法需要阈值而又不依赖于阈值,选取任一阈值对检测对象进行阈值分割,再结合Sobel边缘检测以及经过定制的边界跟踪算法,实现对检测对象轮廓的提取。经实验得出的轮廓检测结果在不同阈值的条件下都呈现出较好的完整性与一致性。因此,方法具有较好的鲁棒性,实现了对检测对象轮廓的完整提取,提高了对象检测算法的检测准确性。

In order to improve the veracity of the object detection algorithm in intelligent video-surveillance systems, the paper puts forward the approach to object contour extraction based on the threshold segmentation and edge detection after analyzing the deficiency of the common methods of object detection. The method needs threshold but no depending on it, by switching any value of threshold to segment the detecting object, and then using the Sobel operator and the customized edge detection to extract the contour of the detecting object. The result of the contour obtained by the approach is integrated and consistent when using different threshold. So, it can get the conclusion that this approach has better robustness and can get the contour of the object integrality which improves the veracity of the object detection algorithm.