基于人脸检测和关键点识别的快速人体组件划分

FAST SEGMENTATION OF HUMAN BODY COMPONENTS BASED ON FACE DETECTION AND KEY POINT RECOGNITION

人体组件划分用来检测视频帧或静态图像中的人体,并将其划分为头部和四肢等独立区域,是人体精确三维重建和动作识别等后续研究的重要基础。提出一种新的人体组件划分算法,算法主要针对无肢体重叠的人体图像,首先利用人脸检测技术快速定位人体大致范围,再通过边缘检测获取准确的人体轮廓,最后设计并使用高效的十字链表存储和检索方法,完成基于轮廓关键点查找的人体组件划分。实验表明该方法具有较好的实时性和准确率。

Segmentation of human body components is used to detect the human bodies in static images or video frames. It separates the human body into individtial regions such as the head, limbs, legs and etc. , and is an important basis in regard to the follow-up researches including the precise three-dimensional human body reconstruction and motion recognition. We present a new segmentation algorithm for human body components, which is mainly for images of human body without overlapped limbs. First, we use face detection technology to quickly locate the approximate position of the human body. Then we obtain accurate body contour by Canny edge detection. Finally we design and implement the efficient storage and retrieval method in orthogonal list format, and achieve the body components segmentation which is based on finding key points of contour. Experiments show that this method has good real-time property and accuracy.