手势特征提取与图像分割的优化研究

手势识别的关键是手势分割。在现实应用中,手势图像因光照强度和复杂背景会存在不同程度的畸变,致使手势分割的精确程度降低以及分割消耗的时间增长。针对这一弊端,提出改进传统手势识别的思路。对手势进行主要特征点的提取,将多个先验图像减至单个先验图像,缩短了手势分割所需的时间;将YCBCR色度空间和粒子滤波跟踪算法相融合,更新参数阈值,优化边缘去噪。该方法将手势分割的准确度提升至96%以上,用时降至2ms以内,精准程度高,所耗时间短,在复杂环境下的识别效率高。

OPTIMIZATION OF GESTURE FEATURE EXTRACTION AND IMAGE SEGMENTATION

The key of gesture recognition is gesture segmentation.In practical applications,due to the illumination intensity and complex background,gesture images have different degree of distortion,resulting in the reduction of the accuracy of gesture segmentation and the increase of segmentation time.In view of this disadvantage,this paper puts forward the idea of improving the traditional gesture recognition.We extracted the main feature points of gesture,and reduced the number of prior images to a single prior image,which shortened the time needed for gesture segmentation.The YCBCR chromaticity space and particle filter tracking algorithm were combined to update parameter threshold and optimize edge denoising·The accuracy of gesture segmentation is improved to more than 96%,and the time is reduced to less than 2 ms.The accuracy is high,the time is short,and the recognition efficiency is high in complex environment.