Due to the influence of illumination and environmental factors,there will be parallax interference in the image in non-rigid motion,which leads to the poor effect of non-rigid motion feature extraction.Therefore,a feature extraction method for non-rigid motion is proposed.The non-rigid moving image is pro-cessed by Gaussian filtering,and the initial feature points are extracted by feature description operator in the Gaussian difference scale space.According to the result of feature point extraction,the cost of non-rigid motion feature extraction is calculated.Guided filtering method is introduced to optimize the parallax effect of non-rigid motion images.Under the constraint of minimum parallax,the features of non-rigid motion are accurately extracted according to the confidence aggregation and propagation strategy.Taking Chinese Wushu as the experimental object,the experiment results show that the feature point extraction accuracy of the proposed method is high,and the parallax extraction error rate does not exceed 2.0％.
关键词
特征点提取/非刚体运动/最小视差约束/特征提取/武术运动
Key words
feature point extraction/non-rigid motion/minimum parallax constraint/feature extraction/Chinese Wushu
维普
万方数据