Leaf Classification And Gait Recognition Based On Improved Active Shape Model

The paper based on the ASM idea of the algorithm, the use of improved algorithms and procedures for classification of leaf shape and gait, in succession to punctuation active shape model to the shape outline of the object sequence that, by the normalized sequence to study the statistical relationship between the shape of inter- similarities and differences, and identify with the recognition rate to reflect the effect of these three basic characteristics of the same time, gait recognition for the plant leaves and the characteristics of the model was improved:increase in coarse classification of the independent parameters and steps to identify from non-normal interference. Leaves, respectively, during the classification process applications cover a point of reference for the radius and biological transformation and rotation transformation scalable, so that the normalized biological characteristics more in line; in the classification by setting the similarity threshold, and strengthen the similar standards, so that reducing the number of false classified non-similar incidents; final design of the three groups of experiments to study changes of each factor independent of the recognition rate of the impact of these improvements, we realize batch identification and classification of plant leaves, provides a the wrong way to check the distribution points and pointed out that research to improve the recognition rate. During gait recognition, silhouette segmentation first, fill, normalized pre-treatment, according to Lee proposed method the human body divided into seven different sub-regions, were calculated centroid of each sub-region to establish active shape model; in Select the feature extraction algorithm will consider two:(1) changes in the factors that half of limb drop out, just consider the legs, especially the front legs of the change; (2) changes in systemic factors. We view the side of Chinese Academy of Sciences database NLPR database classification trying three different methods, namely nearest neighbor classifier (NN), K-nearest neighbor classifier (kNN), and with the recent specimen classifier (ENN) to distinguish between different samples. In this paper, a good gait, presented the program results, which have good prospects for the study.