Human action classification method based on video local feature dictionary

Abstract

The present invention discloses a human action classification method based on a video local feature dictionary. The method comprises a step of extracting a local feature from a training video with a category label, wherein a feature package is formed by the feature vector set in each segment of video, a step of grouping feature packages, using a multi-instance learning method to learn a local feature classifier, using a mode of cross-validation in the multi-instance learning, and marking multiple instances with largest ranking in each package as positive instances in updating the positive instances, a step of taking a learned classifier as the dictionary of feature encoding, and using a maximum pooling method to pool the local feature response to obtain the global vector expression of a video, and a step of using a global feature vector to learn, obtaining the classifier of each action category, and using the classifier to classify the action in a new video. According to the invention, the improvement of the accuracy of estimation is facilitated, the memory of an initial value by classification is avoided, and the accuracy of estimated positive samples is ensured at the same time.

Description

technical field [0001] The invention belongs to the technical field of automatic video analysis, and relates to a method for automatically classifying or recognizing human body movements in videos, in particular to a method for pooling feature sets based on human body movements or scene local features extracted from videos The dictionary, and then use the pooled global vector representation to classify the action. Background technique [0002] Action recognition or classification can be applied to many fields such as detection of video surveillance events, video retrieval, human-computer interaction, and behavior understanding. However, due to factors such as camera movement, complex scene changes, changes in perspective, changes in action speed, and occlusion of human body parts, the visual differences of similar actions are relatively large, and there are high similarities in different types of actions. A great challenge has come. Due to the poor robustness of directly o...

AI Technical Summary

Problems solved by technology

There are two problems in traditional dictionary generation methods: one is that it is difficult to measure the semantic similarity of high-dimensional features by using predefined distance functions such as Euclidean distance and cosine distance, and it is impossible to guarantee that similar vectors are clustered together; the other is local There are a large number of background features and features that are not related to category identification in the features. These redundant features participate in visual expression and increase the noise of the feature space. Traditional clustering methods cannot screen these features.

Images