Behavior recognition method and system based on middle-level features

Abstract

The invention discloses a behavior recognition method and system based on middle-level features. The realization of the method comprises: obtaining a candidate component detector set from a sample image sequence; removing component detectors which are b% of the candidate component detector set and have low discrimination capacities, and obtaining a new candidate component detector set; carrying out sorting in a descending order according to weights of all the component detectors in the new candidate component detector set, and selecting the P component detectors at the front of sorting as middle-level feature extractors of a Class-A behavior class; acquiring middle-level feature extractors of all classes in behavior classes, combining the same into a bag of words (BOW), utilizing the bag of words to extract the sample middle-level features of the sample image sequence, utilizing the sample middle-level features to train a classifier, and obtaining a behavior recognition classifier; and inputting a test image sequence to the behavior recognition classifier, and obtaining the behavior class of the test image sequence. According to the method and system, the recognition capacity is high, the recognition accuracy rate is high, the practicability is high, and the correlation between components is retained.

Description

technical field [0001] The invention belongs to the field of computer vision, and more specifically relates to a behavior recognition method and system based on middle-level features. Background technique [0002] Behavior recognition technology is the core technology in the application fields of video security monitoring, human-computer interaction, video retrieval and analysis, etc., and it has attracted more and more attention from industry and academia. However, behavior analysis in videos poses great challenges due to the great disturbances in behaviors, such as motion blur, scale change, low resolution, background noise, camera motion, and viewpoint changes. [0003] Existing methods mainly include the following two main lines: The first is low-level spatiotemporal local features, such as: spatiotemporal interest points, gradient-based features, and trajectory features. Typically, a large number of local descriptors are extracted from the video training set, then a "b...

AI Technical Summary

Problems solved by technology

[0006] Aiming at the above defects or improvement needs of the prior art, the present invention provides a behavior recognition method and system based on middle-level features, thereby solving the problem that the existing behavior recognition has weak recognition ability, requires a lot of manpower and material resources, is not practical, and is lost. Technical issues of associativity between components

Images