Abstract: Aiming at the problem of multitarget tracking in video behavior recognition, due to the mutual occlusion of the target and the complexity of the environment background, we proposed a method to detect and locate active individuals in video by using a local pattern activity discrimination model, which solved the problem of low recognition accuracy caused by factors such as incomplete trajectory extraction or too complicated background due to occlusion in the trajectorybased method. The model was applied to the video segments of the UEFA Euro 2012 and the 2013—2014 season Spanish League One 38 matches associated with the goal event, and the average accuracy rate was 91.3%. The experimental results show that the method has high accuracy and practicability for the recognition of target behaviors in videos.

Key words: video behavior recognition, local spatiotemporal regression kernel, feature word bag model, activity discrimination