Signature behavior recognition method and system based on deep learning

Abstract

The invention discloses a signature behavior recognition method and system based on deep learning, and the method comprises the steps: receiving video data uploaded by a user, and extracting a plurality of to-be-recognized first images from the video data; according to a preset algorithm, sequentially detecting all the first images, and when it is detected that the sign pen area and the hand area in the current first image intersect and the sign pen rotation angle is larger than a preset angle, judging that the current first image is an effective image; performing behavior recognition on all the effective images in sequence, and obtaining a signature behavior video from the video data according to all behavior recognition results; wherein one effective image corresponds to one behavior recognition result. The first image with the sign pen area and the hand area intersecting and the sign pen rotation angle larger than the preset angle is screened to serve as the effective image, the situation that the behavior that the target is taken up and not signed is wrongly recognized as the signature behavior is avoided, and the signature behavior recognition accuracy is improved.

Description

technical field [0001] The present invention relates to the field of video data processing, in particular to a signature behavior recognition method and system based on deep learning. Background technique [0002] In order to protect the rights and interests of consumers, based on the requirements of regulatory agencies, financial and insurance institutions need to regulate the sales behavior of financial and insurance institutions through audio and video recordings when selling wealth management products and consignment sales of financial products. At present, financial and insurance institutions usually cache video files locally and asynchronously upload them to the cloud for storage after the entire video recording is completed for subsequent compliance review by regulatory authorities. In order to ensure the compliance of business-handling videos, financial and insurance institutions generally use manual methods to review videos. A large amount of video data will be gen...

AI Technical Summary

Problems solved by technology

It takes 10 to 15 minutes to manually review a video, and manual review alone cannot meet the growing business needs

In addition, although special auxiliary equipment can be used to detect signature behavior, the operation of auxiliary equipment is complicated and costly, and the universal applicability is relatively poor, which is limited by the usage scenarios

[0003] Existing signature detection systems usually determine the result of signature recognition based on whether the signature pen and hand appear in the video. It is easy to misjudge the behavior of the target picking up the pen but not signing as a signature behavior, which affects the accuracy of the recognition result.

Images