Face snapshot method and system in video monitoring

Abstract

The invention relates to the field of image processing, and discloses a face snapshot method and system in video monitoring, and the method comprises the steps of obtaining a current frame target image of a video or picture flow; performing human face detection on the current frame of target image, and outputting human face rectangular frame coordinates; performing quality evaluation on the face small image to obtain a quality score of each face in the current frame of target image; performing feature extraction on each face small image to obtain face features of each face in the current frameof target image; respectively matching all the stored face tracks with each face in the current frame of target image; updating all face track information; and performing frame-by-frame snapshot andoptimal face snapshot judgment, and reducing the optimal quality score of the face trajectory by using the attenuation factor alpha. According to the invention, the real-time performance of snapshot and the high quality of snapshot pictures are ensured, the computing resources of the system are saved, the snapshot repetition rate is reduced, the robustness of the whole system is improved, and theaccuracy of subsequent face recognition is improved.

Description

technical field [0001] The invention relates to the field of image processing, in particular to a face capture method and system in video surveillance. Background technique [0002] With the advancement and popularization of smart cities, intelligent monitoring systems are widely used in public scenes with security needs such as traffic security checks, squares, and supermarkets. The system obtains different faces appearing in the current scene through intelligent face capture, extracts face features through the face recognition module, compares them with faces in the library or target face pictures, and judges whether they are the same face. When necessary, the system Immediately send out an alarm response, so as to realize tasks such as real-time investigation of dangerous personnel and fast tracking of suspects. Compared with the traditional manual inspection, the face intelligent capture and recognition system can quickly compare a large number of faces, and at the same...

AI Technical Summary

Problems solved by technology

There are generally two typical capture methods. The first one is to select the frame with the best quality as the optimal face capture after a trajectory ends. This method captures fewer faces and captures higher quality images. , but the system must wait until the end of the trajectory to perform subsequent processing, which cannot respond in real time, and is difficult to apply to scenes that require high real-time performance, and a single snapshot cannot guarantee the stability of the capture quality. Errors will directly affect the accuracy of face recognition; the second capture method is to capture at intervals, this method can guarantee a certain real-time performance when the time interval is short, and will also capture high-quality pictures , but there are many repeated faces captured, which will reduce the processing efficiency of subsequent face recognition and cause a waste of computing resources

However, due to the complex and changeable monitoring scene, the human face continues to move in the picture, and there are often situations such as lighting changes, blurring, occlusion, and facial posture changes, which exacerbate the phenomenon of repeated captures and poor quality.

It is difficult to meet the three requirements of real-time capture, small amount of capture, and high-quality capture at the same time. Therefore, the performance of the intelligent monitoring system in practical applications is limited. How to balance the three aspects of speed, quantity and quality has become an urgent problem to be solved in the face capture method question

Images