Real-time face abstract service method and system for intelligent video conference terminal

Abstract

The invention discloses a real-time face abstract service method and system for an intelligent video conference terminal, and belongs to the technical field of face recognition. The method comprises the steps of initializing a model; acquiring a video frame; performing face detection on the preprocessed frame image by using a face detection model; initializing a tracker by utilizing a face detection result, carrying out face tracking on a face, and tracking and capturing position information of the face; obtaining coordinates of key points of the face by utilizing a face alignment model, and then transforming the face into a standard face image by adopting similarity transformation; performing face feature mapping by using a face recognition model, and recognizing a face in the frame image; and inputting the recognized face image into a face gallery, and updating the face gallery through face optimization. The method provided by the invention is good in quality, high in definition, high in calculation speed and good in real-time performance when being used for carrying out face abstract service, can be directly applied to a video conference terminal without a large-scale calculation platform, and reduces the cost.

Description

technical field [0001] The invention belongs to the technical field of face recognition, and in particular relates to a method and system for real-time face summary service of an intelligent video conferencing terminal. Background technique [0002] With the development of economy and technology, the office mode is constantly changing, and the status of video conferencing has gradually become prominent in various fields. In video conferencing, a very important technology is the face summary service, and the faces of the people who will participate in the meeting are displayed in the face gallery. For video conferencing, the face summary is to collect face images according to the video screen, and then enter them into the face gallery. The function of the face gallery is to store the face features of the participants and filter them into high-quality face blocks for display. , therefore, the face summarization service technology faces many technical problems, such as how to ...

AI Technical Summary

Problems solved by technology

[0004] However, using the method in the prior art, although the face summary and video summary can be generated, there are many problems in the method in the prior art, the first is the clarity of the generated face summary, the existing method directly uses human The face recognition method generates a face summary, but due to complex and changeable scenes such as lighting changes, motion blur, and face proportions in the video scene, the recognition rate of the face is seriously affected, resulting in a low degree of clarity of the face summary, which is difficult to meet Scenario requirements

Secondly, in the current face abstract generation methods, most of the neural network models are used to complete face recognition and face detection, while the neural network models used in the prior art are often relatively large, resulting in a very large amount of calculation, which is difficult Deployed directly on the terminal device, in order to quickly detect and recognize faces in the video, it is often necessary to use a large machine platform, such as deploying on a cloud server, but using this method to perform face detection, recognition and other operations It is completed on the cloud server and then returned to the terminal device, which will cause delays and poor real-time performance, which will affect the quality of the face summary service, so it is difficult to meet the needs of scenarios such as video conferencing

[0005] To sum up, the existing face summarization service methods, the quality of face summarization service is poor, the definition is low, and the amount of calculation is large, which makes it difficult to quickly and accurately generate face summaries in real time, and the algorithm model cannot be directly Deployed on terminal devices, large-scale computing equipment is required, with poor real-time performance and high cost

Images