Video recognition and tracking system and method based on depth sensor

Abstract

The invention discloses a depth sensor-based video recognition and tracking system and method thereof, which are characterized by comprising: a pan / tilt camera control module, a tracking module, an identification module, and a recording module; and applied to n+1 depth sensors, two In a classroom environment composed of a PTZ camera, a master and n slaves; place a depth sensor around the speaker area, and place two PTZ cameras above the speaker area and the audience area , where one PTZ camera faces the direction of the speaker area, and the other PTZ camera faces the direction of the audience area, and the remaining n depth sensors are evenly arranged on both sides of the audience area and correspond to the n slave machines respectively. The present invention can automatically identify and track the movement of the speaker, and at the same time, monitor the audience area, automatically identify and switch the audience interacting with the speaker, thereby realizing fully automatic classroom and conference recording.

Description

Technical field [0001] The invention relates to the field of video recording, in particular to a video recognition and tracking system based on a depth sensor and a method thereof. Background technique [0002] At present, most of the methods for taking pictures of courses and meetings at home and abroad are to directly invite photographers to take pictures. Or install a fixed camera in the classroom to shoot. Entrusting photographers to record in classrooms and meetings not only consumes human and material resources, but also diverts the audience's attention and affects the effect. That is, the current video recording has the problem of insufficient automation. [0003] Traditional single-camera tracking methods such as optical flow method, time difference method, or Gaussian background modeling method for people tracking have poor anti-noise performance, easy to confuse the foreground and background, easy to track wrong targets, and difficult to apply to the full frame Real-ti...

AI Technical Summary

Problems solved by technology

Entrusting photographers to record in classrooms and meetings not only consumes manpower and material resources, but also diverts the attention of the audience and affects the effect

That is, the current video recording has a problem of insufficient automation

[0003] Traditional single-camera tracking methods such as optical flow method, time difference method, or Gaussian background modeling method for tracking people have poor anti-noise performance, easy to confuse foreground and background, easy to track wrong targets, and difficult to apply to full frames Real-time processing of video streams, etc.

And the use of multi-eye cameras for three-dimensional space data acquisition requires complex algorithms to synthesize and calculate the data, and the real-time performance is not high

Images