Video analysis method combining target detection with human body posture estimation

Abstract

The invention relates to a video analysis method combining target detection with human body posture estimation, and the video analysis method comprises the steps: obtaining an onsite camera video stream through a streaming media server, intercepting a video frame from the onsite camera video stream, and obtaining a decoded picture; carrying out rapid target detection on each obtained picture by adopting a deep learning algorithm, and finding out target objects of all identifiable categories; judging whether a person exists or not, and sequentially carrying out dangerous behavior analysis and dangerous scene analysis on the picture; in dangerous behavior analysis and dangerous scene analysis, if a dangerous event is found, reporting and pushing an alarm to a display platform in time; and continuously analyzing the video stream in real time. According to the video analysis method, target detection and a human body posture estimation algorithm are combined and introduced into the field ofintelligent monitoring and recognition of infrastructure videos, and dangerous behaviors or scenes can be recognized, and the labor force of manual screen staring in traditional monitoring can be liberated, and the infrastructure safety management efficiency is greatly improved.

Description

technical field

[0001] The invention relates to the technical field of video intelligent recognition, in particular to a video analysis method combining target detection and human body pose estimation. Background technique

[0002] In power grid infrastructure projects, traditional video surveillance is based on manual screen monitoring, and random calls to monitor lines for monitoring. In the era of big data information, the number of monitoring lines is increasing rapidly, but due to the internal personnel positions and staffing of the enterprise, it is impossible for the management personnel of power grid infrastructure projects to increase significantly. Moreover, some danger sources, such as fire sources, potholes, etc., require continuous attention 24 hours a day, covering the entire construction site, which is impossible to complete manually. Therefore, there is an urgent need for automated video analysis techniques capable of dangerous behaviors and scenarios. [0...

Examples

Embodiment Construction

[0032] The technical solution of the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments.

[0033] Such as figure 1 and figure 2 As shown, a video analysis method that combines target detection with human pose estimation includes the following steps:

[0034] S1. Fetch video frames: Obtain live camera video streams through the streaming media server, intercept video frames from them, and obtain decoded pictures.

[0035] S2. Target detection: Use the deep learning algorithm to perform fast target detection on the first frame of the picture, and find all the target objects that can be identified.

[0036] Among them, in the fast target detection, the YOLO target detection model with real-time performance is used to make its detection speed reach 50 frames per second, which meets the real-time detection requirements.

[0037] S3. Positioning of key points of human body posture: If a person is found in the ...

technical field

[0001] The invention relates to the technical field of video intelligent recognition, in particular to a video analysis method combining target detection and human body pose estimation. Background technique

[0002] In power grid infrastructure projects, traditional video surveillance is based on manual screen monitoring, and random calls to monitor lines for monitoring. In the era of big data information, the number of monitoring lines is increasing rapidly, but due to the internal personnel positions and staffing of the enterprise, it is impossible for the management personnel of power grid infrastructure projects to increase significantly. Moreover, some danger sources, such as fire sources, potholes, etc., require continuous attention 24 hours a day, covering the entire construction site, which is impossible to complete manually. Therefore, there is an urgent need for automated video analysis techniques capable of dangerous behaviors and scenarios. [0...

Examples

Embodiment Construction

[0032] The technical solution of the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments.

[0033] Such as figure 1 and figure 2 As shown, a video analysis method that combines target detection with human pose estimation includes the following steps:

[0034] S1. Fetch video frames: Obtain live camera video streams through the streaming media server, intercept video frames from them, and obtain decoded pictures.

[0035] S2. Target detection: Use the deep learning algorithm to perform fast target detection on the first frame of the picture, and find all the target objects that can be identified.

[0036] Among them, in the fast target detection, the YOLO target detection model with real-time performance is used to make its detection speed reach 50 frames per second, which meets the real-time detection requirements.

[0037] S3. Positioning of key points of human body posture: If a person is found in the ...

Images