A forest fire smoke video target detection method based on eigenroots and fluid mechanics

Abstract

The invention discloses a forest fire smoke video target detection method based on characteristic roots and fluid mechanics. Firstly, the dynamic area in the video is extracted by extracting continuous frame images in the video; secondly, the connected area skeleton diagram is calculated for the dynamic area by a morphological algorithm. ; Then extract the suspicious tobacco root feature candidate points in the end points of the continuous frame skeleton images; next, make a judgment based on the Navier-Stokes equation to the tobacco root feature candidate points, and obtain the predicted tobacco image of the tobacco root feature; finally by predicting the image It is fused with the current actual image to determine the area where smoke exists. Due to the certain predictability of the algorithm, it can continuously affect the judgment information of consecutive frames, making the judgment results stable, and can effectively avoid accidental calculation errors caused by the image processing stage.

Description

technical field [0001] The invention belongs to the fields of forest fire prevention and video target detection, and in particular relates to a video-based forest fire detection method. Background technique [0002] The mainstream methods of smoke detection technology based on video images can be roughly divided into three categories: color intensity-based, dynamic detection-based, and texture-based detection. Although new types of detection methods such as feature fusion, multi-feature extraction, and optical flow have emerged during the development of detection methods, the root cause is still the update and fusion of the three major features of color, dynamics, and texture. This characteristic determines the rigid requirements of the method for image quality, and requires specific threshold debugging and detection for different scenarios. Although the emerging feature classification method based on deep learning network is less dependent on the threshold, the large amoun...

AI Technical Summary

Problems solved by technology

More importantly, it ignores the attention to the natural environment around the smoke generation point, such as wind speed, so it cannot bring predictive result determination

Images