Video flame identification method based on key frame and fast support vector machine fusion

Abstract

The invention discloses a video flame identification method based on fusion of a key frame and a fast support vector machine. The method comprises the following steps: collecting a video and extracting image features of a video frame as a classification basis; using a PCA algorithm to obtain main features of the video images, clustering the video frames by using a k-means clustering algorithm, classifying the data objects according to the similarity degree between the calculation frames, and selecting the video frames corresponding to the class centers of different classes as the key frames ofthe classes; obtaining the sum of m key frames of the video, fusing the features of the m key frames by adopting a weighting mode, and extracting the static features and the dynamic features of the video; and taking the static features and the dynamic features as input vectors of a fast support vector machine, and performing classification identification by the fast support vector machine to obtain a final video flame identification result. According to the method, the image processing technology is used for carrying out whole-course real-time automatic monitoring and image analysis on the fire process, the monitoring accuracy and reliability are improved, and the safety of fire operation is ensured.

Description

technical field [0001] The invention relates to the technical field of machine vision, in particular to a video flame recognition method based on fusion of key frames and fast support vector machines. Background technique [0002] Hot work is a very important part of equipment maintenance, and it is of great significance to ensure continuous production, technological repair and other on-site operations. Hot work includes operations such as electric welding, gas welding, and argon arc welding, which are usually accompanied by a large number of sparks flying around during the implementation process. The working environment of iron and steel enterprises is relatively complicated. Inflammable materials such as wire and corner cloth are often left in the corners of the site, and lubricating oil / grease can be seen everywhere on the ground. There are many potential safety hazards during hot work. In response to such problems, at present, major iron and steel enterprises are active...

AI Technical Summary

Problems solved by technology

The manual monitoring method has high labor risk, high work intensity, and poor reliability; the monitoring range of the sensor is small, and the monitored data information is easy to lose, which is not suitable for large-space workshops

Images