Flame identification algorithm of low-rank analysis

Abstract

The invention discloses a flame identification algorithm of low-rank analysis. First, video images are acquired by an infrared camera and preprocessed by Gaussian smoothing filter, and grayscale images are binarized. The external rectangle of a suspicious area of the current frame is extracted and zoomed in. The data of the external rectangle is used as a column vector. The data of external rectangles of a flame area of the continuous n-1 frame behind the current frame is extracted and zoomed to the same size, and the data of the frame is taken as column vectors frame-by-frame. All column vectors form a matrix D which his decomposed into the sum of a low rank matrix A and a sparse error matrix E. The singular value decomposition for the low rank moment A is obtained, and respective contribution rate is calculated by the square of the singular values. The number of singular values is determined by that the cumulative contribution rate is greater than the given threshold. The determined number is new rank rank(D) of the matrix D, and the boundary point theta* is obtained through experiments. An alarm is given off when the number is greater than or equal to theta*. A lot of experiments show that the algorithm has high recognition rate and robustness, and the method is proved to be effective.

Description

technical field [0001] The invention relates to the technical field of flame identification, in particular to a low-rank analysis flame identification algorithm. Background technique [0002] With the continuous enhancement of computer processing capabilities and the increasing requirements of the information society for multimedia information processing, the flame detection technology based on image processing has been developed rapidly. And other characteristics, with significant advantages and broad prospects. [0003] In recent years, many scholars have conducted research on flame recognition algorithms based on image processing. Some existing technologies are based on support vector machines and linear classifiers for flame recognition algorithms; some existing technologies identify flames based on characteristics such as flame temperature, boundary irregularities, contour changes, and area changes, and other existing technologies It is proposed to use wavelet feature...

AI Technical Summary

Problems solved by technology

Some existing technologies are based on support vector machines and linear classifiers for flame recognition algorithms; some existing technologies identify flames based on characteristics such as flame temperature, boundary irregularities, contour changes, and area changes, and other existing technologies It is proposed to use wavelet features and color features, and use neural network to identify flames. This algorithm has high requirements for flame materials, and it is difficult to achieve a relatively complete library of flame materials.

Images