Fire image recognition method based on mixed feature and manifold learning

Abstract

The invention discloses a fire image recognition method based on a mixed feature and manifold learning. The method comprises steps of: firstly dividing an image into a plurality of nonoverlapping sub-blocks by using a block dividing strategy, and detecting whether the sub-blocks are candidate fire areas by establishing two color modules in the HSV color space and a similarity matching method basedon a color histogram; secondly, in view of the significant visual characteristics of a flame area and a smoke area, using a hybrid feature extraction method in combination with local characteristicsand texture features to capture more image details and improve the accuracy of subsequent classifications; and finally, using a manifold learning method to construct a flame manifold and a smoke manifold based on double manifold topology, and establishing a classifier for final determination of the fire image on the two types of image manifolds. The method not only reduces the burden to the classifier caused by high-dimension feature, but also obtains high fire image recognition accuracy.

Description

technical field [0001] The invention relates to an image recognition method, in particular to a fire image recognition method based on hybrid feature and manifold learning. Background technique [0002] With the rapid development of our country's economy and society, a large number of fire incidents occur every day. Fires not only bring casualties to human beings, but also cause huge property losses. Therefore, how to effectively carry out early identification and early warning of fire events has important research value and practical significance. Most of the traditional fire identification methods rely on chemical or gas sensors, but these sensors usually need to be installed in a fixed position in a limited space, which cannot be applied to open or open areas, such as: forest fire identification, straw burning identification, high-rise residential fire identification, etc. in application. Compared with the sensor-based fire recognition method, the image / video-based fire...

AI Technical Summary

Problems solved by technology

In addition, although the method of multi-feature fusion can effectively improve the recognition performance, it may generate a large amount of redundant information, and the burden of subsequent classifiers will be increased due to the increase of feature dimension.

Images