Coal dust particle image recognition method based on improved Fast R-CNN

Abstract

The invention discloses a coal dust particle image recognition method based on an improved Fast R-CNN. The coal dust particle image recognition method comprises the steps of 1, inputting a coal dust particle image into a trained improved Fast R-CNN; wherein a plurality of coal dust particle calibration areas of a training sample are stored in the improved Fast R-CNN network; 2, performing featureextraction on the improved Fast R-CNN network by adopting a convolutional layer of a VGG network to obtain a feature map of the coal dust particle image; 3, the Fast R-CNN network is improved to recognize the coal dust particles in the feature map of the coal dust particle image from the background; and 4, the Fast R-CNN network is improved to input the coal dust particle image of the coal dust particle target identified in the step 3 into two parallel full connection layers, and the position of the coal dust particle target is finely adjusted through a linear ridge regression device. The detection process is efficient, the detection precision is high, and the contour information of the particle sample can be restored to the maximum extent.

Description

technical field [0001] The invention belongs to the technical field of coal dust particle image recognition, and in particular relates to a coal dust particle image recognition method based on improved Fast R-CNN. Background technique [0002] With the frequent occurrence of coal dust explosions, people pay more and more attention to the causes of safety accidents. Coal dust will explode when the concentration of coal dust particles is large and the ambient temperature reaches a certain level. A coal dust explosion will cause the loss of several or even hundreds of miners' precious lives. China's total coal production accounts for about 37% of the world's total, coal dust explosions account for 50% of coal mine accidents, and the death toll accounts for about 70% of the world's total. In particular, the participation of coal dust with a particle size of 70-200 μm will aggravate the destructive power of the explosion. The image of coal dust particles can be identified by th...

AI Technical Summary

Problems solved by technology

[0003] In the research on particle image recognition, there are artificial extraction methods to outline the image outline, for example, Cheng Zhongqi, Shi Jianghao and Mao Yijun published the paper "Deep Learning Based The method proposed in "Realization of Image Retrieval Technology", however, this method is rough in detection time and precision

GUO Guankai, LIU Wei, YU Lingling (Liu Guankai, Liu Wei, Yu Lingling) and others published the paper "Imagesegmentation of touching particles" in the journal "China powder science and technology" (China powder science and technology) based on improved FAST and watershed algorithm "(particle image segmentation based on improved FAST and watershed algorithm), proposes a combination of feature point detection and watershed algorithm to extract the segmentation points of particle images, but this will affect the accuracy of image extraction and detection of multi-featured small areas lower

SUN, Guodong, LIN Kai, GAO Yuan (Sun Guodong, Lin Kai, Gao Yuan) and others published the paper "Research and Implementation of Ore Particles Image Segmentation Based on Improved AffinityGraph (Research and Implementation of Ore Particle Image Segmentation Based on Improved Affinity Graph), proposed a method of linear expression in the neighborhood to improve pixel correlation, and multi-scale feature extraction for ore particles, but limited the depth of extracted features

In the prior art, there is still a lack of a coal dust particle image recognition method that has high recognition accuracy of coal dust particles and can effectively identify coal dust particle images with unclear edges

Images