Button recognition with texture feature based on spiking neural network. Issue 16 (25th September 2018)

• Record Type: Journal Article
• Authors: Zhang, Zhenmin; Wu, Qingxiang; Lai, Xiaoyan; Lin, Xiufang
• Is Part Of: Journal of engineering. Volume 2018:Issue 16(2018)
• Page Start: 1692
• Page End: 1697
• Publication Date: 2018-09-25
• Subjects: matrix algebra -- image texture -- learning (artificial intelligence) -- image recognition -- image classification -- image segmentation -- feature extraction -- image colour analysis -- neural nets; button recognition -- spiking neural network -- neuron network -- neural networks -- nonlinear information processing -- supervised feature extraction -- unsupervised feature extraction -- image classification -- texture segmentation -- image recognition -- co‐occurrence matrix algorithm -- visual images -- texture feature extraction -- image structure characterisation analysis -- buttons -- analogous textures -- button image feature extraction -- button images; Engineering -- Periodicals; Engineering; Electronic journals; Periodicals; 620.005
• DOI: 10.1049/joe.2018.8283
• Languages: English
• ISSNs: 2051-3305

Abstract

Abstract : Spiking neuron network is generally considered as the third generation of neural networks. This type of network is a class of machine learning techniques that exploit many layers of non‐linear information processing for supervised or unsupervised feature extraction, image classification, texture segmentation, and image recognition. On the other hand, the grey‐level co‐occurrence matrix algorithm is widely used in visual images for texture feature extraction and image structure characterisation analysis. For those buttons with the same size, same shape, similar colours, and analogous textures, they cannot be effectively identified by conventional methods. At this time, the spiking neural network trained with the improved GLCM algorithm can be used to achieve button image feature extraction, classification, and recognition. Experiments show that the method proposed here can effectively segment the button images with their texture features.