Authors: Kangshun Li; Weiguang Chen; Ying Huang; Shuling Yang
Addresses: College of Mathematics and Informatics, South China Agricultural University, Guangzhou, China ' College of Mathematics and Informatics, South China Agricultural University, Guangzhou, China ' Key Laboratory of Jiangxi Province for Numerical Simulation and Emulation Techniques, Institute of Mathematical and Computer Sciences, Gannan Normal University, Ganzhou 341000, Jiangxi Province, China ' College of Software Engineering, South China University of Technology, Guangzhou, China
Abstract: Crop disease and insect pest detection and recognition using machine vision can provide precise diagnosis and preventive suggestions. However, the complexity of agricultural pest and disease identification based on traditional bag of words (BOW) models is high and the effect is general. This paper presents a histogram quadric segmentation algorithm based on an evolutionary algorithm to observe the features (colour, texture) of disease spots and to learn from the guided filtering algorithm. This process aims to obtain the precise positions of disease spots in images. Dense-SIFT, which can extract features and spatial pyramid, which can map image features to high-spatial-resolution space, are simultaneously applied in the recognition of crop diseases and insect pests in the BOW model. The experimental results show that the new segmentation algorithm can effectively locate the positions of disease spots in corn images and the improved BOW model substantially increases the recognition accuracy of crop diseases and insect pests.
Keywords: evolutionary algorithm; disease spot segmentation; image recognition; diseases; insect pests.
International Journal of High Performance Computing and Networking, 2019 Vol.14 No.3, pp.274 - 283
Received: 29 May 2017
Accepted: 01 Nov 2017
Published online: 09 Sep 2019 *