Flower type identification method based on machine learning

• Main Authors: CHEN BIN, YANG JING, XIAO BENDU, WU MENGLAN, WANG XIANBAO
• Format: Patent
• Language: Chinese; English
• Published: 15.11.2019
• Subjects: CALCULATING; COMPUTING; COUNTING; HANDLING RECORD CARRIERS; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

The invention discloses a flower type identification method based on machine learning. The method comprises the following steps: firstly, carrying out image preprocessing on each image from a sample library and then segmenting an ROI region; establishing a two-dimensional coordinate system according to the position of the ROI; performing four-side positioning on a flower, determining a rectangularregion position of the flower, extracting SIFT features from an ROI region of each sample, and then performing K-means clustering on the SIFT features of each sample to obtain a keyword of the sample, a weight of the keyword and sample category information; then, using the keywords and the weights of all the samples as input, and the sample categories as output. A multi-class SVM (Support VectorMachine) classifier to obtain a classifier containing the class information of all flower classes, and inputting newly acquired unclassified pictures into the trained multi-class SVM classifier to output the corresponding flow