Multi-branch path recognition algorithm based on linear discriminant analysis

• Main Authors: NIU YUNFENG, WU XIANFU
• Format: Patent
• Language: Chinese; English
• Published: 22.12.2023
• Subjects: CALCULATING; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS

The invention relates to the technical field of AGVs, and discloses a multi-branch path recognition algorithm based on linear discriminant analysis, which comprises the following steps: firstly, projecting an image obtained by an imaging device of an AGV, i.e., a feature vector sample, from a high-dimensional space to a low-dimensional space through linear or nonlinear mapping by adopting an LDA dimension reduction method; finding out a meaningful low-dimensional structure hidden in the high-dimensional observation data; mapping the original sample to a high-dimensional feature space by using a kernel technique through an SVM (Support Vector Machine) method, and then searching a hyperplane with a maximum interval between two types of samples in the space; a plurality of dichotomy SVM classifiers are constructed and combined together, when SVM sub-classifiers of a category i and a category j are constructed, sample data belonging to the category i and the category j are selected from a sample data set as train