Automatic labeling method based on multi-layer dictionary learning

• Main Authors: REN WENCHAO, ZHAO HAIYING
• Format: Patent
• Language: Chinese; English
• Published: 06.01.2023
• Subjects: CALCULATING; COMPUTING; COUNTING; PHYSICS

The invention discloses an automatic labeling method based on multilayer dictionary learning, and the method comprises the following steps: S1, extracting image features, and expressing an image as a feature vector; s2, performing two-layer dictionary learning and information coding on the image data; s3, carrying out pooling and full connection on the coded information; and S4, classifying the final coding information vector by a multiclass SVM (Support Vector Machine), and finally obtaining the annotation information. According to the method, firstly, feature extraction is carried out on an image, then two-layer dictionary learning and information coding are carried out respectively, after coding information is obtained, information enhancement processing is carried out through pooling and a full connection layer, finally, the information is input into a multiclass SVM model for judgment, finally, effective label information is obtained, and automatic labeling of the image is achieved. 本发明公开一种基于多层字典学习的自动标注方