A One-Dimensional Range Image Multi-classifier Fusion Recognition Method Based on Class Confidence

Abstract

The invention discloses a one-dimensional distance image multi-classifier fusion recognition method based on category confidence. First, the Euclidean distance measure is used to extract the K nearest neighbor samples from the preprocessed one-dimensional range image signal sample amplitude features; then, the reliability evaluation of the neighborhood sample category is obtained from the classification results of each sub-classifier, and then each classification is obtained. Finally, the confidence of each category of the sample is obtained from the evaluation matrix to achieve the purpose of classification. The present invention is based on decision fusion theory, combined with the idea of ​​K-nearest neighbor, and uses the nearest neighbor and Bayesian criterion to obtain the category confidence of the target sample to complete the target category division. Compared with decision fusion schemes such as single classifier and traditional voting method, this method has higher recognition accuracy and better robustness, and it is a one-dimensional range image decision fusion recognition method with practical value.

Description

technical field [0001] The invention relates to the technology of using multi-classifier fusion to identify one-dimensional distance images, in particular to a one-dimensional distance image multi-classifier fusion recognition method based on category confidence. Background technique [0002] Automatic target recognition is a key area in radar research applications. High-resolution range profile (HRRP) is an important research direction to achieve real-time classification of targets because it contains rich target features and has a small amount of data. Classifier design is an important research direction in HRRP recognition. Designing a reasonable classifier can effectively improve the recognition accuracy and robustness. However, the performance improvement of a single classifier has always been limited. With the idea of ​​fusion, in recent years, thinking about how to synthesize the recognition results of multiple classifiers for the same target to obtain more accurate ...

AI Technical Summary

Problems solved by technology

Traditional fusion methods include majority voting method, weighted average method, and Bayesian method. The relationship between classifiers and their contribution to the classification results, and it only fuses the results of a single test sample under each classifier, resulting in no significant improvement in the fusion result compared with a single classifier, and the Poor stickiness, difficult to apply well in engineering practice

Images