Multi-modal compound detection target recognition method based on improved one-class SVM algorithm

Abstract

The present invention provides a multi-mode compound detection target recognition method based on the improved One-Class SVM algorithm, aiming at the output signals of the pre-processed passive millimeter wave detectors, long infrared detectors and FMCW radars in the steady-state scanning state, The features with higher discrimination are selected as the classification features, and the improved One-Class SVM algorithm is used to train the training data to construct a classifier with a higher target recognition rate. The features of the signal to be tested are extracted, and the processed feature data is sent to the classifier for target recognition, and the judgment result is obtained. Considering the rapid development of the current high-speed real-time signal processing system and the characteristics of large amount of information in composite detection data processing, the present invention uses the sample and feature weighted One-Class SVM algorithm from the perspective of feature layer fusion to significantly improve the performance of multi-mode composite detection. The target recognition rate of the device.

Description

technical field [0001] The invention belongs to the technical field of target recognition, in particular to a multi-mode compound detection target recognition method and system based on an improved One-Class SVM algorithm weighted by samples and features. Background technique [0002] The fundamental principle of composite detection technology is to use the respective advantages of various sensors such as radar, radiometer, and infrared detector to achieve more accurate identification of targets by fusing and judging sensor information according to certain rules. In the field of data fusion, the feature layer fusion is located between the data layer and the decision-making layer in terms of real-time and information volume, and with the rapid development of high-speed real-time processing technology, it has become a trend of data fusion processing. [0003] The problem of composite detector target recognition is a single-classification problem. The single-classification supp...

AI Technical Summary

Problems solved by technology

[0003] The problem of composite detector target recognition is a single-classification problem. The single-classification support vector machine introduced by Wang Hongbo and others in the study of the learning method of the single-classification support vector machine can effectively solve the single-classification problem. The values ​​are all of the same importance, but in reality, the importance of samples and eigenvalues ​​is somewhat biased, which is not conducive to the identification of targets

Images