A Linear Discriminant Learning Method for One-Dimensional Range Image Feature Extraction of True and False Targets

Abstract

The invention belongs to the technical field of radar target recognition, and in particular relates to a linear discriminant learning method for extracting one-dimensional distance image features of true and false targets. The method of the present invention obtains various sample distribution center vectors through iterative learning, and then uses the learning center vector instead of the sample mean vector to calculate the intra-class scatter matrix and the inter-class scatter matrix, and obtains a linear discriminant learning transformation matrix, when the target sample data distribution is non-Gaussian In the case of distribution, it can still well represent the degree of intra-class aggregation and inter-class separation, which overcomes the disadvantage that the conventional discriminant subspace is only suitable for Gaussian distribution of sample data, thereby improving the performance of target recognition.

Description

Technical field [0001] The present invention belongs to the field of radar target recognition technology, specifically relates to a linear discriminant learning true and false target one-dimensional distance image feature extraction method. Background [0002] In radar target recognition, the discriminant vector subspace method can reduce the difference between the features of similar targets while increasing the difference between the features of heterogeneous targets, so as to extract effective identification features, so that the discriminant vector subspace method has obtained good recognition performance. [0003] However, the discriminant vector subspace method is only suitable for the case where the sample data is Gaussian distribution, and in practice, the sample data distribution may be a non-Gaussian distribution, resulting in a decrease in the recognition performance of the discriminant vector subspace method. The recognition performance of existing conventional discri...

AI Technical Summary

Problems solved by technology

[0003] However, the discriminant vector subspace method is only suitable for the case where the sample data is a Gaussian distribution, and the actual sample data distribution may be non-Gaussian distribution, resulting in a decline in the recognition performance of the discriminant vector subspace method

There is room for further improvement in the recognition performance of existing conventional discriminant vector subspace methods

Images