A Dimension Decomposition-Based Density Clustering Classification Pattern Recognition Method

Abstract

The invention discloses a density clustering class label pattern recognition method based on dimension decomposition. According to the clustering core point index set, the core point matrix is ​​taken out from the training data, and the kth test data in the unmanned aerial vehicle test data matrix is ​​taken out to obtain Take the set of adjacent core points of the kth test data, analyze the set of adjacent core points of the kth test data, identify the clustering labels of the kth test data, and traverse each test data in the UAV test data matrix . The present invention only needs to input the radius of the neighborhood, and gets rid of the trouble of adjusting the hyperparameters of the algorithm. No modeling is required and the algorithm overhead is small.

Description

technical field [0001] The method belongs to the field of pattern recognition, and in particular relates to a pattern recognition method of density clustering and classification based on dimension decomposition. Background technique [0002] The density clustering algorithm DBSCAN has many advantages such as being able to process data of any shape, independently inferring the number of clusters according to the laws of the data itself, and automatically removing noise data, so it is widely used in many fields. [0003] After the DBSCAN algorithm completes the clustering analysis on the original clustering data (training data), it will divide the training data into several clusters and mark them with different class labels, that is, group the data. In practical applications, it is often necessary to judge which group of the training data the new data (test data) belongs to, that is, clustering and labeling pattern recognition of the test data. [0004] Common pattern recogni...

AI Technical Summary

Problems solved by technology

The neural network-based pattern recognition method needs to be modeled, and the model is prone to fall into local optimum and overfitting. When the amount of data increases, the calculation cost is large

The pattern recognition method based on machine learning also needs to learn the regular modeling of training data and class labels, and then identify the test data class labels, but this type of algorithm has the following problems: 1) When the amount of data increases, the calculation cost is relatively high. Large; 2) prone to overfitting or underfitting; 3) troubled by algorithm hyperparameter tuning

Images