Regarding the linear discriminant analysis algorithm LDA algorithm, it can be used for dimensionality reduction and classification, a supervised learning strategy (that is, you have to specify several categories), which is different from PCA (PCA is unsupervised learning).
Recommended learning URL:
https://blog.csdn.net/ruthywei/article/details/83045288
https://blog.csdn.net/itplus/article/details/12038357
The idea of LDA can be summarized in one sentence, that is, "the intra-class variance is the smallest and the inter-class variance is the largest after projection (OTSU algorithm also has a similar concept?) "
The specific study should refer to the two websites given above for in-depth study