Horse image recognition method based on class center SVM and near-end SVM

The invention relates to a horse image recognition method based on a class center SVM and a near-end SVM, and the method comprises the following steps: obtaining horse image data, carrying out the preprocessing, inputting the data into a pre-trained CCPSVM model, and outputting a recognition result, the model comprising a class center SVM classifier and a near-end SVM classifier; the model training process comprises the steps that a horse image data set is acquired and preprocessed; on the basis of the preprocessed horse image data set, the class center of each class is calculated, a linear kernel is adopted to train a class center SVM classifier so as to obtain most relevant horse image data samples in each class, and labels of the corresponding class are distributed to each class center; and taking the most relevant horse image data sample in each category as a training sample of a near-end SVM classifier, and selecting a polynomial kernel to train the near-end SVM classifier to complete training. Compared