Object detection method based on feature redundancy elimination AdaBoost classifier

Abstract

The invention discloses an object detection method based on a feature redundancy elimination AdaBoost classifier. The method comprises the following steps of: (1) classifier training, namely, necessarily training a classifier by using a rich and complete feature set and a great number of training samples before the AdaBoost classifier is used for the first time, and improving a conventional training process by using weight-based feature redundancy elimination; (2) object feature extraction, namely, extracting features such as haar features of an object required to be detected by using a feature extraction algorithm; (3) feature input, namely, inputting the features obtained in the step (2) into the trained classifier; and (4) the obtaining of detection results, namely, performing binary classification on the input features by using the classifier, determining whether the input features belong to the object to be detected or not, and outputting the detection results.

Description

technical field [0001] The present invention relates to an object detection method based on a redundant feature reduction AdaBoost classifier (Adaboost basedweighted features redundancy elimination algorithm referred to as AdaBoostWrea), which relates to an object detection method, in particular to a weight-based redundant feature reduction AdaBoost classifier object detection method. Background technique [0002] Object detection, a binary classification problem in pattern recognition, is also a challenging computer vision problem and is sometimes classified as a research topic in the field of machine learning. Object detection can be widely used in many fields, such as pedestrian detection in intelligent monitoring, face detection, similar image retrieval in video, and so on. Although there have been a large number of literatures that have conducted a comprehensive and in-depth exploration of it, several key issues have not been fully explained, and some have not been res...

AI Technical Summary

Problems solved by technology

Therefore, according to the traditional AdaBoost feature selection process, we found that all irrelevant and redundant features will run through the main process of AdaBoost feature selection, so that the existence of irrelevant and redundant features will consume a lot of computing time and significantly reduce the learning efficiency. speed, and when a feature is selected, another irrelevant or redundant feature may also be selected

In this way, due to the existence of irrelevant features and redundant features, the learning algorithm may cause the classifier to overfit the training samples (that is, the output obtained by the training samples is basically the same as the expected output, but the output of the test sample is different from the expected output of the test sample. but big)

Images