Image texture feature extraction method

Abstract

The invention provides an image texture feature extraction method. The invention provides an image feature extraction, training and detection method. Original image normalization reduces influence of illumination variation, shadows and noisy points, feature extraction adopts a Haar-like method, keeps edge gradient information in an original image, a central pint is compared with points on the upper side, lower side, left side and right side of the central point and four diagonal points, and weight information is added in a feature extraction process, thereby making up for insufficiency of conventional texture features; and a combination mode of features greatly reduces feature dimensionality, compared with a conventional texture features, the final feature extraction method has lower feature dimensions and faster training and detection speed in pedestrian detection, and the detection rate is correspondingly improved.

Description

technical field [0001] The invention relates to the field of image processing, in particular to an image texture feature extraction method. Background technique [0002] Object detection is one of the core problems in computer vision. The object detection method based on statistics mainly uses machine learning to train a classifier from a series of training data, and then uses the classifier to identify the input window. There are two key points in the object detection method, one is feature extraction, and the other is classifier design. The purpose of feature extraction is to reduce the dimensionality of data and obtain features that can reflect the attributes of objects, thereby facilitating classification. A good feature should have the characteristics of strong discrimination ability, simple calculation, strong robustness and simple form. Classifier design belongs to the category of machine learning, and its purpose is to obtain a classifier with low computational co...

AI Technical Summary

Problems solved by technology

In pedestrian detection algorithms based on texture features, most of the literature is dedicated to studying the impact of single texture features on pedestrian detection performance. They continuously optimize the performance of single texture features by changing feature extraction methods, dense descriptions, or pyramid descriptions. , so that its detection performance can be compared with HOG features, or even slightly better than HOG features, but it greatly increases the computational complexity

Images