Pedestrian detection method based on CoLBP co-occurrence features and GSS (gradient self-similarity) features

The invention discloses a pedestrian detection method based on CoLBP co-occurrence features and GSS (gradient self-similarity) features. The method includes the following steps that: the HOG features and LBP features of each frame of image in a video sequence are extracted, paired gradient self-similarity (GSS) features of the local gradient blocks of the images are calculated according to the HOG features, and CoLBP co-occurrence features are obtained according to the LBP features; asymmetric GSS features are removed through adopting a feature generating machine (FGM), so that discriminant gradient self-similarity (DGSS) features are obtained; an HOG feature and CoLBP co-occurrence feature training-based linear SVM (support vector machine) removes negative samples in the images; and as for remaining negative samples and positive samples, pedestrians in each frame of image are detected through using an HOG feature and CoLBP co-occurrence feature training-based Real-AdaBoost classifier, and detection results are obtained. According to the method of the invention, the CoLBP co-occurrence features and the GSS features at higher levels are obtained based on the HOG features and LBP features, and therefore, the reliability of pedestrian feature extraction is increased, and a final detection result also shows that the method of the invention has good detection effect.