A horn bow region positioning method based on the combination of fusion features and SVM

The invention discloses a horn bow region positioning method based on the combination of fusion features and SVM. Firstly, the LBP feature and HOG feature of the training sample image are extracted. Alocal binary pattern histogram is constructed by calculating the relationship between the pixel value of the training sample and its neighborhood to extract the LBP feature. The collected training samples are divided into several pixel units, and then the direction histograms of the edges or gradients of each pixel point in the unit are obtained, and all the direction histograms are integrated toform a HOG feature descriptor. Then the LBP feature and HOG feature are fused by feature fusion technology. Then, the SVM classifier is used to train and obtain the training parameters. Finally, thetarget image is scanned by multi-scale sliding window, and the region of horn arch is extracted. The invention has the advantages of high positioning accuracy and easy implementation. 本发明公开了种基于融合特征和SVM结合的羊角弓区域定位方法。方法为：首先提取训练样本图