An image recognition method based on hierarchical feature extraction

• Main Authors: LI WEIFU, FAN HONGHUA, FU YINGXIONG, SHI WEI
• Format: Patent
• Language: Chinese; English
• Published: 14.12.2018
• Subjects: CALCULATING; COMPUTING; COUNTING; HANDLING RECORD CARRIERS; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

The invention relates to an image recognition method based on hierarchical feature extraction, comprising the following steps: 1, defining a multi-scale frame and transformation of adjacent scales according to the features of an image data set; 2, constructing templates under different scales according to the defined frame and the transformation; 3, constructing features under different scales according to that templates under different scales; 4, fusing the multi-scale features to construct a classifier for image recognition. The technical proposal provided by the invention can effectively obtain both representative and discriminative template sets to extract more effective image features, and can obtain higher recognition accuracy with lower computational complexity. 本发明涉及种基于分层特征提取的图像识别方法，包括：步骤1，根据图像数据集的特征定义多尺度的框架和相邻尺度的变换；步骤2，根据定义的所述框架和所述变换构造不同尺度下的模板；步骤3，根据不同尺度下的所述模板构造不同尺度下的特征；步骤4，融合多尺度的所述特征，构造分类器进行图像识别。本发明提供的技术方案可有效地获得既有代表性、又有判别性的模板集合，来提取更有效的图像特征，能够以较低的计算复杂度来获取更高的识别精度。