图像识别的特征提取，一般可以分为两个层次：一层是底层的特征提取，另一层

The feature extraction of image recognition can generally be divided into two levels: one layer is the feature extraction of the bottom layer, and the other layer is the high-level feature extraction. The low-level feature extraction is the basis of image analysis. Commonly used color features, shape features and texture features have the characteristics of simple calculation and stable performance. High-level feature extraction is generally based on the height of the semantic level, such as face recognition, people Behavior analysis, etc., all these need to be extracted based on the bottom layer, and can be obtained through machine learning.

图像识别的特征提取，一般可以分为两个层次：一层是底层的特征提取，另一层是高层 次的特征提取。底层的特征提取是图像分析的基础，常用的有颜色特征、形状特征和纹理特征，具有计算简单和性能稳定的特点，高层的特征提取，一般是基于语义层次的高度，如人脸识别、人的行为分析等，这些都需根据底层的提取结果，并通过机器学习才能得到。

Generally, the feature extraction of image recognition can be divided into two levels: one is the bottom level feature extraction, the other is the high level feature extraction. The feature extraction in the bottom layer is the basis of image analysis. The common features are color feature, shape feature and texture feature, which are simple in calculation and stable in performance. The feature extraction in the top layer is generally based on the height of semantic level, such as face recognition, human behavior analysis, etc., which can only be obtained by machine learning based on the extraction results in the bottom layer.<br>