中文说明：HOG特征计算的步骤：（1）将输入的彩图转换为灰度图；（2）采用Gamma校正法对输入图像进行颜色空间的标准化（归一化）；目的是调节图像的对比度，降低图像局部的阴影和光照变化所造成的影响，同时可以抑制噪音的干扰；（3）计算梯度；主要是为了捕获轮廓信息，同时进一步弱化光照的干扰。（4）将梯度投影到单元的梯度方向；目的是为局部图像区域提供一个编码，（5）将所有单元格在块上进行归一化；归一化能够更进一步对光照、阴影和边缘进行压缩，通常，每个单元格由多个不同的块共享，但它的归一化是基于不同块的，所以计算结果也不一样。因此，一个单元格的特征会以不同的结果多次出现在最后的向量中。我们将归一化之后的块描述符就称之为HOG描述符。（6）收集得到检测空间所有块的HOG特征；该步骤就是将检测窗口中所有重叠的块进行HOG特征的收集，并将它们结合成最终的特征向量供分类使用。
English Description：
The steps of hog feature calculation are as follows: (1) convert the input color image into gray image; (2) normalize the color space of the input image by gamma correction method; (3) calculate the gradient; the main purpose is to capture the contour At the same time, the interference of light is further weakened. (4) The gradient is projected to the gradient direction of the cell; the purpose is to provide a coding for the local image region; (5) all cells are normalized on the block; normalization can further compress the illumination, shadow and edge. Generally, each cell is shared by multiple different blocks, but its normalization is based on different blocks, so the calculation results are not the same . Therefore, the features of a cell will appear in the final vector many times with different results. We call the normalized block descriptor hog descriptor. (6) The hog features of all blocks in the detection space are collected; this step is to collect the hog features of all overlapping blocks in the detection window, and combine them into the final feature vector for classification.