图像预处理包括图像平滑、图像分割、二值化、图像形态学处理和轮廓提取。轮廓提取是使用八邻域边界跟踪算法。在手势的特征提取部分，使用边界傅立叶描绘子的手势特征提取算法。针对手势的识别，采用了基于 BP 神经网络的识别方法-Image pre-processing, including image smoothing, image segmentation, binarization, morphological image processing and contour extraction.

常用的opencv图像特征提取算法。包括直方图、边缘检测、角点检测、直线提取、圆形提取等。并包括不太成熟的BP神经网络识别纯底色数字。-Commonly used opencv image feature extraction algorithm. Including histograms, edge detection, corner detection, line extraction, circular extraction. And includes less mature pure b

道路场景识别，通过对样本图像处理和特征提取，再通过bp神经网络进行学习，最后通过学习后得到的权值进行样本识别。-Road scene recognition, through the sample image processing and feature extraction, and then through bp neural network learning, and finally by learning the weights obtained after the sample ide

cnn卷积神经网络进行在图像方面的应用，可以进行训练，有训练结果。-Cnn convolution neural network in the image of the application, can be trained, there are training results