Traditional image processing has always relied on the features of manual design, but such features only describe and represent the low-level edge information in the image, and it is difficult to extract the in-depth information. Deep learning in data-driven way, using a series of nonlinear transformation, and extracted from the original data and various convolution neural network technology as the unmanned and focus of research in the fields of mobile robot, and will have broad prospect of application and research of visual value algorithm, the traditional algorithms of computer vision has been replaced by deep neural network algorithm. The computer vision algorithms are divided into traditional algorithms and neural network algorithms. According to the different classification of computer vision algorithms, the latest research achievements and applications of all kinds of algorithms in the field of image recognition are systematically described.