In the field of image processing, recognition is one of the major challenges for researchers in recent years. The goal of recognition is to detect and extract features in images to classify samples into different classes. One problem in this field is self-driving robot cars using image processing. Our aim in this paper is to present a deep learning method to solve the lane recognition problem. We will build a self-driving robot car basing on convolutional neural network with the use of many different layers to automatically extract the best features in the image. We have done the work and have achieved high performance results.