Algorithm Design of Self-driving Vehicle Based on Visual Sensing Technology

In order to realize the vision of a safe, efficient, green and civilized smart car power, solve the traffic safety and congestion problems, and fully meet the people’s growing needs for a better life, the task of building intelligent transportation is increasingly urgent. In recent years, with the introduction of deep learning theory, 3D target detection technology based on visual sensing technology has developed rapidly in the field of autonomous driving. In recent years, self-driving cars, a new generation of cars equipped with advanced sensors and other devices and using new technologies such as artificial intelligence, have become the focus of attention of all parties. At present, unmanned vehicles mainly use lidar to acquire 3D point cloud of the scene in the driving area, which makes 3D target detection technology based on deep neural network attract more and more attention. Because the auto-driving system needs to control the car in real time, it requires that the visual perception algorithm can run in real time, and the on-board computing platform of the auto-driving car belongs to the mobile computing platform, and its computing power is very limited. Under the existing technology, relying on a single sensor can not obtain accurate detection results in a complex environment. Although many kinds of 3D target detection networks have been proposed in the literature, the existing detection networks still face the problem of insufficient detection accuracy of small target objects. Based on the theory of vision sensing technology and the key algorithms of automatic driving visual perception system, this paper develops a research on the algorithm of automatic driving vehicle based on vision sensing technology.