In the third wave of artificial intelligence, the concept of AI is now well known and is accelerating penetration in many fields. However, currently we are still in the weak artificial intelligence stage. To realize strong artificial intelligence, upgrading object recognition to scene understanding is the key. To realize scene understanding, we must first understand the relationship between objects, such as the most basic three-dimensional spatial relationship. As one of the important application directions of artificial intelligence, in recent years, the long-range depth camera lidar of self-driving cars has been recognized as a mainstream solution for cars to understand three-dimensional space. However, lidar is currently large and expensive, which makes the further popularization of this technology full of challenge. So, how can we solve these problems so that the laser pointer radar can further promote the integration and development between cars and artificial intelligence?
From “weak” to “strong” artificial intelligence, the most basic thing is to understand the three-dimensional spatial relationship. The use of a depth camera-based vision system for visual navigation, identifying the external environment, planning paths, and avoiding obstacles has become an automatic task. Recognized solutions in artificial intelligence fields such as driving, AR / VR, and robots. Of course, the depth camera is also divided into a close-range depth camera and a long-range depth camera. The close-range depth camera has high accuracy and short working distance, and can be used for industrial detection, disorderly sorting, face recognition, and the like. Long-range depth cameras have low recognition accuracy and long working distances, and can be used for gesture, motion and attitude recognition, unmanned obstacle avoidance, and path planning. Among them, the laser pointer radar commonly used in automatic driving systems belongs to long-range depth cameras.