Reporter 26th from China Aerospace Science and Engineering Second chamber two learned that the department recently successfully developed "based on depth learning intelligent driving system", with a few centimeters of embedded chip size, can realize real-time and accurate intelligent perception of the environment, in the target recognition accuracy rate to reach the world advanced level. Public data show that the accuracy of intelligent driving target recognition is up to 90.55%, but it takes 4 seconds to process a picture, and the algorithm accuracy of two intelligent driving systems is 90.05%, but it only takes 0.03 seconds to process a picture. At present, two teams have broken through a series of core technologies, including multi-target detection and identification, removable area segmentation, lane detection, etc., and have formed technical barriers in depth neural network compression, depth neural network compilation tool chain and intelligent acceleration engine. Guo Rui, the team leader, said that the scene semantic segmentation is to let the computer understand it "see" the image represents what, and the depth of neural network can be learned to extract high-level semantic features, to eliminate the backlight, Shangguang, shadow, defects and other factors to target detection interference. Guo Rui said that the focus of the team has shifted to the "intelligent driving system based on deep learning" engineering and product work, and with the relevant automotive plants to carry out functional testing and production trial, is expected to complete small-scale production by the end of this year. In addition, the team is developing a fusion of visible light, infrared and millimeter-wave radar of the new smart sensors to meet the driving environment of military vehicles in the light of complex and changeable, night marching prohibited lighting and other special scenarios, the military vehicles to satisfy the full weather, all-weather auxiliary driving needs. The team's advanced technologies, such as intelligent perception and intelligent decision-making, are gradually playing a role in the target detection and identification of weapon system and intelligent accusation.