| （英） |
In the World Congress On ITS 2013 Tokyo, it was stated that CV(Connected Vehicle) and AV(Automated Vehicle) will converge in the near future. Then no doubt the cooperation between probe cars and big data, smartphones will raise the movements towards intelligent vehicles as the collective intelligence, which is partly begun as the field experiments conducted by UMTRI and as the other ones by IBM. On the other hand, the R&D of neuro computer conducted by DARPA SyNAPSE project has also shown the sign of the beginning of industrializations. In this article, we firstly propose the multi-viewpoint driving scene recognition which is expected to drastically reduce the amounts of tasks to collect and teach huge size of GT (ground truth) which has been serious bottleneck of conventional machine learning. Also it is expected to improve drastically the performance of environment understanding according to the scene information. Secondly, we investigate the feasibilities of technical realization in both aspects of current level and future level, especially regarding the evolution process of ET (estimated truth) towards GT. |