Detecting approaching cars via artificial insect vision

摘要

The application of computer vision or image processing technology is typically limited to expensive machines in big facilities. One reason for this phenomenon is the cost of hardware required by this technology. In order to develop computer vision algorithms runnable on inexpensive hardware, we look at the vision system of insects, and discuss four ideas we could take into account when developing such algorithms. On the basis of this discussion, we specify a design for an approaching-car detection system for car driver assistance. This system is attached to the rear of a car, and faces backwards. If other cars approach from the rear, the system informs the driver. We investigated the performance of the system using various experiments and demonstrated its effectiveness. The amount of computation the system must carry out is approximately one ten-thousandth of that required by a conventional image processing algorithm. It is executable on a small and inexpensive processor for embedding purposes.