Skip to main content

For prevention of traffic accidents, image processing technique has been widely studied to recognize pedestrians from the image taken by a camera; however, the accuracy of recognition is degraded by bad weather condition. In contrast, millimeter radar is tolerant to weather condition though with lower angular resolution; JRC has made test software to recognize the target in the radar signal. Based on machine learning, the recognition software classifies each target moving in open space into three types of targets: pedestrian, bicycle, and car. The recognition succeeded in classifying targets with the accuracy of approximately 90%. Through the experiments, JRC confirmed that the effective information for target classification is included in radar data.

Yoichi Tomiki
Yukinobu Tokieda
Takuma Nishiyama
Hiroki Sugawara