A Sensor Fusion of the Millimeter-Wave Radar and Camera-Based Visual Processing for Detection of Lanes in the Automated Driving

摘要

The recent trend of the automated driving such as Google Car Project, a high spec laser range finder like LiDARs is necessary to complete the Simultaneous Localization and Mapping (SLAM), which is crucial for the accurate determination of running direction and orientation in accordance with the digital map information linking with GPS. However such a high spec laser range finder is costly, which sometimes higher than the usual price of the car. An expecting technology instead of the excessive dependency on the high spec sensor device is the complementary approach to combine different types of sensors, known as the sensor fusion; however the appropriate combination is still unclear even researchers made a long efforts to investigate. In the present study, we focused on the validation method to clarify the appropriate combination by proposing a system of 76 GHz millimeter-wave radar and image processing according to the two stage model with the initial filter of contour extractions by 2D wavelet analysis and the discrete outline information judgement. The two stage model was capable to detect from lanes to cracks on the road, but the threshold control is sensitive to control, which affords to determine it by information from the millimeter-wave radar. Further validation will be done in the real automated driving.