Radar->Camera 坐标转换。 标定结果check [ADS]

FROM baidu Apollo

简单来说： 就是确保融合的精度，radar只能提供位置信息--置信度较高， XY,Z(option), 在鸟瞰图BEV 下，

比较radar 和Lidar的位置信息是否一致（对齐）如果offset 过大，说明转换矩阵不能用，需要重新标定获得。

RADAR-->LIDAR , CAMRA-->LIDAR, 这样就可以知道摄像头和RADAR 之间的转换矩阵。

radar-->camra ? ， 目前L2 应用量产的项目中，摄像头的（单目）的安装位置已知， radar已知。 就可以直接转换了。--不需要外部标定。

Radar-to-Camera Calibration

• Background Information: To verify the extrinsic output, use the LiDAR in the system as a medium. This approach enables you to obtain:

  • The extrinsic parameter of the radar relative to the LiDAR through the extrinsic value of the radar relative to the camera

  • The extrinsic value of the camera relative to the LiDAR

    You can then draw a bird's-eye-view fusion image, which fuses the radar data and the LiDAR data in the LiDAR coordinate system. You can use the alignment of the radar data and the LiDAR data in the bird's-eye-view fusion image to judge the accuracy of the extrinsic parameter. In the fusion image, all of the small white points indicate the LiDAR point cloud, while the large green solid circles indicate radar objects.

• Validation Method:

  The alignment of the radar object and the LiDAR data in the bird's-eye-view fusion image shows the accuracy of the extrinsic parameter. If most of the targets coincide, it is satisfactory. However, if over 40% targets (especially vehicles) do not align, it is not satisfactory and you need to re-calibrate.

• Examples: As shown in the following examples, Figure 6 meets the precision requirements of the extrinsic parameter, and Figure 7 does not.

Figure 6. Good Camera-to-Radar Calibration Validation Result

Figure 7. Bad Camera-to-Radar Calibration Validation Result