From BAIDU Apollo

https://github.com/ApolloAuto/apollo/blob/master/docs/quickstart/apollo_2_0_sensor_calibration_guide.md

Camera-to-Camera Calibration

• Background Information: In the warp image, the green channel is produced from the wide-angle camera image, and the red and blue channels are produced from the telephoto camera image. Users can compare the alignment result of the warp image to validate the precision of the calibrated extrinsic parameter.

• Validation Method:

  In the fusion area of the warp image, judge the alignment of the scene 50 meters away from the vehicle. If the images coincide completely, the extrinsic parameter is satisfactory. However, if a pink or green ghost (displacement) appears, the extrinsic parameter is in error. 

• 组合的图片--两个摄像头融合后的图片，RGB ,绿色由广角（近焦距）镜头提供， G/B-颜色由长焦镜头提供。如果拼接有问题的，说明标定没做好

•  

• When the error is greater than a certain range (for example, 20 pixels, determined by the actual usage), you need to re-calibrate the extrinsic parameter. Under general circumstances, due to the parallax, some dislocations may occur in the horizontal with close objects, but the vertical direction is not affected. This is a normal phenomenon.

• **Examples: **As shown in the following examples, Figure 2 meets the precision requirements of the extrinsic parameter, and Figure 3 does not.

Figure 2. Good Calibration Result for Camera-to-Camera Calibration

 

Figure 3. Bad Calibration Result for Camera-to-Camera Calibration