A compact omnidirectional stereo camera for autonomous driving

Compact and low-cost devices are needed for autonomous driving to image and measure distances to objects 360-degree around. We have been developing an omnidirectional stereo camera exploiting two hyperbolic mirrors and a single set of a lens and sensor, which makes this camera cost efficient. This paper presents a new calibration method for this camera. Based on the original calibration method by Mei and Rives, we improve the calibration accuracy by considering higherorder radial distortion, detailed tangential distortion, an image sensor tilt, and a lens-mirror offset. Our method is applied to our prototype and reduces the root mean square of the calibration accuracy by 1.2 times and 2.2 times for the upper- and lower-view images, respectively. The distance error is less than 8% up to objects 14 meters apart, which is improved more than seven times compared to the original method, although the error is still larger than the target value of 5%. We consider that the remaining calibration error is due to distortion of the glass cylinder and a degraded optical resolution. As future work, we plan to make further improvements in the calibration and optical resolution of the prototype. In addition, a rectification method for cylindrically expanded images needs to be developed.