Image calibration method and image calibration device

摘要

To alleviate the strictness degree required for a relative positional relationship between a calibration index and a vehicle in an image calibration method and an image calibration device. The image calibration device includes a calibration index 20, four cameras 10, a distortion correction processing unit 30, a viewpoint transformation processing unit 40, and a calibration processing unit 50. The calibration index 20 includes two mutually parallel lines and a distance between which is known and one line orthogonal to the two lines. The cameras 10 take peripheral area images S1 to S4 of peripheral areas R1 to R4 including the calibration index 20. The distortion correction processing unit 30 corrects the peripheral area images S1 to S4 by using internal parameters M to obtain distortion-corrected images P1 to P4. The viewpoint transformation processing unit 40 performs viewpoint transformation processing on the distortion-corrected images P1 to P4 by using external parameters N to obtain a viewpoint-transformed composite image Q0. The calibration processing unit 50 performs calibration in the viewpoint transformation processing. The calibration processing unit 50 includes an internal parameter correction unit 51 and an external parameter correction unit 52. The internal parameter correction unit 51 corrects the internal parameters M. The external parameter correction unit 52 corrects the external parameters N.

主权项

1. An image calibration method including:
taking peripheral area images of peripheral areas of a vehicle respectively with a plurality of cameras installed on the vehicle, the peripheral areas including a calibration index provided in advance on a road surface on which the vehicle is arranged, performing distortion correction on the peripheral area images respectively taken by the cameras on the basis of internal parameters according to optical characteristics of the cameras, performing viewpoint transformation processing on a plurality of distortion-corrected images obtained by the distortion correction on the basis of external parameters according to states of attachment of the cameras to the vehicle, the viewpoint transformation processing being performed so that the plurality of distortion-corrected images are processed into a single viewpoint-transformed composite image obtained when the vehicle is looked down from above the vehicle, and calibrating the viewpoint-transformed composite image by adjusting the external parameters on the basis of an image of the calibration index in the viewpoint-transformed composite image, in the viewpoint transformation processing, the image calibration method characterized in that the calibration index includes at least two lines which are mutually parallel and a distance between which is known and at least one line orthogonal to each of the two lines, and is formed so that each of the peripheral area images includes at least images corresponding to the two mutually parallel lines and an image corresponding to the one line orthogonal to each of the two lines and so that among the lines included in the calibration index, at least one line located in a shooting range of one of the cameras and at least one line located in a shooting range of a different one of the cameras are located on a straight line, for each of the plurality of distortion-corrected images, or viewpoint-transformed composite image, correction of adjusting the internal parameters is performed so that images corresponding to the two mutually parallel lines and an image corresponding to the one orthogonal line included in the distortion-corrected image, or the viewpoint-transformed composite image extend linearly, a new viewpoint-transformed composite image is obtained by using new distortion-corrected images obtained by using the corrected internal parameters, and the correction of adjusting the external parameters is performed so that the new viewpoint-transformed composite image thus obtained satisfies the following conditions (1) to (4):(1) the images corresponding to the two mutually parallel lines are mutually parallel;(2) a distance between the images corresponding to the two mutually parallel lines is the known distance between the lines in the calibration index;(3) the image corresponding to the one orthogonal line is orthogonal to the images corresponding to the two parallel lines; and(4) images corresponding to the line located on the straight line in the calibration index are located on a straight line.