Apparatus and method of generating top-view image
AssigneeHyundai Mobis Co., Ltd.
InventorSeong Soo Lee
Life of this patent
Estimating the term…
Abstract
Provided is a method of generating a top-view image. The method includes: obtaining an input image by photographing a peripheral road surface of the vehicle by using a camera when a vehicle is driving, detecting at least two lanes from the input image, detecting a vanishing point at which the at least two lanes intersect, calculating a first rotation angle of the camera by using the vanishing point, calculating a second rotation angle of the camera for converting the input image into a top-view image represented in a top-view image coordinate system, based on the first rotation angle, and rotation-converting the input image by using the second rotation angle to generate the top-view image.
Claims9
A method of generating a top-view image, comprising:
- obtaining an input image by photographing a peripheral road surface of a vehicle by using a camera when the vehicle is driving;
- detecting at least two lanes from the input image;
- detecting a vanishing point at which the at least two lanes intersect;
- calculating a first rotation angle of the camera by using the vanishing point;
- calculating a second rotation angle of the camera for converting the input image into a top-view image represented in a top-view image coordinate system, based on the first rotation angle;
- rotation-converting the input image by using the second rotation angle to generate the top-view image;
- and
- verifying the top-view image for alignment between a vertical axis of the vehicle and the top-view image coordinate system after the generating of the top-view image;
- wherein the verifying of the top-view image comprises:
- detecting a lane from a previous top-view image generated from a previous frame and detecting a lane from a current top-view image generated from a current frame;
- determining whether a direction of the lane detected from the previous top-view image matches a direction of the lane detected from the current top-view image;
- determining whether an interval between the lane detected from the previous top-view image and the vertical axis of the vehicle matches an interval between the lane detected from the current top-view image and the vertical axis of the vehicle;
- and
- determining the alignment as being made between the vertical axis of the vehicle and the top-view image coordinate system when the direction of the lane detected from the previous top-view image matches the direction of the lane detected from the current top-view image, and the interval between the lane detected from the previous top-view image and the vertical axis of the vehicle matches the interval between the lane detected from the current top-view image and the vertical axis of the vehicle.
An apparatus for generating a top-view image, the apparatus comprising:
- a camera configured to obtain input images by photographing a peripheral road surface of a vehicle when the vehicle is driving;
- a lane detector configured to detect at least two lanes from each of the input images;
- a vanishing point detector configured to detect a vanishing point at which the at least two lanes intersect;
- a rotation angle calculator configured to calculate a rotation angle for converting each of the input images into an image corresponding to a top-view viewpoint by using the vanishing point;
- an image rotation unit configured to rotation-converting the input images by using the rotation angle;
- an image synthesizer configured to synthesize the input images rotation-converted at the rotation angle to generate a top-view image;
- and
- a top-view image verifier configured to verify the top-view image for alignment between a vertical axis of the vehicle and the top-view image coordinate system;
- wherein the top-view image verifier is configured to verify whether the alignment between the vertical axis of the vehicle and the top-view image coordinate system is made, based on a previous top-view image of a previous frame and a current top-view image of a current frame.
Dependencies are inferred from the claim text.