Egomotion estimation system and method
An egomotion estimation system may include: a stereo camera suitable for acquiring a stereo image; a map generation unit suitable for generating a depth map and a disparity map using the stereo image; a feature point extraction unit suitable for extracting a feature point from a moving object in the stereo image using the disparity map; a motion vector detection unit suitable for detecting a motion vector of the extracted feature vector point; an error removing unit suitable for removing an error of the detected motion vector; and an egomotion determination unit suitable for calculating and determine an egomotion using the error-removed motion vector.
1. An egomotion estimation system based on a computer for estimating an egomotion of a moving system mounted a stereo camera comprising:
the stereo camera configured to acquire a stereo image; and
the computer configured to estimate the egomotion of the moving system,
wherein the computer generates depth information indicating a distance between objects based on a disparity in the stereo image, the depth information being visualized as a depth map and a disparity maps,
calculates a pixel value accumulated in the disparity map, when the pixel value is equal to or less than a reference value, determine an corresponding object as a road surface, remove the corresponding object from the stereo image, and extract a feature point from a moving object in the stereo image using the disparity map,
detects a motion vector of the extracted feature point by applying an optical flow to the extracted feature point,
removes an error of the detected motion vector using a forward-backward error removing algorithm,
calculates egomotion parameters using the error-removed motion vector,
calculates motion vector for the feature point through the egomotion parameters,
calculates an error in the egomotion parameters for the feature point by summing absolute value of difference between the motion vector detected through the optical flow and the motion vector calculated through the egomotion parameters, and
determines the egomotion by repeating the steps when the error is within a predetermined range.
2. The egomotion estimation system of claim 1 , wherein the computer calculates and determines the egomotion by applying random sample consensus (RANSAC) to the error-removed motion vector.
3. The egomotion estimation system of claim 1 , wherein the stereo camera comprises a first and second cameras mounted on the moving system.
4. The egomotion estimation system of claim 3 , wherein the first and second cameras are 2D cameras capturing first and second images for the same object and wherein the stereo camera generates a stereo, image from said first and second images of the same object.
5. The egomotion estimation system of claim 3 , wherein the first and second cameras are 3D cameras capturing first and second images for the same object and wherein the stereo camera generates a stereo image from said first and second images of the same object.
6. An egomotion estimation method comprising:
acquiring a stereo image from a stereo camera mounted on a moving system;
generating depth information indicating a distance between objects based on a disparity in the stereo image, the depth information being visualized as a depth map and a disparity map;
calculating a pixel value accumulated in the disparity map;
when the pixel value is equal to or less than a reference value, determining an corresponding object as a road surface, and removing the corresponding object from the stereo image;
extracting a feature point from a moving object in the stereo image using the disparity map;
detecting a motion vector of the extracted feature point by applying an optical flow to the extracted feature point;
removing an error of the detected motion vector using a forward-backward error removing algorithm; and
calculating and determining an egomotion of the moving system using the error-removed motion vector,
wherein the calculating and determining the egomotion includes:
calculating egomotion parameters using, the error-removed motion vector;
calculating motion vector for the feature point through the egomotion parameters;
calculating an error in the egomotion parameters for the feature point by summing absolute value of difference between the motion vector detected through the optical flow and the motion vector calculated through the egomotion parameters; and
determining the egomotion by repeating the steps when the error is thin a predetermined range.
7. The egomotion estimation method of claim 6 , wherein the calculating and determining of the egomotion is performed by applying RANSAC to the error-removed motion vector.
8. The egomotion estimation method of claim 6 , wherein the stereo camera comprises a first and second cameras mounted on the moving system.
9. The egomotion estimation method of claim 8 , wherein the first and second cameras are 2D cameras capturing first and second images for the same object and wherein the stereo camera generates a stereo image from said first and second images of the same object.
10. The egomotion estimation system of claim 8 , wherein the first and second cameras are 3D cameras capturing first and second images for the same object and wherein the stereo camera generates a stereo image from said first and second images of the same object.