Abstract: A system comprises a LiDAR sensor and a processor coupled to the LiDAR. The processor is configured to derive a regression line, and generate the outline of the static object by connecting the classified valid points except for the classified removal points. Wherein the processor is further configured to calculate an angle defined by one segment and another segment adjacent to the segment, select an i-lth point, which is a start point of a segment including an ith point, as a start point, when an angle between segments at both sides with respect to the ith point is greater than the reference angle, and select a jth point as an end point, when an angle between segments at both sides with respect to the jth point is less than or equal to the reference angle and when there is the start point before the jth point.

Abstract: A free space detection apparatus for a vehicle is provided to detect a free space based on a lidar sensor and recognize a driving environment. The apparatus includes a point data processor that removes noise of point data acquired from the LIDAR sensor and derives a half line with respect to an obstacle point from which noise has been removed based on coordinates of the obstacle point. A map processor generates a map for space classification, checks whether the obstacle point or the half line with respect to the obstacle point is present in a cell of the map and determines cell states of the map. A free space detector detects a free space based on the cell states of the map.

Abstract: An apparatus for recognizing an object based on a lidar sensor and a method thereof can extract a valid contour point of an object from a point cloud generated by the lidar sensor and recognize a shape of the object based on a contour line composed of the extracted valid contour point of the object, such that it is possible to recognize the shape of an object located on a road with high accuracy, thereby improving the driving stability of an autonomous vehicle. The apparatus includes the lidar sensor that generates the point cloud around an autonomous vehicle, and a controller that extracts the valid contour point of the object from the point cloud and recognizes the shape of the object based on the contour line composed of the extracted valid contour point.

Abstract: An object orientation determination method includes generating pixel data on each of a plurality of unit pixels included in a region of interest of a point cloud acquired from an object, generating a plurality of candidate boxes using the generated pixel data, and determining, as a heading angle of an oriented bounding box, an inclination of a candidate box having a smallest cost among costs calculated on the plurality of candidate boxes. A cost of each of the plurality of candidate boxes is calculated based on positions of respective sides of a corresponding one of the plurality of candidate boxes and the pixel data.

Abstract: A method of analyzing the shape of an object using a LiDAR sensor includes determining the first to Mth shapes of first to Mth layers (where M is a positive integer of 2 or more) related to a target object using clustered LiDAR points and analyzing the determined first to Mth shapes according to a predetermined priority to determine the shape of the target object.

Abstract: A device and method are configured to selectively remove specific points included in multiple layers of data received from a sensor. A vehicle may include a LiDAR sensor as the sensor, and a signal processor configured to select a static object included in the multiple layers, to select target layers from among the multiple layers, to define a shape formed by points included in each of the selected target layers, to select a reference layer based on the shape formed by the points, and to remove, based on a distance between a contour formed by the reference layer and each of points of remaining layers and a reference distance, corresponding points from at least one of the remaining layers.

Abstract: A system for extracting an outline of a static object generates a regression line on the basis of continuity of angles between segments interconnecting points obtained from a LiDAR sensor, which indicates a boundary of a static object, removes points where a distance departing from the regression line is greater than or equal to a certain reference distance, and extracts an outline composed of remaining valid points, thus removing points caused by forest or the like except for the static object to be recognized among points clustered and recognized as one object together with the static object, such as a guardrail or a building, and extracting the outline of the static object to be recognized at high reliability.

Abstract: A free space detection apparatus for a vehicle is provided to detect a free space based on a lidar sensor and recognize a driving environment. The apparatus includes a point data processor that removes noise of point data acquired from the LIDAR sensor and derives a half line with respect to an obstacle point from which noise has been removed based on coordinates of the obstacle point. A map processor generates a map for space classification, checks whether the obstacle point or the half line with respect to the obstacle point is present in a cell of the map and determines cell states of the map. A free space detector detects a free space based on the cell states of the map.

Abstract: An apparatus for recognizing an object based on a lidar sensor and a method thereof can extract a valid contour point of an object from a point cloud generated by the lidar sensor and recognize a shape of the object based on a contour line composed of the extracted valid contour point of the object, such that it is possible to recognize the shape of an object located on a road with high accuracy, thereby improving the driving stability of an autonomous vehicle. The apparatus includes the lidar sensor that generates the point cloud around an autonomous vehicle, and a controller that extracts the valid contour point of the object from the point cloud and recognizes the shape of the object based on the contour line composed of the extracted valid contour point.

Abstract: A collision avoidance control system and method are provided. The system includes a GPS receiver that obtains location information of a vehicle, a navigation system having map information, and a sensor unit that senses a target vehicle located near a roundabout. The sensor obtains traveling information of the target vehicle and forward view image information of the vehicle. A controller then calculates an estimated collision point based on the map information, the location information of the vehicle and the traveling information of the target vehicle to determine a risk of collision based on an absolute value of a difference between an arrival time of the vehicle to the estimated collision point and an arrival time of the target vehicle to the estimated collision point. The speed of the vehicle is then adjusted in response to the determined risk of collision.

Abstract: A method of estimating a front obstacle detection distance includes detecting an obstacle from a front image of a vehicle, extracting coordinates of a lower end portion of a detection box of the detected obstacle, calculating a projected line on real world coordinates based on the extracted coordinates of the lower end portion, searching for a point or a line through which the projected line passes on a detailed map, and calculating a distance to the point or the line through which the projected line passes on the detailed map from a position of the vehicle.

Abstract: A method of estimating a front obstacle detection distance includes detecting an obstacle from a front image of a vehicle, extracting coordinates of a lower end portion of a detection box of the detected obstacle, calculating a projected line on real world coordinates based on the extracted coordinates of the lower end portion, searching for a point or a line through which the projected line passes on a detailed map, and calculating a distance to the point or the line through which the projected line passes on the detailed map from a position of the vehicle.

Abstract: A collision avoidance control system and method are provided. The system includes a GPS receiver that obtains location information of a vehicle, a navigation system having map information, and a sensor unit that senses a target vehicle located near a roundabout. The sensor obtains traveling information of the target vehicle and forward view image information of the vehicle. A controller then calculates an estimated collision point based on the map information, the location information of the vehicle and the traveling information of the target vehicle to determine a risk of collision based on an absolute value of a difference between an arrival time of the vehicle to the estimated collision point and an arrival time of the target vehicle to the estimated collision point. The speed of the vehicle is then adjusted in response to the determined risk of collision.

Abstract: A matting method and apparatus for extracting a foreground object are disclosed. The matting method includes extracting one or more pixels from an input image and a trimap; calculating a first processing cost by extracting one or more nonlocal neighboring pixels; calculating a second processing cost by extracting one or more local neighboring pixels; and performing matting on the input image by estimating an opacity value for each of the extracted pixels based on the first and second processing costs. Accordingly, it is possible to perform high-quality image matting, and thus increase a performance in extracting a to foreground object from the input image, so that the matting method and apparatus can be applied to multiple-purpose image editing software.

Abstract: A matting method and apparatus for extracting a foreground object are disclosed. The matting method includes extracting one or more pixels from an input image and a trimap; calculating a first processing cost by extracting one or more nonlocal neighboring pixels; calculating a second processing cost by extracting one or more local neighboring pixels; and performing matting on the input image by estimating an opacity value for each of the extracted pixels based on the first and second processing costs. Accordingly, it is possible to perform high-quality image matting, and thus increase a performance in extracting a to foreground object from the input image, so that the matting method and apparatus can be applied to multiple-purpose image editing software.

Abstract: Provided are an apparatus and method for correcting distortion of an image. The method includes finding center point vectors including lens distortion center information about an image and dividing the image into at least two blocks on the basis of the center point vectors, and determining representative directions of the at least two divided blocks and removing outlier vectors not indicating the representative directions from the blocks to estimate a lens distortion center. Accordingly, accurate image correction is enabled by estimating distortion centers of images generated by devices equipped with any lens. Also, image content improving lens distortion correction performance is differentiated from image content degrading the same by picking out image content improving lens distortion center estimation performance, so that a superior image correction effect to other image correction techniques can be obtained.

Abstract: Provided are an apparatus and method for processing an image. The apparatus includes a detector configured to form one group with pieces of pixel information on an edge of a received image, detect line segments, and generate a plurality of curves using the detected line segments, an optimizer configured to apply a weight to a cost function based on the plurality of generated curves according to distance from the center of the image, define a curve having a shorter length than a threshold value as an outlier and a curve having a greater length than the threshold value as an inlier, remove the outlier, and estimate a distortion level parameter, and a corrector configured to correct a position of a pixel with a correction value using the distortion level parameter. Accordingly, it is possible to correct lens distortion of even an image that has not been directly photographed, and to correct images generated by devices equipped with any lens rather than a specific camera or photographing apparatus.

Abstract: A method for detecting and correcting a chromatic aberration, and an apparatus and method for processing an image using the same are provided. A chromatic aberration is detected by dividing an edge portion of the image into static areas where a color varies below a reference value and a dynamic area where a color varies above the reference value, calculating a intensity difference value between at least one reference color and a comparative color for each of the static areas and the dynamic area, and determining at least one pixel having a intensity difference value exceeding a range of intensity difference values in the static areas among pixels in the dynamic area, as a chromatic aberration region. The chromatic aberration is corrected by correcting the intensity difference value for the pixels in the chromatic aberration region using a correction value in the intensity difference value range in the static area.

Abstract: A method and apparatus for processing an image to create a grayscale image are provided. The method includes converting, in a color image consisting of a plurality of pixels, a pixel value of each pixel in a RGB color space into a pixel value in a new color space through color space conversion, extracting color information using the pixel value in the new color space, calculating a luminance change amount of each pixel using the extracted color information, calculating a luminance value for the grayscale image by combining an original luminance value of each pixel and the luminance change amount, and creating the grayscale image according to the calculated luminance value. In converting color image into grayscale image and representing the grayscale image, the luminance change amount is set according to a property of each color perceived by humans and added to original luminance value, thereby explicitly discriminating among objects.

Abstract: Provided are an apparatus and method for correcting distortion of an image. The method includes finding center point vectors including lens distortion center information about an image and dividing the image into at least two blocks on the basis of the center point vectors, and determining representative directions of the at least two divided blocks and removing outlier vectors not indicating the representative directions from the blocks to estimate a lens distortion center. Accordingly, accurate image correction is enabled by estimating distortion centers of images generated by devices equipped with any lens. Also, image content improving lens distortion correction performance is differentiated from image content degrading the same by picking out image content improving lens distortion center estimation performance, so that a superior image correction effect to other image correction techniques can be obtained.