Abstract: Improved processing of sensor data (e.g., LiDAR data) can be used to distinguish between free space and objects/hazards. Autonomous vehicles can use such information for performing autonomous driving and/or parking operations. LiDAR data can include a plurality of range measurements (e.g., forming a 3D point cloud). Each of the range measurements can correspond to a respective LiDAR channel and azimuth angle. The processing of LiDAR data can include identifying one or more of the plurality of range measurements as ground points or non-ground points based on one or more point criteria. The one or more point criteria can include a ground projection criterion and an angular variation criterion.

Abstract: In some embodiments, a computer-readable medium stores executable code, which, when executed by a processor, causes the processor to: capture an image of a finder pattern using a camera; locate a predetermined point within the finder pattern; use the predetermined point to identify multiple boundary points on a perimeter associated with the finder pattern; identify fitted boundary lines based on the multiple boundary points; and locate feature points using the fitted boundary lines.

Abstract: In some embodiments, a computer-readable medium stores executable code, which, when executed by a processor, causes the processor to: capture an image of a finder pattern using a camera; locate a predetermined point within the finder pattern; use the predetermined point to identify multiple boundary points on a perimeter associated with the finder pattern; identify fitted boundary lines based on the multiple boundary points; and locate feature points using the fitted boundary lines.

Abstract: The present invention is related to systems and methods for detecting an occluded object based on the shadow of the occluded object. In some examples, a vehicle of the present invention can capture one or more images while operating in an autonomous driving mode, and detecting shadow items within the captured image. In response to detecting a shadow item moving towards the direction of vehicle travel, the vehicle can reduce its speed to avoid a collision, should an occluded object enter the road. The shadow can be detected using image segmentation or a classifier trained using convolutional neural networks or another suitable algorithm, for example.

Abstract: Disclosed examples include three-dimensional imaging systems and methods to reconstruct a three-dimensional scene from first and second image data sets obtained from a single camera at first and second times, including computing feature point correspondences between the image data sets, computing an essential matrix that characterizes relative positions of the camera at the first and second times, computing pairs of first and second projective transforms that individually correspond to regions of interest that exclude an epipole of the captured scene, as well as computing first and second rectified image data sets in which the feature point correspondences are aligned on a spatial axis by respectively applying the corresponding first and second projective transforms to corresponding portions of the first and second image data sets, and computing disparity values of a stereo disparity map according to the rectified image data sets to construct.

Abstract: Improved processing of sensor data (e.g., LiDAR data) can be used to distinguish between free space and objects/hazards. Autonomous vehicles can use such information for performing autonomous driving and/or parking operations. LiDAR data can include a plurality of range measurements (e.g., forming a 3D point cloud). Each of the range measurements can correspond to a respective LiDAR channel and azimuth angle. The processing of LiDAR data can include identifying one or more of the plurality of range measurements as ground points or non-ground points based on one or more point criteria. The one or more point criteria can include a ground projection criterion and an angular variation criterion.

Abstract: A system for use in a vehicle, the system comprising one or more sensors, one or more processors operatively coupled to the one or more sensors, and a memory including instructions, which when executed by the one or more processors, cause the one or more processors to perform a method. The method comprising capturing a current point cloud with the one or more sensors, determining an estimated location and an estimated heading of the vehicle, selecting one or more point clouds based on the estimated location and heading of the vehicle, simplifying the current point cloud and the one or more point clouds, correlating the current point cloud to the one or more point clouds, and determining an updated estimate of the location of the vehicle based on correlation between the current point cloud and the one or more point clouds.

Abstract: Improved processing of sensor data (e.g., LiDAR data) can be used to distinguish between free space and objects/hazards. Autonomous vehicles can use such information for performing autonomous driving and/or parking operations. LiDAR data can include a plurality of range measurements (e.g., forming a 3D point cloud). Each of the range measurements can correspond to a respective LiDAR channel and azimuth angle. The processing of LiDAR data can include identifying one or more paths as candidate ground paths based on one or more path criteria. The processing of LiDAR data can also include identifying one or more of the plurality of range measurements as ground points or non-ground points based on the one or more paths identified as candidate ground paths and based on one or more point criteria.

Abstract: In some embodiments, a computer-readable medium stores executable code, which, when executed by a processor, causes the processor to: capture an image of a finder pattern using a camera; locate a predetermined point within the finder pattern; use the predetermined point to identify multiple boundary points on a perimeter associated with the finder pattern; identify fitted boundary lines based on the multiple boundary points; and locate feature points using the fitted boundary lines.

Abstract: A method of generating a high dynamic range (HDR) image is provided that includes capturing a long exposure image and a short exposure image of a scene, computing a merging weight for each pixel location of the long exposure image based on a pixel value of the pixel location and a saturation threshold, and computing a pixel value for each pixel location of the HDR image as a weighted sum of corresponding pixel values in the long exposure image and the short exposure image, wherein a weight applied to a pixel value of the pixel location of the short exposure image and a weight applied to a pixel value of the pixel location in the pixel long exposure image are determined based on the merging weight computed for the pixel location and responsive to motion in a scene of the long exposure image and the short exposure image.

Abstract: A vehicle control system for moving a vehicle to a target location is disclosed. According to examples of the disclosure, a camera captures one or more images of a known object corresponding to the target location. An on-board computer having stored thereon information about the known object can process the one or more images to determine vehicle location with respect to the known object. The system can use the vehicle's determined location and a feedback controller to move the vehicle to the target location.

Abstract: A method of generating a high dynamic range (HDR) image is provided that includes capturing a long exposure image and a short exposure image of a scene, computing a merging weight for each pixel location of the long exposure image based on a pixel value of the pixel location and a saturation threshold, and computing a pixel value for each pixel location of the HDR image as a weighted sum of corresponding pixel values in the long exposure image and the short exposure image, wherein a weight applied to a pixel value of the pixel location of the short exposure image and a weight applied to a pixel value of the pixel location in the pixel long exposure image are determined based on the merging weight computed for the pixel location and responsive to motion in a scene of the long exposure image and the short exposure image.

Abstract: A method for autonomously controlling a vehicle is disclosed. In some examples, a vehicle can maneuver out of a parking space in an autonomous and unmanned operation. While parking, the vehicle can capture first one or more images of its surroundings and store the images in a memory included in the vehicle. Upon starting up, the vehicle can capture second one or more images of its surroundings and compare them to the first one or more images to determine if there is an object, person, or animal proximate to the vehicle, for example. In some examples, in accordance with a determination that there is no object, person, or vehicle present that was not present during parking, the vehicle can autonomously move from the parking space with or without a user present in the vehicle.

Abstract: A system that performs a method is disclosed. The system receives information about a map, which includes information about one or more zones in the map. While navigating a vehicle along a driving path within the map, the system receives information about the location of the vehicle in the map. The system estimates an error bounds of the location of the vehicle, and determines in which of the one or more zones in the map the error bounds is located within. In response to the determination: in accordance with a determination that the error bounds is located within a first zone in the map, the system causes the vehicle to perform a driving operation. In accordance with a determination that the error bounds is located within a second zone of the one or more zones in the map, the system causes the vehicle to perform a different driving operation.

Abstract: Camera-based autonomous parking is disclosed. An autonomous parking procedure can include detecting parking lines in images captured by a camera on a vehicle. The vehicle can be localized with respect to the parking lines based on location data for the vehicle from a GPS receiver and a location determination for the vehicle based on detected ends of the parking lines. The vehicle can further determine an occupancy state of one or more parking spaces formed by the two or more parking lines using a range sensor on the vehicle and select an empty space. A region of interest including the selected space can be identified and one or more parking lines of the selected space can be detected in an image of the region of interest. The vehicle can autonomously move to reduce errors between the location of the vehicle and the final parking position within the selected space.

Abstract: A system that performs a method is disclosed. The system detects one or more characteristics about a vehicle via a first set of one or more sensors and determines one or more characteristics about the vehicle's surroundings via a second set of one or more sensors. The system also detects a vehicle failure via the first set of one or more sensors. In response to detecting the vehicle failure via the first one or more sensors, the system selects one or more road hazard indicators using the one or more characteristics about the vehicle's surroundings. After selecting the one or more road hazard indicators using the one or more characteristics about the vehicle's surroundings, the system deploys the one or more road hazard indicators.

Abstract: Camera-based autonomous parking is disclosed. An autonomous parking procedure can include detecting parking lines in images captured by a camera on a vehicle. The vehicle can be localized with respect to the parking lines based on location data for the vehicle from a GPS receiver and a location determination for the vehicle based on detected ends of the parking lines. The vehicle can further determine an occupancy state of one or more parking spaces formed by the two or more parking lines using a range sensor on the vehicle and select an empty space. A region of interest including the selected space can be identified and one or more parking lines of the selected space can be detected in an image of the region of interest. The vehicle can autonomously move to reduce errors between the location of the vehicle and the final parking position within the selected space.

Abstract: A method, apparatus and a system multi-camera image processing method. The method includes performing geometric alignment to produce a geometric output by estimating fish eye distortion correction parameters, performing initial perspective correction on related frame, running corner detection in the overlapping areas, locating the stronger corner, calculating BRIEF descriptors for features and match feature point from two cameras using BRIEF scores, performing checks and rejecting wrong feature matches, finding perspective matrices to minimize distance between matched features; and creating a geometric lookup table.

Abstract: A method of transforming an N-bit raw wide dynamic range (WDR) Bayer image to a K-bit raw red-green-blue (RGB) image wherein N>K is provided that includes converting the N-bit raw WDR Bayer image to an N-bit raw RGB image, computing a luminance image from the N-bit raw RGB image, computing a pixel gain value for each luminance pixel in the luminance image to generate a gain map, applying a hierarchical noise filter to the gain map to generate a filtered gain map, applying the filtered gain map to the N-bit raw RGB image to generated a gain mapped N-bit RGB image, and downshifting the gain mapped N-bit RGB image by (N?K) to generate the K-bit RGB image.

Abstract: A method of automatically focusing a projector in a projection system is provided that includes projecting, by the projector, a binary pattern on a projection surface, capturing an image of the projected binary pattern by a camera synchronized with the projector, computing a depth map from the captured image, and adjusting focus of the projector based on the computed depth map.