Abstract: Autonomous vehicles can include systems and apparatus for performing signal processing on point cloud data from Light Detection and Ranging (LiDAR) devices located on the autonomous vehicles. A method includes obtaining, by a computer located in an autonomous vehicle, a combined point cloud data that describes a plurality of areas of an environment in which the autonomous vehicle is operating; determining that a first set of points from the combined point cloud data are located within fields of view of cameras located on the autonomous vehicle; assigning one or more labels to a second set of points from the first set of points in response to determining that the second set of points are located within bounding box(es) around object(s) in images obtained from the cameras; and causing the autonomous vehicle to operate based on characteristic(s) of the object(s) determined from the second set of points.

Abstract: Vehicles can include systems and apparatus for performing signal processing on sensor data from radar(s) and LiDAR(s) located on the vehicles. A method includes obtaining and filtering radar point cloud data of an area in an environment in which a vehicle is operating on a road to obtain filtered radar point cloud data; obtaining a light detection and ranging point cloud data of at least some of the area, where the light detection and ranging point cloud data include information about a bounding box that surrounds an object on the road; determining a set of radar point cloud data that are associated with the bounding box that surrounds the object; and causing the vehicle to operate based on one or more characteristics of the object determined from the set of radar point cloud data.

Abstract: A system method for occluding contour detection using a fully convolutional neural network is disclosed. A particular embodiment includes: receiving an input image; producing a feature map from the input image by semantic segmentation; learning an array of upscaling filters to upscale the feature map into a final dense feature map of a desired size; applying the array of upscaling filters to the feature map to produce contour information of objects and object instances detected in the input image; and applying the contour information onto the input image.

Abstract: A system and method for lateral vehicle detection is disclosed. A particular embodiment can be configured to: receive lateral image data from at least one laterally-facing camera associated with an autonomous vehicle; warp the lateral image data based on a line parallel to a side of the autonomous vehicle; perform object extraction on the warped lateral image data to identify extracted objects in the warped lateral image data; and apply bounding boxes around the extracted objects.

Abstract: An autonomous vehicle includes audio sensors configured to detect audio in an environment around the autonomous vehicle and to generate audio signals based on the detected audio. A processor in the autonomous vehicle receives the audio signals and compares a time domain or frequency domain representation of the audio signals to a corresponding representation of a known emergency vehicle siren. The comparison causes the processor to output a first determination indicating whether the audio signals are indicative of an emergency vehicle siren. The processor also applies a trained neural network to the audio signals that causes the processor to output a second determination indicating whether the audio signals are indicative of the emergency vehicle siren. If the first determination or the second determination indicates presence of an emergency vehicle siren in the environment around the autonomous vehicle, the autonomous vehicle is caused to perform an action.

Abstract: A system and method for vehicle wheel detection is disclosed. A particular embodiment can be configured to: receive training image data from a training image data collection system; obtain ground truth data corresponding to the training image data; perform a training phase to train one or more classifiers for processing images of the training image data to detect vehicle wheel objects in the images of the training image data; receive operational image data from an image data collection system associated with an autonomous vehicle; and perform an operational phase including applying the trained one or more classifiers to extract vehicle wheel objects from the operational image data and produce vehicle wheel object data.

Abstract: A system and method for occluding contour detection using a fully convolutional neural network is disclosed. A particular embodiment includes: receiving an input image; producing a feature map from the input image by semantic segmentation; learning an array of upscaling filters to upscale the feature map into a final dense feature map of a desired size; applying the array of upscaling filters to the feature map to produce contour information of objects and object instances detected in the input image; and applying the contour information onto the input image.

Abstract: A system and method for vehicle wheel detection is disclosed. A particular embodiment can be configured to: receive training image data from a training image data collection system; obtain ground truth data corresponding to the training image data; perform a training phase to train one or more classifiers for processing images of the training image data to detect vehicle wheel objects in the images of the training image data; receive operational image data from an image data collection system associated with an autonomous vehicle; and perform an operational phase including applying the trained one or more classifiers to extract vehicle wheel objects from the operational image data and produce vehicle wheel object data.

Abstract: A system and method for taillight signal recognition using a convolutional neural network is disclosed. An example embodiment includes: receiving a plurality of image frames from one or more image-generating devices of an autonomous vehicle; using a single-frame taillight illumination status annotation dataset and a single-frame taillight mask dataset to recognize a taillight illumination status of a proximate vehicle identified in an image frame of the plurality of image frames, the single-frame taillight illumination status annotation dataset including one or more taillight illumination status conditions of a right or left vehicle taillight signal, the single-frame taillight mask dataset including annotations to isolate a taillight region of a vehicle; and using a multi-frame taillight illumination status dataset to recognize a taillight illumination status of the proximate vehicle in multiple image frames of the plurality of image frames, the multiple image frames being in temporal succession.

Abstract: A system and method for online real-time multi-object tracking is disclosed. A particular embodiment can be configured to: receive image frame data from at least one camera associated with an autonomous vehicle; generate similarity data corresponding to a similarity between object data in a previous image frame compared with object detection results from a current image frame; use the similarity data to generate data association results corresponding to a best matching between the object data in the previous image frame and the object detection results from the current image frame; cause state transitions in finite state machines for each object according to the data association results; and provide as an output object tracking output data corresponding to the states of the finite state machines for each object.

Abstract: A system and method for taillight signal recognition using a convolutional neural network is disclosed. An example embodiment includes: receiving a plurality of image frames from one or more image-generating devices of an autonomous vehicle; using a single-frame taillight illumination status annotation dataset and a single-frame taillight mask dataset to recognize a taillight illumination status of a proximate vehicle identified in an image frame of the plurality of image frames, the single-frame taillight illumination status annotation dataset including one or more taillight illumination status conditions of a right or left vehicle taillight signal, the single-frame taillight mask dataset including annotations to isolate a taillight region of a vehicle; and using a multi-frame taillight illumination status dataset to recognize a taillight illumination status of the proximate vehicle in multiple image frames of the plurality of image frames, the multiple image frames being in temporal succession.

Abstract: A system and method for determining car to lane distance is provided. In one aspect, the system includes a camera configured to generate an image, a processor, and a computer-readable memory. The processor is configured to receive the image from the camera, generate a wheel segmentation map representative of one or more wheels detected in the image, and generate a lane segmentation map representative of one or more lanes detected in the image. For at least one of the wheels in the wheel segmentation map, the processor is also configured to determine a distance between the wheel and at least one nearby lane in the lane segmentation map. The processor is further configured to determine a distance between a vehicle in the image and the lane based on the distance between the wheel and the lane.

Abstract: A system and method for online real-time multi-object tracking is disclosed. A particular embodiment can be configured to: receive image frame data from at least one camera associated with an autonomous vehicle; generate similarity data corresponding to a similarity between object data in a previous image frame compared with object detection results from a current image frame; use the similarity data to generate data association results corresponding to a best matching between the object data in the previous image frame and the object detection results from the current image frame; cause state transitions in finite state machines for each object according to the data association results; and provide as an output object tracking output data corresponding to the states of the finite state machines for each object.

Abstract: A system and method for determining car to lane distance is provided. In one aspect, the system includes a camera configured to generate an image, a processor, and a computer-readable memory. The processor is configured to receive the image from the camera, generate a wheel segmentation map representative of one or more wheels detected in the image, and generate a lane segmentation map representative of one or more lanes detected in the image. For at least one of the wheels in the wheel segmentation map, the processor is also configured to determine a distance between the wheel and at least one nearby lane in the lane segmentation map. The processor is further configured to determine a distance between a vehicle in the image and the lane based on the distance between the wheel and the lane.

Abstract: A system and method providing truck-mounted sensors to detect trailer following vehicles and trailer conditions are disclosed. A system of an example embodiment comprises: a vehicle control subsystem installed in an autonomous truck, the vehicle control subsystem comprising a data processor; and a truck-mounted sensor subsystem installed on a portion of a tractor of the autonomous truck to which a trailer is attachable, the truck-mounted sensor subsystem being coupled to the vehicle control subsystem via a data connection, wherein the truck-mounted sensor subsystem is configured to emit electromagnetic waves propagating in a space under the trailer, to generate object data representing objects detected by receiving a reflection of the electromagnetic waves, and to transfer the object data to the vehicle control subsystem.

Abstract: A system and method for lateral vehicle detection is disclosed. A particular embodiment can be configured to: receive lateral image data from at least one laterally-facing camera associated with an autonomous vehicle; warp the lateral image data based on a line parallel to a side of the autonomous vehicle; perform object extraction on the warped lateral image data to identify extracted objects in the warped lateral image data; and apply bounding boxes around the extracted objects.

Abstract: A system and method for lateral vehicle detection is disclosed. A particular embodiment can be configured to: receive lateral image data from at least one laterally-facing camera associated with an autonomous vehicle; warp the lateral image data based on a line parallel to a side of the autonomous vehicle; perform object extraction on the warped lateral image data to identify extracted objects in the warped lateral image data; and apply bounding boxes around the extracted objects.

Abstract: A system and method for instance-level lane detection for autonomous vehicle control are disclosed.

Abstract: A system and method for semantic segmentation using hybrid dilated convolution (HDC) are disclosed. A particular embodiment includes: receiving an input image; producing a feature map from the input image; performing a convolution operation on the feature map and producing multiple convolution layers; grouping the multiple convolution layers into a plurality of groups; applying different dilation rates for different convolution layers in a single group of the plurality of groups; and applying a same dilation rate setting across all groups of the plurality of groups.

Abstract: A system and method for instance-level lane detection for autonomous vehicle control includes: receiving training image data from a training image data collection system; performing a training phase to train a plurality of tasks associated with features of the training image data, the training phase including extracting roadway lane marking features from the training image data, causing the plurality of tasks to generate task-specific predictions based on the training image data, determining a bias between the task-specific prediction for each task and corresponding task-specific ground truth data, and adjusting parameters of each of the plurality of tasks to cause the bias to meet a pre-defined confidence level; receiving image data from an image data collection system associated with an autonomous vehicle; and performing an operational phase including extracting roadway lane marking features from the image data, causing the plurality of trained tasks to generate instance-level lane detection results.