A method for 3D auto-labeling of objects with predetermined structural and physical constraints includes identifying initial object-seeds for all frames from a given frame sequence of a scene. The method also includes refining each of the initial object-seeds over the 2D/3D data, while complying with the predetermined structural and physical constraints to auto-label 3D object vehicles within the scene. The method further includes linking the auto-label 3D object vehicles over time into trajectories while respecting the predetermined structural and physical constraints.

System and methods for automated incident identification and reporting while operating a vehicle on the road using a device. The device identifies incidents using artificial intelligence neural networks trained to detect, classify, segment, and/or extract other information pertaining to objects of interest representing incidents. Additionally, a system and method for further storing, transmitting, processing, organizing and accessing the information graphically with respect to incident type, location, date and time during operation of the vehicle along the road.

Described herein are systems, devices, and methods for localizing and recognizing an object in an environment. In an example, a mobile cleaning robot comprises a drive system to move the mobile cleaning robot about an environment, an imaging sensor to take images of an object in the environment from different perspectives. The multiple observations include images of the object that is at least partially occluded by an obstacle. A controller circuit of the mobile robot can, for multiple different locations in a map of the environment, calculate respective fractional visibility values using the plurality of images. The fractional visibility values each represent a probability of the object being visible through the corresponding location. The controller circuit can localize and recognize the object based on the fractional visibility values at the multiple locations on the map.

Systems and methods use cameras to provide autonomous navigation features. In one implementation, a driver-assist object detection system is provided for a vehicle. One or more processing devices associated with the system receive at least two images from a plurality of captured images via a data interface. The device(s) analyze the first image and at least a second image to determine a reference plane corresponding to the roadway the vehicle is traveling on. The processing device(s) locate a target object in the first two images, and determine a difference in a size of at least one dimension of the target object between the two images. The system may use the difference in size to determine a height of the object. Further, the system may cause a change in at least a directional course of the vehicle if the determined height exceeds a predetermined threshold.

A vehicular vision system includes a camera disposed at an in-cabin side of a windshield of a vehicle and viewing forward of the vehicle. The vehicular vision system, responsive at least in part to image processing of multiple frames of captured image data, detects an object present exterior of the vehicle that is moving relative to the vehicle. The system, when the vehicle is moving, and based at least in part to received vehicle motion data indicative of motion of the vehicle when the vehicle is moving and image processing of multiple frames of captured image data, (i) estimates object trajectory of the detected object based at least in part on corresponding object features present in multiple frames of image data captured by the camera and (ii) determines motion of the detected object relative to the moving vehicle based on the estimated object trajectory.

Systems and methods are provided for constructing, using, and updating the sparse map for autonomous vehicle navigation. A system may comprise a processor and a memory. The memory may include instructions, which when executed on the processor, cause the processor to maintain a map; determine, based on analysis of image data, an existence of a non-transient condition that is inconsistent with the map, the image data from a camera integrated with the autonomous vehicle; and update the map.

A method includes calculating a stereo-disparity map between the initial image and the final image, for each column of the stereo-disparity map, calculating an average value of the stereo-disparities of the pixels of the column, calculating the slope and/or constant factor of a linear function approximating variations of said average values; and calculating said information relative to the relative speed between the object and the camera, based on the slope and/or the constant factor.

A system is comprised of one or more processors coupled to memory. The one or more processors are configured to receive image data based on an image captured using a camera of a vehicle and to utilize the image data as a basis of an input to a trained machine learning model to at least in part identify a distance of an object from the vehicle. The trained machine learning model has been trained using a training image and a correlated output of an emitting distance sensor.

A method and system provide for temporal fusion of depth maps in an image space representation. A series of depth maps are obtained/acquired from one or more depth sensors at a first time. A first Gaussian mixture model (GMM) is initialized using one of the series of depth maps. A second depth map is obtained from the depth sensors at a second time. An estimate of the motion of the depth sensors, from the first time to the second time, is received. A predictive GMM at the second time is created based on a transform of the first GMM and the estimate of the motion. The predictive GMM is updated based on the second depth map.

A method comprises: collecting first point cloud data under a static scenario around a target collecting point; building a static background mesh model under the static scenario around the target collecting point according to the first point cloud data; collecting a second point cloud data of a target frame under a dynamic scenario after a dynamic obstacle moves around the target collecting point; annotating the point cloud of dynamic obstacle in the second point cloud data corresponding to the target frame, according to the static background mesh model. Through the technical solution of the present disclosure, it is feasible to automatically annotate the point cloud of the dynamic obstacle, effectively save manpower and annotation time spent in annotating the dynamic obstacle point cloud, and thereby effectively improve the efficiency of annotating the dynamic obstacle.