The present disclosure provides a spatial parking place detection method and device, a storage medium and a program product, which relate to the field of data processing and, in particular, to the fields of computer vision, autonomous parking and autonomous driving. A specific implementation lies in: acquiring ultrasonic data around a vehicle collected by an ultrasonic sensor on the vehicle, and image data around the vehicle collected by an image collection apparatus; determining a first spatial parking place around the vehicle according to the ultrasonic data, and determining a second spatial parking place around the vehicle according to the image data; fusing the first spatial parking place and the second spatial parking place that are located at an identical position to determine a spatial parking place at that position; and checking the availability of the detected spatial parking place, and determining available spatial parking places of the vehicle.

An obstacle recognition device of a vehicle provided with a camera capturing an image around the vehicle, includes an acquiring unit sequentially acquiring the image captured by the camera; a feature point extracting unit extracting a plurality of feature points of an object included in the image; a calculation unit calculating each motion distance of the plurality of feature points between the image previously acquired and the image currently acquired by the acquiring unit; a first determination unit determining whether each motion distance of the feature points is larger than or equal to a first threshold; a second determination unit determining whether each motion distance of the feature points is larger than or equal to a second threshold; and an obstacle recognition unit recognizing an obstacle.

An obstacle identification apparatus acquires an image that is captured by a camera that is mounted to a vehicle. The obstacle identification apparatus calculates a first gradient that is a gradient in a first direction of a luminance value of pixels in the image and a second gradient that is a gradient of the luminance value in a second direction orthogonal to the first direction of the first gradient. Based on the first gradient and the second gradient, the obstacle identification apparatus estimates a shadow boundary that is a boundary between an own-vehicle shadow that is a shadow of the own vehicle and an object outside the vehicle. Based on the estimated shadow boundary, the obstacle identification apparatus estimates the own-vehicle shadow.

Generally, the disclosed systems and methods implement improved detection of objects in three-dimensional (3D) space. More particularly, an improved 3D object detection system can exploit continuous fusion of multiple sensors and/or integrated geographic prior map data to enhance effectiveness and robustness of object detection in applications such as autonomous driving. In some implementations, geographic prior data (e.g., geometric ground and/or semantic road features) can be exploited to enhance three-dimensional object detection for autonomous vehicle applications. In some implementations, object detection systems and methods can be improved based on dynamic utilization of multiple sensor modalities. More particularly, an improved 3D object detection system can exploit both LIDAR systems and cameras to perform very accurate localization of objects within three-dimensional space relative to an autonomous vehicle. For example, multi-sensor fusion can be implemented via continuous convolutions to fuse image data samples and LIDAR feature maps at different levels of resolution.

Systems and methods determining velocity of an object associated with a three-dimensional (3D) scene may include: a LIDAR system generating two sets of 3D point cloud data of the scene from two consecutive point cloud sweeps; a pillar feature network encoding data of the point cloud data to extract two-dimensional (2D) bird's-eye-view embeddings for each of the point cloud data sets in the form of pseudo images, wherein the 2D bird's-eye-view embeddings for a first of the two point cloud data sets comprises pillar features for the first point cloud data set and the 2D bird's-eye-view embeddings for a second of the two point cloud data sets comprises pillar features for the second point cloud data set; and a feature pyramid network encoding the pillar features and performing a 2D optical flow estimation to estimate the velocity of the object.

Methods, systems, and devices for object detection are described. An example method for object detection is provided which may include capturing at least three polarization angles of a scene. The method may include translating polarization parameters associated with the at least three polarization angles to a reference angle to create a vector map and resolving the vector map into parallel components and perpendicular components, wherein the parallel components are parallel to a plane of incidence of light in the scene and the perpendicular components are perpendicular. The method may further include determining a range map based at least in part on the parallel and perpendicular components, detecting an object present in the scene using the range map and an airlight scattering polarization component, and outputting an indication of the object.

A vehicle periphery monitoring device includes a plurality of sensors including a rear camera, a rear right-side camera, and a rear left-side camera, a display unit that is provided inside a vehicle cabin, memory, and a processor that is coupled to the memory. The processor is configured so as to acquire a rear image that includes a first image of a vehicle rear side acquired by the rear camera, a second image of a vehicle rear right-side acquired by the rear right-side camera, and a third image of a vehicle rear left-side acquired by the rear left-side camera, acquire from the plurality of sensors relative positions of a target that is present in areas including the vehicle rear side, the vehicle rear right-side, and the vehicle rear left-side relative to a host vehicle, determine whether or not the acquired relative positions of the target are mutually consistent with each other in the plurality of sensors, display a single composite image that is created by combining the first image, the second image, and the third image at the display unit in a case in which it is determined that the relative positions of the target are mutually consistent with each other, and display the first image, the second image, and the third image individually and adjacently to each other on the display unit in a case in which it is determined that the relative positions of the target are not mutually consistent with each other.

In various examples, systems and methods are disclosed that detect hazards on a roadway by identifying discontinuities between pixels on a depth map. For example, two synchronized stereo cameras mounted on an ego-machine may generate images that may be used extract depth or disparity information. Because a hazard's height may cause an occlusion of the driving surface behind the hazard from a perspective of a camera(s), a discontinuity in disparity values may indicate the presence of a hazard. For example, the system may analyze pairs of pixels on the depth map and, when the system determines that a disparity between a pair of pixels satisfies a disparity threshold, the system may identify the pixel nearest the ego-machine as a hazard pixel.

Systems and methods for training an adapter network that adapts a model pre-trained on synthetic images to real-world data are disclosed herein. A system may include a processor and a memory in communication with the processor and having machine-readable that cause the processor to output, using a neural network, a predicted scene that includes a three-dimensional bounding box having pose information of an object, generate a rendered map of the object that includes a rendered shape of the object and a rendered surface normal of the object, and train the adapter network, which adapts the predicted scene to adjust for a deformation of the input image by comparing the rendered map to the output map acting as a ground truth.

Methods and system for training a neural network for depth estimation in a vehicle. The methods and systems receive respective training image data from at least two cameras. Fields of view of adjacent cameras of the at least two cameras partially overlap. The respective training image data is processed through a neural network providing depth data and semantic segmentation data as outputs. The neural network is trained based on a loss function. The loss function combines a plurality of loss terms including at least a semantic segmentation loss term and a panoramic loss term. The panoramic loss term includes a similarity measure regarding overlapping image patches of the respective image data that each correspond to a region of overlapping fields of view of the adjacent cameras. The semantic segmentation loss term quantifies a difference between ground truth semantic segmentation data and the semantic segmentation data output from the neural network.