Embodiments is disclosed to detect a locked heading direction of a pedestrian and to predict a path for the pedestrian using the locked heading direction. According to one embodiment, a system perceives an environment of an autonomous driving vehicle (ADV) using one or more image capturing devices. The system detects a pedestrian in the perceived environment. The system determines a facing direction of the pedestrian relative to the ADV as one of left/right side, front, or back. If the facing direction of the pedestrian is determined to be front or back facing, the system determines a lane nearest to the pedestrian. The system projects the pedestrian onto the nearest lane to determine a lane direction at the projection. The system determines a heading direction for the pedestrian locking to the lane direction of the nearest lane based on a predetermined condition.

A method includes generating a first warped image based on a pose and a depth estimated from a current image and a previous image in a sequence of images captured by a camera of the agent. The method also includes estimating a motion of dynamic object between the previous image and the target image. The method further includes generating a second warped image from the first warped image based on the estimated motion. The method still further includes controlling an action of an agent based on the second warped image.

A method and system by which a bounding box disposed around a segmented object in a camera (or other perception sensor) 2D image can be used to produce an estimate for both the location of the object—its position relative to the position of the camera that obtained the image (i.e., translation)—and the angle of rotation of the surface that the object is located on. The method and system may be used by an advanced driver assistance system (ADAS), an autonomous driving (AD) system, or the like. The input includes a simple camera (or other perception sensor) 2D image, with the ego vehicle generating 2D or 3D bounding boxes for objects detected at the scene. The output includes, for each object, its estimated distance from the ego vehicle camera/perception sensor and the angle of rotation of the surface underneath the object relative to the surface underneath the ego vehicle.

A method and device for identifying a stereoscopic object, and a vehicle and a storage medium are described. They are used for solving the problem that a monocular camera cannot determine whether an object to be identified is a stereoscopic object. The method is applied to a vehicle, and the method comprises: during driving, photographing an object to be identified by means of a monocular camera on the vehicle, so as to obtain a plurality of images (S11); determining, according to the plurality of images, an imaging change rule of the object to be identified that is projected onto an imaging plane of the monocular camera, wherein the imaging changes along with a change in the distance between the object to be identified and the vehicle (S12); and if the imaging change rule matches a quadratic curve, determining that the object to be identified is a stereoscopic object (S13).

Aerial vehicles that are equipped with one or more imaging devices may detect obstacles that are small in size, or obstacles that feature colors or textures that are consistent with colors or textures of a landing area, using pairs of images captured by the imaging devices. Disparities between pixels corresponding to points of the landing area that appear within each of a pair of the images may be determined and used to generate a reconstruction of the landing area and a difference image. If either the reconstruction or the difference image indicates the presence of one or more obstacles, a landing operation at the landing area may be aborted or an alternate landing area for the aerial vehicle may be identified accordingly.

According to one aspect, scene classification may be provided. An image capture device may capture a series of image frames of an environment from a moving vehicle. A temporal classifier may classify image frames with temporal predictions and generate a series of image frames associated with respective temporal predictions based on a scene classification model. The temporal classifier may perform classification of image frames based on a convolutional neural network (CNN), a long short-term memory (LSTM) network, and a fully connected layer. The scene classifier may classify image frames based on a CNN, global average pooling, and a fully connected layer and generate an associated scene prediction based on the scene classification model and respective temporal predictions. A controller of a vehicle may activate or deactivate vehicle sensors or vehicle systems of the vehicle based on the scene prediction.

An agricultural method includes providing a positive air pressure chamber to prevent outside contaminants from entering the chamber; growing crops in a plurality of cells in the chamber, each cell having multi-grow benches or levels, each cell further having connectors to vertical hoists for vertical movements in the chamber; maintaining pre-set temperature, humidity, carbon dioxide, watering and lighting levels to achieve predetermined plant growth; using motorized transport rails to deliver benches for operations including seeding, harvesting, grow media recovery, and bench wash; dispensing seeds in the cell with a mechanical seeder coupled to the transport rails; growing the crops with computer controlled nutrients, light and air level; and harvesting the crops and delivering the harvested crop at a selected outlet of the chamber.

A system may include at least one processor including circuitry and a memory. The memory may include instructions executable by the circuitry to cause the at least one processor programmed to receive at least one identifier associated with a condition having at least one dynamic characteristic. The at least one identifier may be determined based on acquisition, from a camera associated with a host vehicle, of at least one image representative of an environment of the host vehicle, and analysis of the at least one image to identify the condition in the environment, and analysis of the at least one image to determine the at least one identifier associated with the condition. The at least one processor may also be programmed to update a database record to include the at least one identifier associated with the condition, and distribute the database record to at least one entity.

An obstacle detection method and device, an apparatus and a storage medium are provided, which are related to a field of intelligent transportation. The specific implementation includes: acquiring position information of a two-dimensional (2D) detection frame and position information of a three-dimensional (3D) detection frame of an obstacle in an image; converting the position information of the 3D detection frame of the obstacle into position information of a 2D projection frame of the obstacle; and optimizing the position information of the 3D detection frame of the obstacle by using the position information of the 2D detection frame, the position information of the 3D detection frame and the position information of the 2D projection frame of the obstacle in the image. Accuracy of results of predicting a 3D position of an obstacle by a roadside, on-board sensing device, or other sensing devices may be improved.

An on-road obstacle detection device that includes: a memory; and a processor, the processor being connected to the memory and being configured to: assign a semantic label to each pixel in an image using a first discriminator that has been pre-trained using images in which an on-road obstacle is not present; and detect an on-road obstacle based on a probability density of the semantic label assigned.