Disclosed is a method of determining a disparity of an image generated by using a multibaseline stereo camera system. The method includes determining a reference parity between a reference image and a target image among multiple images generated by using a multi-baseline stereo camera system, determining an ambiguity region in each of the multiple images on the basis of a positional relationship among the multiple images or among cameras in the multibaseline stereo camera system, and determining a disparity for each of the multiple images by determining a matching point in each of the ambiguity regions of the respective images.

A system for disparity estimation includes one or more feature extractor modules configured to extract one or more feature maps from one or more input images; and one or more semantic information modules connected at one or more outputs of the one or more feature extractor modules, wherein the one or more semantic information modules are configured to generate one or more foreground semantic information to be provided to the one or more feature extractor modules for disparity estimation at a next training epoch.

A depth map generation device for merging multiple depth maps includes at least three image capturers, a depth map generator, and a mixer. The at least three image capturers form at least two image capture pairs. The depth map generator is coupled to the at least three image capturers for generating a depth map corresponding to each image capturer pair of the at least two image capture pairs according to an image pair captured by the each image capturer. The mixer is coupled to the depth map generator for merging at least two depth maps corresponding to the at least two image capturer pairs to generate a final depth map, wherein the at least two depth maps have different characteristics.

There is provided an image capturing apparatus that captures a plurality of images, calculates a three-dimensional position from the plurality of images, and outputs the plurality of images and information about the three-dimensional position. The image capturing apparatus includes an image capturing unit, a camera parameter storage unit, a position calculation unit, a position selection unit, and an image complementing unit. The image capturing unit outputs the plurality of images using at least three cameras. The camera parameter storage unit stores in advance camera parameters including occlusion information. The position calculation unit calculates three dimensional positions of a plurality of points. The position selection unit selects a piece of position information relating to a subject area that does not have an occlusion, and outputs selected position information. The image complementing unit generates a complementary image, and outputs the complementary image and the selected position information.

Techniques and systems are described herein for determining dynamic lighting for objects in images. Using such techniques and systems, a lighting condition of one or more captured images can be adjusted. Techniques and systems are also described herein for determining depth values for one or more objects in an image. In some cases, the depth values (and the lighting values) can be determined using only a single camera and a single image, in which case one or more depth sensors are not needed to produce the depth values.

A method for matching points between three images of a scene comprises retrieving three images acquired by a sensor, extracting blobs from said reflection in said two images; for each given extracted blob of the first image: selecting a selected epipolar plane; identifying plausible combinations; calculating a matching error; repeating the steps of selecting, identifying and calculating for each epipolar plane of the set of epipolar planes; determining a most probable combination; identifying matching points between the two images; validating the matching points between the two images, said validating comprising for each pair of matching points, determining a projection of the pair of matching points in a third image of the third camera; and providing the validated pairs of matching points.

There is disclosed a system for generating a three-dimensional model of a physical object including a camera skid placed at a known distance from the physical object and moved fully around the physical object at the known distance, a set of cameras on the camera skid for capturing image data at a series of locations fully around the physical object, and a computing device for generating a three-dimensional model of the physical object using the known distance and the image data.

Various embodiments are disclosed for detecting, tracking and counting objects of interest in video. In an embodiment, a method of detecting and tracking objects of interest comprises: obtaining, by a computing device, multiple frames of images from an image capturing device; detecting, by the computing device, objects of interest in each frame; accumulating, by the computing device, multiple frames of object detections; creating, by the computing device, object tracks based on a batch of object detections over multiple frames; and associating, by the computing device, the object tracks over consecutive batches.

A method of sorting is described, and which includes providing a product stream formed of individual objects of interest having feature aspects which can be detected; generating multiple images of each of the respective objects of interest; classifying the feature aspects of the objects of interest; identifying complementary images by analyzing some of the multiplicity of images; fusing the complementary images to form an aggregated region representation of the complementary images; and sorting the respective objects of interest based at least in part upon the aggregated region representation which is formed.

A dual-mode data capture system includes a capture controller, a point cloud generator, a collision detector, a plurality of cameras viewing a capture volume, and a motion sensor to generate a detection signal when an object arrives at a capture position within the volume. The controller: activates a subset of cameras in a collision detection mode to capture sequences of images of the volume; responsive to receiving the detection signal, activates the cameras in a dimensioning mode to capture a synchronous set of images of the capture position. The collision detector: determines whether the sequences of images indicate a potential collision; and responsive to detection of a potential collision, generates a warning. The point cloud generator: receives the synchronous set of images and generates a point cloud representing the object based on the synchronous set of images, for use in determining dimensions of the object.