Abstract: A method for identifying a position of an object in a cargo space includes identifying a region of interest (ROI) from a field of view of the cargo space being captured by a camera, where the camera captures at least one of depth and color of the field of view, extracting a plurality of planes from the ROI, where the plurality of planes correspond to a geometry of the object, clustering similar and nearby planes, where the clustering is based on a weight of two or more planes of the plurality of planes and where the weight is assigned based on a property of orthogonality and a property of dimensions of two or more planes of the plurality of planes, modelling a multi-dimensional bounding box corresponding to the object based on the clustered planes, and identifying a position of the object based on a position of the multi-dimensional bounding box.

Abstract: Systems and methods for generating responses to queries from a client application are disclosed. One example method includes receiving a first query from the client application, determining an authorization context for the first query based at least in part on a set of authorization restrictions corresponding to the client application, generating a response to the first query based at least in part on an intermediate response generated by a query engine, the response redacted based at least in part on the authorization context, and providing the response to the client application.

Abstract: A method of determining volume of a cargo space of a vehicle in real-time includes generating an initial spatial model of the cargo space using images from a plurality of cameras including cameras for capturing at least one of depth and color, positioned in and around the cargo space, where the images are stitched together into a dense point cloud in real-time, generating an updated spatial model of the cargo space using the images, upon detection of items being loaded or unloaded into the cargo space, where the updated spatial model includes changes to the cargo space, estimating a volume of loaded items in the updated spatial model, and determining a remaining volume of the cargo space based on the estimated volume of the loaded items and a total volume of the cargo space, where the total volume is calculated based on the initial spatial model.

Abstract: A method for identifying a position of an object in a cargo space includes identifying a region of interest (ROI) from a field of view of the cargo space being captured by a camera, where the camera captures at least one of depth and color of the field of view, extracting a plurality of planes from the ROI, where the plurality of planes correspond to a geometry of the object, clustering similar and nearby planes, where the clustering is based on a weight of two or more planes of the plurality of planes and where the weight is assigned based on a property of orthogonality and a property of dimensions of two or more planes of the plurality of planes, modelling a multi-dimensional bounding box corresponding to the object based on the clustered planes, and identifying a position of the object based on a position of the multi-dimensional bounding box.

Abstract: A system of calibrating a multi-camera device that include two or more planar calibration targets, arranged in a non-planar orientation. The system mechanically moves the calibration targets relative to the multi-camera device to change the relative orientation. The system records images with the multi-camera device at various orientations. The recorded images contain one or more of the constituent planar calibration targets. The system identifies and extracts the pixel coordinates of each constituent planar calibration target from each image.

Abstract: A system of calibrating a multi-camera device that include two or more planar calibration targets, arranged in a non-planar orientation. The system mechanically moves the calibration targets relative to the multi-camera device to change the relative orientation. The system records images with the multi-camera device at various orientations. The recorded images contain one or more of the constituent planar calibration targets. The system identifies and extracts the pixel coordinates of each constituent planar calibration target from each image.