A pattern matching unit carries out a pattern matching between a photographed image obtained by photographing a workpiece with a monocular camera and a first plurality of models having a plurality of sizes and a plurality of angles, and selects a model having a size and an angle with the highest degree of matching. A primary detection unit detects a position and an angle of an uppermost workpiece based on the selected model. An actual load height calculation unit calculates an actual load height of the uppermost workpiece based on a hand height. A secondary detection unit re-detects the position and the angle of the uppermost workpiece based on a model having a size and an angle with the highest degree of matching selected by carrying out a pattern matching between the photographed image and a second plurality of models selected or newly created based on the actual load height.

Techniques are described for dynamically correcting image distortion during imaging of a patterned sample having repeating spots. Different sets of image distortion correction coefficients may be calculated for different regions of a sample during a first imaging cycle of a multicycle imaging run and subsequently applied in real time to image data generated during subsequent cycles. In one implementation, image distortion correction coefficients may be calculated for an image of a patterned sample having repeated spots by: estimating an affine transform of the image; sharpening the image; and iteratively searching for an optimal set of distortion correction coefficients for the sharpened image, where iteratively searching for the optimal set of distortion correction coefficients for the sharpened image includes calculating a mean chastity for spot locations in the image, and where the estimated affine transform is applied during each iteration of the search.

Devices, methods, and non-transitory program storage devices are disclosed herein to provide for improved perspective distortion correction for wide field of view (FOV) video image streams. The techniques disclosed herein may be configured, such that the distortion correction applied to requested region of interest (ROI) portions taken from individual images of the wide FOV video image stream smoothly transitions between applying different distortion correction to ROIs, depending on their respective FOVs. In particular, the techniques disclosed herein may modify the types and/or amounts of perspective distortion correction applied, based on the FOVs of the ROIs, as well as their location within the original wide FOV video image stream. In some cases, additional perspective distortion correction may also be applied to account for tilt in an image capture device as the wide FOV video image stream is being captured and/or the unwanted inclusion of “invalid” pixels from the wide FOV image.

An image processing system has one or more memories and image processing circuitry coupled to the one or more memories. The image processing circuitry, in operation, compares a first image to feature data in a comparison image space using a matching model. The comparing includes: unwarping keypoints in keypoint data of the first image; and comparing the unwarped keypoints and descriptor data associated with the first image to the feature data of the comparison image. The image processing circuitry determines whether the first image matches the comparison image based on the comparing.

A method for training a face recognition model includes: acquiring a plurality of first training images being uncovered face images, and acquiring a plurality of covering object images; generating a plurality of second training images by separately fusing the plurality of covering object images with the uncovered face images; and training the face recognition model by inputting the plurality of first training images and the plurality of second training images into the face recognition model.

In various examples, live perception from wide-view sensors may be leveraged to detect features in an environment of a vehicle. Sensor data generated by the sensors may be adjusted to represent a virtual field of view different from an actual field of view of the sensor, and the sensor data—with or without virtual adjustment—may be applied to a stereographic projection algorithm to generate a projected image. The projected image may then be applied to a machine learning model—such as a deep neural network (DNN)—to detect and/or classify features or objects represented therein. In some examples, the machine learning model may be pre-trained on training sensor data generated by a sensor having a field of view less than the wide-view sensor such that the virtual adjustment and/or projection algorithm may update the sensor data to be suitable for accurate processing by the pre-trained machine learning model.

In one embodiment, a method includes accessing an image of a damaged object. The method further includes determining, using a plurality of image segmentation models, a plurality of objects in the image. The method further includes determining, using a plurality of visual inference models and the determined plurality of objects from the image segmentation models, a repair-relevant property vector for the damaged object in the image. The repair-relevant property vector includes a plurality of damaged object properties. The method further includes generating a repair report using the repair-relevant property vector and a price catalogue. The repair report includes an indication of the damaged object and a price associated with the repair or replacement of the damaged object. The method further includes providing the generated report for display on an electronic display device.

System and techniques for vehicle environment modeling with a camera are described herein. A device for modeling an environment comprises: a hardware sensor interface to obtain a sequence of unrectified images representative of a road environment, the sequence of unrectified images including a first unrectified image, a previous unrectified image, and a previous-previous unrectified image; and processing circuitry to: provide the first unrectified image, the previous unrectified image, and the previous-previous unrectified image to an artificial neural network (ANN) to produce a three-dimensional structure of a scene; determine a selected homography; and apply the selected homography to the three-dimensional structure of the scene to create a model of the road environment.

In some arrangements, product packaging is digitally watermarked over most of its extent to facilitate high-throughput item identification at retail checkouts. Imagery captured by conventional or plenoptic cameras can be processed (e.g., by GPUs) to derive several different perspective-transformed views—further minimizing the need to manually reposition items for identification. Crinkles and other deformations in product packaging can be optically sensed, allowing such surfaces to be virtually flattened to aid identification. Piles of items can be 3D-modelled and virtually segmented into geometric primitives to aid identification, and to discover locations of obscured items. Other data (e.g., including data from sensors in aisles, shelves and carts, and gaze tracking for clues about visual saliency) can be used in assessing identification hypotheses about an item. Logos may be identified and used—or ignored—in product identification. A great variety of other features and arrangements are also detailed.

The present application discloses a coding pattern, coding and reading methods for the same, a calibration board, and a calibration method. In the present application, the coding pattern comprises four positioning blocks, wherein three of the positioning blocks are located at three corner portions of the coding pattern, and the remaining positioning block only contacts one edge of the coding pattern, thereby forming an asymmetrical distribution configuration of the four positioning blocks. The invention can replace a two-dimensional code standard in the related art, saving the licensing and manufacturing costs of using two-dimensional code generation software in the related art, and is not subject to usage restrictions of the two-dimensional code generation software in the related art. The invention enables accurate and fast positioning of the coding pattern, facilitates quick discovery of an initial coding block position in a coding region, and allows a customizable size of the coding pattern and a customizable size of a coding region in the coding pattern according to a data volume of an application scenario, thereby providing flexibility in configuring data information recorded in the coding pattern. In addition, coding of the coding region is simple, thereby improving efficiency.