The present invention relates to a method for predicting a collision between an examination subject and an imaging apparatus, and an imaging apparatus. The prediction method may include: acquiring an image package of the examination subject via a multi-modal camera system, the image package including a depth image and a thermal image of the examination subject, and the multi-modal camera system including a depth camera module and a thermal camera module; acquiring a 2D contour of the examination subject based on segmentation processing performed on the thermal image; generating a 3D contour of the examination subject based on the 2D contour of the examination subject and the depth image of the examination subject; and estimating, based on the 3D contour of the examination subject, whether the examination subject will collide, on a movement path thereof, with an imaging apparatus scanning the examination subject. The imaging apparatus provided in the present invention can achieve the same prediction effect.

Methods and image processing systems are provided for determining a dominant gradient orientation for a target region within an image. A plurality of gradient samples are determined for the target region, wherein each of the gradient samples represents a variation in pixel values within the target region. The gradient samples are converted into double-angle gradient vectors, and the double-angle gradient vectors are combined so as to determine a dominant gradient orientation for the target region.

The present disclosure relates to systems, non-transitory computer-readable media, and methods for adapting generative neural networks to target domains utilizing an image translation neural network. In particular, in one or more embodiments, the disclosed systems utilize an image translation neural network to translate target results to a source domain for input in target neural network adaptation. For instance, in some embodiments, the disclosed systems compare a translated target result with a source result from a pretrained source generative neural network to adjust parameters of a target generative neural network to produce results corresponding in features to source results and corresponding in style to the target domain.

Captured samples of a physical structure or other scene are mapped to a predetermined multi-dimensional coordinate space, and spatially-adjacent samples are organized into array cells representing subspaces thereof. Each cell is classified according to predetermined target-identifying criteria for the samples of the cell. A cluster of spatially-contiguous cells of common classification, peripherally bounded by cells of different classification, is constructed, and a boundary demarcation is defined from the peripheral contour of the cluster. The boundary demarcation is overlaid upon a visual display of the physical scene, thereby visually demarcating the boundaries of a detected target of interest.

A processor, method, and non-transitory computer-readable storage medium for processing template data and search data according to a search window applied to the search data. The search window comprising a set of offset positions. The processing is performed by a block matching engine (BME) that produces a tensor with difference values, and a convolutional engine (CE) that performs a convolutional operation on the tensor. The processing is performed in an iterative interleaved fashion, by dividing the set of offset positions into a plurality of subsets of offset positions. In parallel with processing of the first X offset positions by the CE, the BME generates the next X channels of the tensor, and which are subsequently pipelined through to the CE via an internal storage, etc.

A system for creating online vectorized maps for autonomous vehicles includes an image sensor and an electronic control unit (ECU). The image sensor captures a series of image frames. The ECU includes a memory, a central processing unit (CPU), and a transceiver. The memory stores a semantic segmentation deep learning model and a vectorization post-processing module as computer readable code. The CPU executes the semantic segmentation deep learning model and the vectorization post-processing module to output a vectorized map of an external environment of a vehicle. The transceiver uploads the vectorized map to a server such that the vectorized map can be accessed by a second vehicle that uses the vectorized map to traverse the external environment.

Methods and image processing systems are provided for determining a dominant gradient orientation for a target region within an image. A plurality of gradient samples are determined for the target region, wherein each of the gradient samples represents a variation in pixel values within the target region. The gradient samples are converted into double-angle gradient vectors, and the double-angle gradient vectors are combined so as to determine a dominant gradient orientation for the target region.

Methods are used by a thermal acoustic imaging (TAI) inspection system to identify potential defects within a component scanned, such as an engine fan blade. The inspection system generates a TAI scan having a plurality of frames. Indications are identified in at least one frame of the plurality of frames. An extractor model is applied to the plurality of frames to extract a plurality of spatial features corresponding to the indication. The plurality of spatial features is concatenated from each frame to generate a time series value of the values for the features. The plurality of spatial features is combined into multiple sequence data. The multiple sequence data includes a plurality of temporal features. The multiple sequence data is provided to train a neural network model to predict defects. For trained neural network models, the multiple sequence data is used to predict if the indication is a defect.

A filling rate measurement method includes: obtaining a space three-dimensional model generated by measuring a first storage having an opening and a first storage space in which a measurement target is to be stored, the measuring being performed using a range sensor facing the first storage; obtaining a storage three-dimensional model that is a three-dimensional model of the first storage in which the measurement target is not stored; extracting a target portion corresponding to the measurement target from the space three-dimensional model using the space three-dimensional model and the storage three-dimensional model; calculating a first three-dimensional coordinate system; estimating a target three-dimensional model using the target portion and the first three-dimensional coordinate system, the target three-dimensional model being a three-dimensional model of the measurement target in the first storage space; and calculating a first filling rate of the measurement target with respect to the first storage space.

A system for creating online vectorized maps for autonomous vehicles includes an image sensor and an electronic control unit (ECU). The image sensor captures a series of image frames. The ECU includes a memory, a central processing unit (CPU), and a transceiver. The memory stores a semantic segmentation deep learning model and a vectorization post-processing module as computer readable code. The CPU executes the semantic segmentation deep learning model and the vectorization post-processing module to output a vectorized map of an external environment of a vehicle. The transceiver uploads the vectorized map to a server such that the vectorized map can be accessed by a second vehicle that uses the vectorized map to traverse the external environment.