A method for training a machine learning model for recognizing an object topology of an object from an image of the object. The method includes obtaining a 3D model of the object, wherein the 3D model comprises a mesh of vertices connected by edges, wherein each edge has a weight which specifies proximity of two vertices connected by the edge in the object; determining a descriptor for each vertex of the mesh by searching descriptors for the vertices which minimize the sum, over pairs of connected vertices, of distances between the descriptors of the pair of vertices weighted by the weight of the edge between the pair of vertices; generating training data image pairs, wherein each training data image pair comprises a training input image showing the object and a target image; and training the machine learning model by supervised learning using the training data image pairs as training data.

A computer-implemented method for learning an autoencoder notably is provided. The method includes obtaining a dataset of images. Each image includes a respective object representation. The method also includes learning the autoencoder based on the dataset. The learning includes minimization of a reconstruction loss. The reconstruction loss includes a term that penalizes a distance for each respective image. The penalized distance is between the result of applying the autoencoder to the respective image and the set of results of applying at least part of a group of transformations to the object representation of the respective image. Such a method provides an improved solution to learn an autoencoder.

Embodiments of the present disclosure provide methods for determining an object location of an object. In the method, a group of measurement locations for a group of feature marks in the object are collected from a group of sensors, respectively. A group of estimation locations are obtained for the group of feature marks based on the object location and a group of offsets between the group of estimation locations and the object location, respectively. An error function is generated based on the group of measurement locations and the group of estimation locations. The object location is determined based on the error function. With these embodiments, performance and accuracy for determining the object location may be greatly increased.

A system and method for estimating dimensions of an approximately cuboidal object from a 3D image of the object acquired by an image sensor of the vision system processor is provided. An identification module, associated with the vision system processor, automatically identifies a 3D region in the 3D image that contains the cuboidal object. A selection module, associated with the vision system processor, automatically selects 3D image data from the 3D image that corresponds to approximate faces or boundaries of the cuboidal object. An analysis module statistically analyzes, and generates statistics for, the selected 3D image data that correspond to approximate cuboidal object faces or boundaries. A refinement module chooses statistics that correspond to improved cuboidal dimensions from among cuboidal object length, width and height. The improved cuboidal dimensions are provided as dimensions for the object. A user interface displays a plurality of interface screens for setup and runtime operation.

Systems and methods for determining an alignment of a trailer relative to a docking bay or a vehicle bay door using dynamic depth filtering. Image data and position data is captured by a 3D camera system with an at least partially downward-facing field of view. When a trailer is approaching the docking bay or door, the captured image data includes a top surface of the trailer. A dynamic height range is determined based on an estimated height of the top surface of the trailer in the image data and a dynamic depth filter is applied to filter out image data corresponding to heights outside of the dynamic height range. An angular position and/or lateral offset of the trailer is determined based on the depth-filtered image data.

A method and a device for 3D shape matching based on a local reference frame are proposed. After acquiring a 3D point cloud and feature points in the method, the feature point set is projected to a plane, and feature transformation is performed on the projected points by using at least one factor from the distances between the 3D points and the feature points, the distances between the 3D points and the projected points, and the average distances between the 3D points and its 1-ring neighboring points to acquire a point distribution with a larger variance in a certain direction than the projected point set, and the local reference frame is determined based on the transformed point distribution. The 3D local feature descriptor established based on this local reference frame can encode the 3D local surface information more robustly, so as to obtain a better 3D shape matching effect.

A method for controlling personal protective equipment. The method includes positioning of a person equipped with personal protective equipment next to an electrical enclosure, wherein a specific personal protective equipment requirement is defined for the electrical enclosure. The method further includes scanning the personal protective equipment in a contact-less manner while the person is next to the electrical enclosure. Thereby scanned personal protective equipment information are generated. The method further includes comparing the scanned personal protective equipment information with the specific personal protective equipment requirement and evaluating, based on the comparison, whether the personal protective equipment is in accordance with the specific personal protective equipment requirement. The method further includes providing a feedback indicating whether or not the personal protective equipment is in accordance with the specific personal protective equipment requirement.

Described herein are a system and a method for object recognition via a computer vision application, the system including at least the following components: an object to be recognized, the object having object specific reflectance and luminescence spectral patterns, a light source which is configured to project at least one light pattern on a scene which includes the object to be recognized, a sensor which is configured to measure radiance data of the scene including the object when the scene is illuminated by the light source, a data storage unit which includes luminescence spectral patterns together with appropriately assigned respective objects, and a data processing unit.

The embodiments of the present disclosure provide a visual positioning method, the method may include obtaining a positioning image collected by an imaging device; obtaining a three-dimensional (3D) point cloud map associated with an area where the imaging device is located; determining a target area associated with the positioning image from the 3D point cloud map based on the positioning image; and

determining positioning information of the imaging device based on the positioning image and the target area.

Generating a 3D model may include determining, based on sensor data from a touch sensor on a first device, a touch event, wherein the touch event comprises a touch on the first device by a touching object. Generating a 3D model may also include, in response to a touch event, obtaining a first image of the touching object by a first camera of the first device, and obtaining, from a second device, a second image of the touching object, wherein the first image of the touching object captures a first view of the touching object, and wherein the second image of the touching object captures a second view of the touching object. A model of the touching object is generated based on the first image and the second image.