The present disclosure provides a data processing method, an apparatus, an electronic device and a medium, which relates to the technical fields of autonomous driving, electronic maps, deep learning, image processing, and the like. The method includes: a computing device inputs a reference image and a captured image into a feature extraction model; obtain, a set of reference descriptors based on the first descriptor map; determine a plurality of sets of training descriptors; obtain a predicted pose of the vehicle by inputting the plurality of training poses and a plurality of similarities into a pose prediction model; and train the feature extraction model and the pose prediction model. When applied to a vehicle localization system, the trained feature extraction model and pose prediction model according to some embodiments of the present disclosure can improve accuracy and robustness of vehicle localization, thereby boosting the performance of the vehicle localization system.

A probabilistic model is provided based on an output of a matching procedure that matches a particular object to representations of objects, where the probabilistic model relates a probability of an object being present to a number of matching features. The probabilistic model is used for detecting whether a particular object is present in received visual data.

A main object determination apparatus includes an image acquisition unit configured to acquire images captured at different timings, a selection unit configured to select main object candidate(s) from objects in the images, a determination unit configured to determine whether the main object candidate(s) each selected at the respective different timings are the same, and an input unit configured to receive an operation. In a case where the determination unit determines that the main object candidate(s) selected by the selection unit in an image of interest and one or more images captured within a predetermined time before the image of interest is captured are the same, the determination unit determines the main object candidate(s) to be a main object. In a case where the input unit receives an instruction to specify a new main object, the determination unit switches the main object according to the instruction.

Methods and apparatus to match images using semantic features are disclosed. An example apparatus includes a semantic labeler to determine a semantic label for each of a first set of points of a first image and each of a second set of points of a second image; a binary robust independent element features (BRIEF) determiner to determine semantic BRIEF descriptors for a first subset of the first set of points and a second subset of the second set of points based on the semantic labels; and a point matcher to match first points of the first subset of points to second points of the second subset of points based on the semantic BRIEF descriptors.

An estimation apparatus according to an embodiment of the present disclosure includes a memory and a hardware processor coupled to the memory. The hardware processor is configured to: acquire first point cloud data; generate, from the first point cloud data, second point cloud data in which an attention point and at least one observation point are combined, the attention point gaining attention as a target of attribute estimation; and estimate an attribute of the attention point by calculating, for each attribute, a belonging probability of belonging to the attribute by using the second point cloud data.

A method, system and computer-readable storage medium for performing a cognitive information processing operation. The cognitive information processing operation includes: receiving data from a plurality of data sources; processing the data from the plurality of data sources to provide cognitively processed insights via an augmented intelligence system, the augmented intelligence system executing on a hardware processor of an information processing system, the augmented intelligence system and the information processing system providing a cognitive computing function; performing an explainability with recourse operation, the explainability with recourse operation providing an assurance explanation regarding the cognitive computing function; and, providing the cognitively processed insights to a destination, the destination comprising a cognitive application, the cognitive application enabling a user to interact with the cognitive insights.

An example system includes a feature extraction engine. The feature extraction engine is to determine a plurality of scale-dependent features for a portion of a target. The system also includes a signature-generation engine to select a subset of the plurality of scale-dependent features based on a strength of each feature. The signature-generation engine also is to store a numeric representation of the portion of the target and the subset of the plurality of scale-dependent features.

A method to train a model for feature point detection involves obtaining a first image and a second image. The method involves generating a first score map for the first image and a second score map for the second image using the model. The method involves selecting a first plurality of interest points in the first image based on the first score map. The method involves selecting a second plurality of interest points in the second image based on the second score map. Pairwise matching of a first interest point among the first plurality of interest points with a second interest point among the second plurality of interest points is performed. Correctness of the pairwise matches is checked based on a ground truth, to generate a reward map. The score map and the reward map are compared and used to update the model.

Autonomous machine navigation techniques may generate a three-dimensional point cloud that represents at least a work region based on feature data and matching data. Pose data associated with points of the three-dimensional point cloud may be generated that represents poses of an autonomous machine. A boundary may be determined using the pose data for subsequent navigation of the autonomous machine in the work region. Non-vision-based sensor data may be used to determine a pose. The pose may be updated based on the vision-based pose data. The autonomous machine may be navigated within the boundary of the work region based on the updated pose. The three-dimensional point cloud may be generated based on data captured during a touring phase. Boundaries may be generated based on data captured during a mapping phase.

A method for processing images of an agricultural field is disclosed that enables accurate phenotype measurements for crops planted in each research plot of the agricultural field. The method comprises receiving a plurality of input images of the agricultural field, calculating and refining object space coordinates for matched key points in the input images and an object space camera pose for each input image, calculating and refining object space center points for the research plots based on a user-defined plot layout, and generating output images of individual research plots that are centered, cropped, orthorectified, and oriented in alignment with planted rows of crops. Based on the output images, accurate phenotype measurements for crops planted in each research plot can be determined. The method advantageously minimizes row-offset errors, variations in canopy cover and color between images, geometric and radiometric distortion, and computational memory requirements, while facilitating parallelized image processing and analysis.