Centroid Prediction Using Segmented Sensor Data and Local Semantics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing centroid prediction methods in sensor data, such as those used in autonomous vehicles, are inaccurate due to the inclusion of extraneous data points within bounding boxes, which affects subsequent operations like tracking, object recognition, and planning.
Innovation Solution
Utilizing a segmentation network to segment and mask out non-target data points from sensor data, followed by a centroid network that leverages semantic and local information to accurately determine the centroid of the target object.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional centroid prediction methods are used on raw sensor data within bounding boxes, then the processing is simple and fast, but the accuracy is low due to inclusion of extraneous data points
Solution Approach 1:
The patent applies segmentation by dividing the sensor data processing into distinct stages: first segmenting the bounding box data to identify target versus background regions, then separately processing target data points through the centroid network. This segmentation allows the system to exclude extraneous background points while maintaining processing efficiency through modular architecture.
Solution Approach 2:
The patent introduces an intermediary segmentation network that acts as a mediator between the raw sensor data and the centroid prediction network. This intermediary processes the bounding box data to identify and filter target-specific data points, thereby improving centroid accuracy without requiring the centroid network itself to be overly complex.
2Measurement precision
If segmentation network is introduced to filter data points before centroid prediction, then the accuracy improves by excluding extraneous points, but the processing time and computational load increase
Solution Approach 1:
The patent applies preliminary action by performing segmentation and filtering of data points before the centroid prediction step. The segmentation network pre-processes the bounding box data to identify target-specific points in advance, so that the subsequent centroid network receives already-filtered data, reducing its computational burden and overall processing time.
Solution Approach 2:
The patent extracts only the relevant target data points from the bounding box by using the segmentation network to identify and separate target points from background points. This extraction process removes extraneous data before centroid calculation, improving accuracy while the efficient extraction algorithm minimizes the time penalty.
Data Source
AI summary
The present disclosure generally relates to improved centroid predictions. In some aspects, a method of the disclosed technology includes: collecting, from a sensor, sensor data comprising data points; segmenting, via a first network, the sensor data into a first portion of the sensor data and a second portion of the sensor data; determining a semantic label for the first portion of the sensor data; determining local semantic information for each data point of the first portion of the sensor data; removing, using a point mask, the second portion of the sensor data; and determining, via a second network, a centroid of the first portion of the sensor data based on the first portion of the sensor data, the semantic label, and the local semantic information. Systems and machine-readable media are also provided.