Automatic Sensor Data Labeling for Occluded Objects
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for labeling sensor data in autonomous vehicles are either time-consuming and costly due to human intervention or require complex algorithms, and they struggle to label objects beyond the effective perceptive range of sensors or when occluded.
Innovation Solution
A method for automatically generating labels for sensor data by identifying first and second sensor data from different sensors and locations, determining if an object is static, and using the label from the first sensor data to generate a label for the second sensor data, even if the object is beyond the effective perceptive range or occluded.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If human intervention is used for labeling sensor data, then labeling accuracy can be ensured, but the process becomes time-consuming and costly
Solution Approach 1:
The system performs self-labeling by automatically generating labels for sensor data without requiring human intervention. The processor identifies objects in sensor data from multiple sensors at different times and locations, determines object stability, and automatically creates labels based on this analysis, enabling the system to serve itself rather than requiring human annotators.
Solution Approach 2:
The system performs preliminary analysis by examining sensor data from multiple time points and locations before generating labels. It pre-determines object stability and characteristics using data from various sensors, which then enables accurate labeling without requiring time-consuming human review for each individual label.
2Adaptability or versatility
If complex algorithms are used for labeling, then labeling capability is improved, but processing complexity and computational cost increase
Solution Approach 1:
The system merges data from multiple sensors (LIDAR, camera, radar) and multiple time points into a unified labeling process. By combining information from different sources and integrating them through a coordinated processing approach, the system achieves enhanced labeling capability without requiring each individual algorithm to be overly complex.
Solution Approach 2:
The processing system performs multiple functions through a single integrated framework: it processes data from different sensor types, handles objects at various distances, accounts for occlusion, and generates labels for diverse object categories. This multi-functional approach reduces overall system complexity compared to having separate specialized algorithms for each task.
3Ease of manufacture
If sensor perceptive range is limited, then sensor design is simplified, but objects beyond the range cannot be labeled
Solution Approach 1:
The system uses intermediate processing steps that bridge the gap between sensor limitations and detection requirements. By analyzing data from multiple sensors and multiple time points, and by determining object stability, the system can infer the presence and characteristics of objects that may be beyond the immediate perceptive range of individual sensors, effectively extending detection capability without changing the sensors themselves.
Solution Approach 2:
The system performs preliminary detection and stabilization analysis before final labeling. By pre-examining object characteristics across multiple time points and determining which objects remain stable, the system can identify objects that persist in the environment even when they fall outside the immediate perceptive range of individual sensors at any single moment.
4Measurement precision
If objects are occluded, then detection accuracy decreases, but labeling capability should remain
Solution Approach 1:
The system uses intermediate temporal data as a mediator to overcome occlusion problems. By examining sensor data from multiple time points, the system can determine object stability and infer the presence of occluded objects through their consistent appearance across time, even when they are temporarily hidden from direct sensor view at any single moment.
Solution Approach 2:
The system performs preliminary stability determination by analyzing object presence across multiple time points before final labeling. This preliminary analysis identifies objects that maintain consistent characteristics over time, allowing the system to label occluded objects based on their temporal persistence and inferred stability rather than relying solely on direct simultaneous detection.
Data Source
AI summary
Aspects of the disclosure provide for automatically generating labels for sensor data. For instance, first sensor data for a vehicle may be identified. This first sensor data may have been captured by a first sensor of the vehicle at a first location during a first point in time and may be associated with a first label for an object. Second sensor data for the vehicle may be identified. The second sensor data may have been captured by a second sensor of the vehicle at a second location at a second point in time outside of the first point in time. The second location is different from the first location. A determination may be made as to whether the object is a static object. Based on the determination that the object is a static object, the first label may be used to automatically generate a second label for the second sensor data.