Object Velocity Detection by Fusing Image and Range Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in accurately detecting the position and velocity of objects in their environment due to limitations in existing sensor data processing, particularly with range data being sparse and unable to detect objects at large distances, and image data being prone to false positives/negatives.
Innovation Solution
A computer-implemented method that combines image data and range data to synchronize sensor data, determining state data and flow data to accurately calculate the position and velocity of objects by identifying corresponding image features and using range data to enhance radial motion detection, while image data enhances tangential motion detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If range data is used to detect objects, then radial motion detection is improved, but detection capability at large distances deteriorates due to sparse data
Solution Approach 1:
The patent combines image data from cameras with range data from LIDAR or other depth sensors to create a fused representation of the scene. This merging allows the system to leverage the complementary strengths of each sensor type: image data provides detailed visual information for tangential motion detection, while range data provides depth information for radial motion detection, thereby overcoming the limitations of each individual sensor at large distances.
Solution Approach 2:
The system employs multiple sensing modalities (image sensors and range sensors) that can serve multiple functions. The image data not only provides visual classification information but also contributes to motion detection through optical flow analysis. Similarly, range data serves both for depth estimation and radial velocity measurement, creating a multi-functional sensing system that improves overall detection reliability.
2Measurement precision
If image data is used to detect objects, then tangential motion detection is improved, but false positives and negatives increase
Solution Approach 1:
The system uses a feedback mechanism where detection results from one sensor type are used to validate and refine detections from the other sensor type. The fused sensor system cross-validates object detections, motion measurements, and velocity estimates, allowing the system to filter out false positives and confirm true detections through inter-sensor consistency checks.
Solution Approach 2:
The patent creates a composite sensing system that integrates data from heterogeneous sensor types (image sensors and range sensors). This composite approach combines the visual information richness of image data with the depth accuracy of range data, producing a more reliable and accurate object detection and velocity measurement system than either sensor type could achieve alone.
3Device complexity
If only image data is used, then system complexity is reduced, but ability to detect radial motion deteriorates
Solution Approach 1:
The system segments the velocity measurement task into two orthogonal components: tangential motion detected by image data through optical flow, and radial motion detected by range data through depth changes. This segmentation allows each sensor type to specialize in detecting specific motion components, with the fusion algorithm combining these measurements to compute full 3D velocity vectors.
4Device complexity
If only range data is used, then system complexity is reduced, but ability to detect objects at large distances deteriorates
Solution Approach 1:
The patent merges image data and range data into a unified detection framework. At large distances, where range data becomes sparse, the system compensates by utilizing the visual information from image sensors to maintain object detection and tracking, while the range data provides available depth constraints to reduce ambiguity in the fused estimation.
Data Source
AI summary
Systems and methods for movement detection are provided. In one example embodiment, a computer-implemented method includes obtaining image data and range data representing a scene external to an autonomous vehicle, the image data including at least a first image and a second image that depict the scene. The method includes identifying a set of corresponding image features from the image data, the set of corresponding image features including a first feature in the first image having a correspondence with a second feature in the second image. The method includes determining a respective distance for each of the first feature and the second feature based at least in part on the range data. The method includes determining a velocity associated with a portion of a scene represented by the set of corresponding image features based at least in part on the respective distance for the first feature and the second feature.