LiDAR Clustering and Radar Selection for Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current object detection methods for applications like assisted and autonomous driving are computationally expensive and require accurate, reliable results, which is challenging due to the need for high computational resources and the limitations of LiDAR sensors in long-range object detection.
Innovation Solution
A computer-implemented method that combines camera, LiDAR, and radar sensor data to efficiently detect objects by clustering LiDAR measurements, using radar to select relevant LiDAR data, and filtering hypotheses based on image and sensor data fusion, with pre-processing techniques like FFT and SAMV/IAA for radar data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional object detection methods are used to ensure accurate and reliable results, then detection reliability is improved, but computational cost increases significantly
Solution Approach 1:
The patent segments the LiDAR measurements into multiple clusters representing different spatial regions or objects. By dividing the complete LiDAR point cloud into manageable clusters, the system can process only relevant portions of data for each detection task, reducing overall computational load while maintaining detection reliability through targeted processing of significant regions.
Solution Approach 2:
The patent introduces radar measurements as an intermediary that bridges camera and LiDAR data. The radar detection serves as a mediator to select and filter LiDAR measurements, allowing the system to focus computational resources on LiDAR data points that are confirmed by radar, thereby reducing unnecessary computations while ensuring reliable detection results.
2Length of stationary object
If LiDAR sensor is used for long-range object detection, then detection range is improved, but detection accuracy deteriorates due to LiDAR limitations
Solution Approach 1:
The patent merges radar and LiDAR measurement systems to create a complementary detection approach. Radar provides robust long-range detection capability with good accuracy for distant objects, while LiDAR adds spatial detail and positioning information. By combining both systems, the patent achieves both extended detection range and maintained accuracy that neither system could achieve alone.
Solution Approach 2:
The patent uses radar measurements as an intermediary to validate and supplement LiDAR detections at long ranges. The radar detection acts as a mediator that confirms the presence and position of distant objects, allowing the system to overcome LiDAR's accuracy limitations at extended ranges by cross-validating with radar information.
3Reliability
If multiple sensor data sources are integrated to improve detection robustness, then detection reliability is improved, but system complexity increases
Solution Approach 1:
The patent introduces a radar-based selection mechanism as an intermediary layer that simplifies the integration of multiple sensors. Instead of directly fusing all sensor data which would increase complexity, the radar measurements serve as a mediator to filter and select relevant LiDAR measurements, creating a hierarchical processing approach that reduces system complexity while maintaining robustness.
Solution Approach 2:
The patent extracts and separates the essential function of long-range detection to the radar system, while using LiDAR for supplementary spatial information. By taking out the primary detection function to radar and using LiDAR only for enhanced positioning and detail, the system achieves robust detection without requiring full integration of all sensor capabilities, thereby reducing overall system complexity.
Data Source
AI summary
A computer implemented method for determining a property of an object comprises the following steps carried out by computer hardware components: acquiring an image of a scene comprising the object; acquiring a plurality of lidar measurements of the scene; clustering the plurality of lidar measurements into a plurality of groups of lidar measurements; acquiring a radar measurement of the scene; identifying which of the plurality of groups of lidar measurements corresponds to the radar measurement; and determining the property of the object based on the image and the identified group of lidar measurements.