Radar Tensor Heatmaps for Valid Hit Point Cloud Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional radar sensor systems generate suboptimal point clouds due to invalid radar hits not being represented by peak energies or energies below a predefined threshold, leading to inaccuracies in object detection, including false positives and missed detections.
Innovation Solution
A radar system that uses processing circuitry to construct a radar tensor, feed it into a neural network to generate a heatmap, and apply thresholding to identify valid radar hits, thereby improving the accuracy of point cloud generation and object detection by assigning labels to bins based on heatmap values exceeding thresholds.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional peak analysis and thresholding are used to generate point clouds, then processing speed is maintained, but detection accuracy deteriorates due to invalid radar hits not being represented by peak energies or energies below threshold
Solution Approach 1:
A heatmap is introduced as an intermediary representation between the radar tensor and the point cloud. The heatmap encodes detection confidence values for each bin, allowing the system to identify valid radar hits more accurately without relying solely on peak energy detection or fixed thresholding. This intermediary layer resolves the contradiction by providing additional information that improves detection accuracy while maintaining a structured processing approach.
Solution Approach 2:
The system changes from using only energy values as the primary parameter for hit detection to using heatmap-encoded confidence values. By transforming the detection criterion from simple energy thresholding to confidence-based classification, the system achieves better detection accuracy. The heatmap values are computed based on energy values but provide a more nuanced representation that reduces false positives and missed detections.
2Measurement precision
If neural network processing is added to generate heatmaps, then detection accuracy improves, but processing time and computational load increase
Solution Approach 1:
The neural network generates the heatmap in advance as a preprocessing step before point cloud generation. By computing the heatmap confidence values beforehand, the system prepares detection information that can be efficiently used during the main detection process. This preliminary action allows the system to achieve improved detection accuracy without adding significant overhead to the critical detection path.
Solution Approach 2:
The neural network creates a copied representation of the radar tensor in the form of a heatmap, which contains encoded detection confidence information. This copy allows the system to work with pre-processed confidence values rather than raw energy data, improving detection accuracy while enabling efficient processing since the neural network computation is performed once per radar frame.
3Ease of operation
If conventional thresholding is used, then processing simplicity is maintained, but false positives and missed detections increase
Solution Approach 1:
The heatmap provides feedback information in the form of confidence values that indicate the likelihood of each bin representing a valid radar hit. This feedback mechanism allows the system to make more informed detection decisions compared to simple thresholding. The confidence values are computed based on the radar tensor data and provide a reliability metric that reduces false positives and missed detections while maintaining a relatively simple processing framework.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The technologies described herein relate to a radar system that is configured to generate point clouds based upon radar tensors generated by the radar system. More specifically, the radar system is configured to generate heatmaps based upon radar tensors, wherein a neural network receives the radar tensors as input and constructs the heatmaps as output. Point clouds are generated based upon the heatmaps. A computing system detects objects in an environment of an autonomous vehicle (AV) based upon the point clouds, and the computing system further causes the AV to perform a driving maneuver based upon the detected objects.