Multi-Sensor Fusion Ghost-Free 3D Target Association
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current multi-sensor systems face challenges in achieving accurate and precise target location, tracking, and imaging in multiple-target settings due to high performance sensor-to-sensor association requirements, particularly in autonomous vehicle applications where ghost targets can lead to inaccurate navigation.
Innovation Solution
A system that includes a computing device capable of receiving and processing multiple pairs of sensor signals from diverse sensors, identifying valid pairs, and generating a representation of targets by filtering out non-admissible pairs and ghosts, using a dictionary and threshold-based filters to associate valid 4-D spatial addresses with voxel centers, thereby enabling low-ghost or ghost-free imaging.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple geographically diverse sensors are used for imaging, then measurement precision and target discrimination capability are improved, but device complexity and computational requirements increase
Solution Approach 1:
The patent segments the complex multi-sensor fusion problem into distinct processing stages: (1) receiving sensor signals from multiple geographically diverse sensors, (2) generating 4-D spatial addresses from sensor coordinates, (3) filtering addresses through a pre-stored dictionary of admissible addresses, and (4) generating voxel representations. This segmentation allows each stage to be optimized independently, reducing overall system complexity while maintaining high measurement precision through coordinated multi-sensor operation.
Solution Approach 2:
The patent applies preliminary action by pre-computing and storing dictionaries of admissible 4-D spatial addresses before actual target detection occurs. These dictionaries contain all possible valid coordinate combinations from multiple sensors, allowing rapid filtering during operation without real-time complex calculations. This pre-computation significantly reduces computational burden during actual target location while maintaining high precision through accurate address validation.
2Measurement precision
If high performance sensor-to-sensor association is achieved, then target discrimination capability is improved, but processing time and computational load increase
Solution Approach 1:
The patent pre-computes and stores dictionaries containing all admissible 4-D spatial addresses from multiple sensors before actual target detection. During operation, the system simply checks whether detected coordinate combinations exist in these pre-stored dictionaries, transforming a complex real-time association problem into a simple lookup operation. This eliminates time-consuming real-time calculations while maintaining high target discrimination capability through accurate address validation.
Solution Approach 2:
The patent creates simplified representations (copies) of the complex multi-sensor association problem by storing dictionaries of admissible addresses that capture all valid coordinate relationships. Instead of performing complex geometric calculations during target detection, the system uses these address copies for rapid validation, significantly reducing processing time while preserving the ability to discriminate between true targets and false associations.
3Reliability
If ghost targets are eliminated through filtering, then reliability of target detection is improved, but false alarm rate may increase if filtering is too aggressive
Solution Approach 1:
The patent implements feedback through a two-stage filtering process that provides multiple opportunities to validate target detections. First, 4-D spatial addresses are checked against the admissible address dictionary to eliminate obvious ghosts. Second, voxel representations are generated and can be further validated against expected target characteristics. This multi-stage feedback mechanism reliably eliminates ghosts while maintaining sensitivity to true targets, as each stage only filters out addresses that fail specific validation criteria rather than applying aggressive uniform filtering.
Solution Approach 2:
The patent applies local quality by treating different spatial regions and address types with different validation criteria. The admissible address dictionary contains region-specific valid coordinate combinations, allowing the system to apply appropriate filtering thresholds for different geographic areas and sensor configurations. This localized approach ensures reliable ghost elimination in each region while minimizing false alarms that might result from overly aggressive global filtering.
4Device complexity
If low-cost Boolean radar systems are used, then device complexity is reduced, but imaging quality and ghost suppression capability deteriorate
Solution Approach 1:
The patent merges multiple low-cost Boolean radar systems into a coordinated multi-sensor network, where each radar operates independently but their results are fused through the 4-D spatial address matching process. By combining the detection capabilities of multiple simple radars and validating their coordinate associations against pre-computed admissible addresses, the system achieves imaging quality and ghost suppression comparable to complex single-sensor systems while maintaining the simplicity and low cost of individual Boolean radars.
Data Source
AI summary
A system performs operations including receiving pairs of sensor signals indicative of imaging of an environment, each pair of sensor signals including (i) a first sensor signal received from a first sensor and including first spatial coordinates and a first signal characteristic and (ii) a second sensor signal including second spatial coordinates and a second signal characteristic. The operations include identifying valid pairs of sensor signals, including determining that an address including the first coordinates of the first sensor signal of a given pair and the second coordinates of the second sensor signal of the given pair corresponds to an admissible address in a set of admissible addresses stored in a data storage. The operations include identifying, from among the valid pairs of sensor signals, pairs of sensor signals that satisfy a threshold, including determining that a value based on a combination of the first signal characteristic of a given pair and the second signal characteristic of the given pair satisfies a threshold value. The operations include generating a representation of a target in the environment based on the identified pairs of sensor signals that satisfy the threshold.