Cooperative LIDAR Feature Sharing for Occlusion-Aware Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object detection systems in computer vision struggle with non-line-of-sight and partial occlusion issues, particularly in single-vehicle setups, and conventional cooperative methods face challenges like increased communication costs and lack of consensus among vehicles.
Innovation Solution
A method involving cooperative vehicles that align point-clouds with a global coordinate system, project them onto a 2D/3D image plane, and use Translation Mod Alignment to pad or shift feature maps for accurate feature sharing, followed by a CNN-based detection process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If raw sensed data is communicated from other observers to improve object detection performance, then detection accuracy improves, but communication cost increases significantly
Solution Approach 1:
The patent extracts only the essential feature representations from raw sensed data using CNN feature extraction networks. Instead of communicating complete point cloud data or raw images, the system extracts and transmits only the relevant feature maps and detection results, significantly reducing communication overhead while maintaining detection accuracy.
Solution Approach 2:
The patent inverts the conventional approach by having vehicles transmit detection results and feature representations rather than raw sensed data. This inversion transforms the communication paradigm from data-intensive raw sensor sharing to efficiency-oriented feature and result sharing, achieving the same collaborative detection benefit with minimal communication cost.
2Measurement precision
If detection results from other observers are communicated to improve object detection, then detection performance improves partially, but complete resolution of occlusion issues is not achieved
Solution Approach 1:
The patent merges detection results and feature maps from multiple cooperative vehicles through feature map fusion and result aggregation. By combining the detection outputs from multiple observers, the system achieves more complete object detection and occlusion handling than any single vehicle could achieve alone, improving both accuracy and reliability.
Solution Approach 2:
The patent introduces a centralized server as an intermediary that collects, processes, and fuses detection results from multiple vehicles. This intermediary coordinates the collaborative detection process, managing feature map fusion and result aggregation to ensure complete and accurate object detection across the cooperative vehicle network.
3Measurement precision
If feature maps are transmitted between vehicles for cooperative detection, then detection accuracy improves, but communication capacity requirements increase
Solution Approach 1:
The patent segments the feature map transmission by selectively transmitting only relevant feature channels and regions. Instead of transmitting complete high-dimensional feature maps, the system divides and transmits only the necessary feature segments, reducing communication data volume while preserving detection accuracy.
Solution Approach 2:
The patent changes the parameter representation of feature maps by encoding them in compressed formats and adjusting resolution and channel dimensions. This parameter transformation reduces the data size of transmitted feature maps while maintaining the essential information needed for accurate cooperative detection.
Data Source
AI summary
Situational awareness as a necessity in connected and autonomous vehicles domain is a subject of significant research in recent years. The driver's safety is directly dependent on robustness, reliability and scalability of such systems. Cooperative mechanisms have provided a solution to improve situational awareness by utilizing communication networks. These mechanisms mitigate problems such as occlusion and sensor range limitation. However, the network capacity is a factor determining the maximum amount of information being shared among cooperative entities. A focus of this work is to reduce the network capacity requirements while maintaining the desirable object detection performance by utilizing and modifying the concept of feature sharing. Described here is a mechanism to further improve object detection performance by utilizing novel decentralized parallel frameworks and a new shared data alignment method to allow parallel and cooperative processing of sensed data in multiple locations yielding significant improvements.