Message Passing Graph Neural Network for Multi-Object Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Efficient and accurate multi-object tracking across different frames of lidar data or images remains a challenging problem for autonomous vehicles, which is crucial for safe navigation and operation.
Innovation Solution
A message passing graph system is employed to track multiple objects by constructing layers of feature nodes and edges across a sequence of images, allowing information to be shared in both forward and backward directions, utilizing neural networks to establish associations and predict trajectories.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multi-object tracking is implemented across different frames of sensor data, then object detection accuracy improves, but computational complexity and processing time increase significantly
Solution Approach 1:
The tracking problem is segmented into independent detection tasks for each frame, where each frame's detections are processed separately. The graph neural network then segments the temporal relationships by creating edges only between relevant detections across frames, rather than computing all possible associations. This reduces computational complexity while maintaining tracking accuracy.
Solution Approach 2:
The system dynamically adjusts the graph structure and message passing operations based on the specific tracking scenario. The graph neural network adaptively determines which edges to create between detections in different frames, optimizing computational resources for the most relevant associations rather than processing all possible pairs statically.
2Measurement precision
If multi-object tracking is implemented across different frames of sensor data, then object detection accuracy improves, but processing speed decreases
Solution Approach 1:
Detections are pre-processed within each frame independently before temporal association. The graph neural network performs preliminary message passing operations to establish local associations, which are then combined to form complete tracking trajectories. This preliminary structuring accelerates the final trajectory computation.
Solution Approach 2:
The message passing operations continue iteratively across the graph structure, allowing information to flow continuously through multiple hops between detections. This continuous refinement of associations improves tracking accuracy while the parallel nature of graph operations maintains processing efficiency.
3Measurement precision
If information is shared across time through message passing, then tracking accuracy improves, but information loss from past frames increases
Solution Approach 1:
The graph neural network implements bidirectional message passing, allowing information to flow both forward and backward through time in the graph structure. This feedback mechanism enables later detections to inform earlier associations, correcting potential information loss and improving overall tracking accuracy by leveraging temporal context from all directions.
Data Source
AI summary
A sequence of images generated at respective times by one or more sensors configured to sense an environment through which objects are moving relative to the one or more sensors is received. A message passing graph having a multiplicity of layers associated with the sequence of images is constructed. A neural network supported by the message passing graph is trained. The training includes performing a pass through the message passing graph in a forward direction including by adding a new feature node based on a feature detection and a new edge node and performing a pass through the message passing graph in a backward direction, including by updating at least one edge node of the message passing graph. Multiple features are tracked through the sequence of images, including passing messages through the message passing graph.