Event-Based TTC Calculation for High-Speed Vehicle Collision Warning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for calculating Time to Collision (TTC) in autonomous driving are inefficient and inaccurate, particularly at high speeds, due to limitations in image acquisition speed, distance and speed calculation accuracy, and resource-intensive image processing, which delays response times in forward collision warnings.
Innovation Solution
A method using a Dynamic Vision Sensor (DVS) to generate a timestamp matrix from event data, scanning events in a predetermined direction to calculate time gradients, and determining spatial positions to compute TTC, thereby accelerating and improving the accuracy of TTC calculations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If conventional camera is used to acquire images for TTC calculation, then the system structure is simple, but the image acquisition speed is limited and response time is long
Solution Approach 1:
The patent transitions from conventional frame-based camera imaging to event-based DVS imaging, fundamentally changing the temporal sampling parameter from fixed frame rates to asynchronous event-triggered sampling. This enables the system to capture only dynamic changes at microsecond-level precision, achieving ultra-high speed acquisition without proportionally increasing system complexity.
2Measurement precision
If rectangular box method is used to calculate distance and speed, then the calculation process is simple, but the accuracy is insufficient
Solution Approach 1:
The patent extracts only the essential dynamic information (event coordinates and timestamps) from the complex image data, discarding redundant static background information. By focusing solely on pixel changes that indicate object movement, the system achieves high measurement precision while maintaining computational efficiency through selective information extraction.
3Productivity
If conventional image processing method is used, then comprehensive information is processed, but large quantity of resources are occupied and response speed is affected
Solution Approach 1:
The patent extracts only the essential dynamic information (event coordinates and timestamps) from the complex image data, discarding redundant static background information. By focusing solely on pixel changes that indicate object movement, the system achieves high measurement precision while maintaining computational efficiency through selective information extraction.
Solution Approach 2:
Instead of processing complete images, the system processes only the partial information necessary for TTC calculation - specifically the event data representing dynamic changes. This partial processing approach significantly reduces computational resource consumption while maintaining the accuracy needed for collision warning.
Data Source
AI summary
The present disclosure provides a method for calculating a TTC for an object and a vehicle for a calculation device, including: generating a timestamp matrix in accordance with a series of event data from a DVS coupled to the vehicle; scanning events in the timestamp matrix in a predetermined scanning direction, so as to calculate a time gradient of each event in the predetermined scanning direction; creating at least one subset consisting of a first quantity of consecutive events in the predetermined scanning direction and each with a positive time gradient; calculating a spatial position of each event in each subset in accordance with intrinsic and extrinsic parameter matrices of the DVS; and calculating the TTC in accordance with the spatial position and the timestamp of each event in the subset. The present disclosure further provides the calculation device and a corresponding vehicle.