Infrared Camera Extrinsic Calibration for All-Weather V2I
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing V2I systems face challenges in extrinsic calibration, particularly with infrared cameras, due to weather conditions and the inability to discern common road markings, leading to poor performance in all-weather conditions and night-time scenarios, and the buildup of noise in images over time.
Innovation Solution
A method and system for extrinsic calibration using an infrared camera that involves identifying points on the road, measuring distances, determining a road model, and calculating a transformation between camera and world coordinate systems to accurately determine object distances, optimized for infrared cameras mounted on infrastructure.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional sensors (CMOS cameras, radar, LIDAR) are used in V2I systems, then the systems can capture visual information under fair weather conditions, but they fail to perform adequately under foul weather, night conditions, or when exposed to glare and reflections
Solution Approach 1:
The patent changes the operating wavelength parameter of the camera from visible spectrum (CMOS) to infrared spectrum. This parameter change allows the sensor to operate independently of visible light conditions, making it immune to glare, reflections, and darkness, while maintaining reliability under all-weather conditions including fog, rain, and snow.
2Reliability
If video cameras with active lighting are used at night, then object detection is possible, but electrical energy consumption increases significantly
Solution Approach 1:
The infrared camera detects thermal radiation emitted naturally by objects (vehicles, pedestrians) without requiring external active lighting. The system uses the inherent thermal energy of the targets themselves, eliminating the need for additional power-consuming light sources while maintaining detection capability during nighttime.
3Measurement precision
If LIDAR is used to provide depth perception and 3D imaging, then distance measurement capability is improved, but the sensing range is limited and performance is obfuscated in rainy or foggy conditions
Solution Approach 1:
The patent replaces the active illumination-based LIDAR system with a passive infrared detection system. Instead of emitting laser pulses and measuring reflections (mechanical/optical active system), the infrared camera passively detects thermal radiation emitted by objects, providing depth perception and distance measurement capabilities that are immune to rain and fog interference.
4Reliability
If infrared cameras are used for all-weather detection, then immunity to light sources and weather factors is achieved, but the ability to discern common road markings and textures visible in the visible spectrum is lost
Solution Approach 1:
The system employs multiple sensor types working in parallel: infrared cameras for all-weather detection and positioning, and visible spectrum cameras for capturing road markings and textures. This multi-functional approach allows the system to maintain reliability under all conditions while preserving access to visual information about road markings when visible.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enhances object detection, tracking, and distance determination in all-weather conditions without active lighting, providing accurate distance measurements to vehicles and pedestrians up to 250 meters, invariant to other light sources and weather factors.
Implementation Method 1
The use of infrared camera allows for detection of, and the estimating of distance to, moving objects, such as vehicles and pedestrians
Data Source
AI summary
A method for calibrating a vehicle to infrastructure system using an infrared camera overlooking a road, comprises: obtaining from the camera a first image, each image of the camera having points identifiable in an image coordinate system and corresponding to objects in a real world scene; identifying in the first image a plurality of points each of which corresponds to a respective object that is touching the road's surface; obtaining a measurement that corresponds to a distance from a location of the camera to each respective one of the points in a camera coordinate system (CCS); determining a model of the road in the CCS based on the points and their obtained distances; and calculating a transformation between the CCS and a world coordinate system (WCS) using the road model, the transformation being usable to determine a distance in the WCS to an object in a second image.