Sensor Fusion Object Tracking Under LIDAR Exhaust Interference
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object recognition systems for autonomous vehicles face challenges in accurately determining the position of objects due to errors caused by exhaust gas, seasonal variations, and temperature changes, which can lead to inaccurate tracking and increased risk of accidents.
Innovation Solution
An object recognition apparatus and method that utilizes a combination of LIDAR, camera, and radar sensors to generate a synthetic fusion track. This involves identifying LIDAR tracks and fusion tracks based on specific conditions such as distribution shape, width, ratio of track widths, and class information to reduce errors caused by exhaust gas and environmental variations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Difficulty of detecting and measuring
If LIDAR is used to detect object position, then measurement capability is improved, but measurement precision deteriorates due to exhaust gas interference
Solution Approach 1:
The patent combines LIDAR, camera, and radar sensors to create a fusion tracking system. The LIDAR provides 3D spatial information, the camera provides visual recognition and class information, and the radar provides velocity data. By merging these multiple sensing modalities, the system compensates for LIDAR's vulnerability to exhaust gas interference while maintaining its detection capability.
Solution Approach 2:
The patent introduces an intermediary validation mechanism where camera-based object recognition serves as a mediator to verify LIDAR-detected objects. When the camera identifies an object at the same location as LIDAR, it confirms the LIDAR detection is valid and not caused by exhaust gas interference, thus improving measurement precision.
2Device complexity
If single sensor tracking is used, then device complexity is reduced, but reliability deteriorates due to environmental errors
Solution Approach 1:
The patent merges LIDAR, camera, and radar into a unified fusion tracking system. Each sensor type provides complementary information that compensates for the weaknesses of others under environmental variations. The LIDAR provides precise 3D positioning, the camera provides environmental context and object classification, and the radar provides velocity information, together achieving reliable tracking despite environmental challenges.
Solution Approach 2:
The patent changes the operational parameters of the sensing system by using multiple sensors with different physical measurement principles. Instead of relying on a single sensor's parameters, the system utilizes diverse parameters (optical, electromagnetic, visual) that are less susceptible to the same environmental errors, thereby improving reliability under varying conditions.
3Productivity
If LIDAR track is used alone, then tracking speed is improved, but measurement precision deteriorates due to exhaust gas and seasonal variations
Solution Approach 1:
The patent merges LIDAR's fast tracking capability with camera and radar validation. The LIDAR continues to provide rapid 3D tracking, while the camera and radar provide parallel validation streams that filter out false detections caused by exhaust gas and seasonal variations, maintaining both tracking speed and precision.
Solution Approach 2:
The patent implements feedback validation where camera and radar data continuously verify LIDAR tracking results. When environmental factors cause LIDAR errors, the feedback from other sensors detects the discrepancy and corrects the tracking, maintaining measurement precision without sacrificing LIDAR's tracking speed.
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
The proposed solution improves the accuracy of object position tracking by reducing errors due to exhaust gas and environmental factors, thereby enhancing the safety and convenience of vehicle operation by minimizing the risk of accidents.
Implementation Method 1
a light detection and ranging (LIDAR) device
Implementation Method 2
a radar
Data Source
AI summary
An object recognition apparatus includes a LIDAR, a camera, a radar, and a processor. The processor may identify a LIDAR track, identify a fusion track, and generate a synthetic fusion track including at least one of a longitudinal position, a lateral position, a width, a length, or a heading represented by the fusion track and corresponding to the object based on determining that at least one of the following conditions is satisfied: a distribution shape of LIDAR points included in the LIDAR track satisfies a distribution condition, a width of the fusion track satisfies a width condition, a ratio of a width of the LIDAR track to the width of the fusion track satisfies a ratio condition, or class information of the object according to the fusion track satisfies a class condition.