Vehicle Recognition Fusion for LIDAR Splash False Positives
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle automatic control systems face challenges in accurately distinguishing between vehicle splashes and actual targets during rainfall or wet road conditions, leading to potential erroneous recognition and control actions.
Innovation Solution
A recognition device that integrates LIDAR and camera data to exclude targets within a predetermined distance, detected only by LIDAR and not by the camera, with a width and detection duration below specific thresholds, and considers rainfall, road surface wetness, wiper operation, and two-wheeled vehicle presence to improve splash determination accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If image processing is performed on camera images to determine splashes, then splash detection capability is provided, but accuracy in splash determination deteriorates because heavy rain hitting the windshield may be erroneously recognized as targets
Solution Approach 1:
The patent combines LIDAR and camera detection results to perform splash determination. The LIDAR detects targets in the traveling direction while the camera captures images, and their results are merged to distinguish actual targets from splashes. This multi-sensor fusion approach resolves the contradiction by providing both splash detection capability and accurate target recognition, as splashes can be identified as LIDAR-detected targets that are not visible in camera images.
Solution Approach 2:
The patent introduces an exclusion processing mechanism that acts as an intermediary between detection and recognition. This mechanism excludes targets detected by LIDAR but not by the camera from being recognized as actual targets, thereby preventing erroneous recognition of splashes as real objects while maintaining accurate target detection capability.
2Measurement precision
If exclusion processing is performed to remove false positive targets, then target recognition accuracy is improved, but processing load increases due to multiple condition checks
Solution Approach 1:
The patent performs preliminary splash determination using LIDAR and camera data before final target recognition. By pre-identifying potential splashes through the exclusion processing based on simple criteria (LIDAR detection without camera detection), the system reduces the processing load during main recognition operations while maintaining high accuracy.
Solution Approach 2:
The exclusion processing is applied selectively based on local conditions - only to targets that meet specific criteria (detected by LIDAR within predetermined distance but not detected by camera). This localized application of exclusion rules avoids unnecessary processing of all detected targets, thereby reducing overall processing complexity while maintaining recognition accuracy.
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 the accuracy of splash determination, reducing erroneous control actions and processing load by selectively applying exclusion processing based on multiple conditions, thereby improving the reliability of vehicle control systems.
Implementation Method 1
a light detection and ranging (LIDAR)
Implementation Method 2
the position of an object in the vicinity of the vehicle is detected by using a light detection and ranging (LIDAR)
Implementation Method 3
a camera disposed so that the detection range overlaps that of the LIDAR
Data Source
AI summary
A recognition device includes an acquisition unit configured to acquire a first detection result from a LIDAR that detects a target present in a traveling direction of a vehicle and to acquire a second detection result from a camera disposed so that the detection range overlaps that of the LIDAR, and a recognition unit configured to recognize a target present around the vehicle on the basis of the first detection result and the second detection result, in which the recognition unit executes exclusion processing of excluding a target, that satisfies all of a first condition indicating that a target is present within a predetermined distance from the vehicle, a second condition indicating that a target is detected by the LIDAR and is not detected by the camera, and a third condition indicating that a width of a target is equal to or less than a predetermined value and a detection duration thereof is equal to or less than a predetermined time, from a result of recognition recognized as a target present around the vehicle.