LiDAR Vehicle Masking for Accurate Near-Vehicle Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
LiDAR sensors in autonomous vehicles often produce noise points that interfere with accurate object detection due to reflections from the vehicle itself, making it difficult to distinguish actual objects from noise in the surrounding environment.
Innovation Solution
A vehicle object detection system that generates vehicle masking data by identifying and excluding the vehicle's area from LiDAR data using a reference point and virtual ground coordinates, transforming sensor data to remove noise and enhance object detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If LiDAR is installed to cover host vehicle area to detect objects close to the vehicle, then detection coverage is improved, but noise points from vehicle reflections increase
Solution Approach 1:
The patent segments the LiDAR detection space into two distinct regions: a first space containing the host vehicle and a second space excluding the host vehicle. By generating separate masking data for each space, the system can selectively process points from different regions differently, allowing comprehensive detection coverage while filtering out vehicle reflection noise through spatial segmentation.
2Object-generated harmful factors
If laser pulses are reflected by the vehicle and points are recognized as noise, then noise filtering is achieved, but actual object points may be mistakenly removed
Solution Approach 1:
The patent introduces a masking data generation unit that creates virtual masking data representing the host vehicle's spatial occupancy. This masking data acts as an intermediary tool that selectively filters points based on their spatial relationship with the vehicle. By using this intermediary masking mechanism rather than direct noise recognition, the system accurately distinguishes between actual objects and vehicle reflections, preventing false removal of valid object points while effectively filtering noise.
3Measurement precision
If vehicle masking data is generated to remove vehicle area, then noise removal precision is improved, but processing complexity increases
Solution Approach 1:
The patent performs preliminary generation of masking data that represents the host vehicle's spatial occupancy before processing the actual LiDAR detection data. By pre-computing the vehicle's spatial mask and storing it for subsequent operations, the system eliminates the need for complex real-time calculations during object detection. This preliminary action simplifies the overall processing complexity while maintaining high noise removal precision through accurate spatial masking.
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 reliability of autonomous driving by accurately distinguishing between vehicle reflections and actual objects, improving the precision of object detection and reducing noise interference.
Implementation Method 1
LiDAR sensor that emits laser pulses to detect and measure the environment
Implementation Method 2
laser pulses reflected by the vehicle
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
An apparatus for controlling autonomous driving of a vehicle is introduced. The apparatus may comprise a sensor mounted at a reference point on the vehicle, configured to acquire sensing data. The apparatus may also comprise processors and a memory storing programs that, when executed, configured to cause the apparatus to generate a first cluster of points based on the reference point and sensing data, create a first range image based on the first cluster of points, produce a second range image based on a distance value and a virtual reference point on the ground, extract first and second areas from the range images, determine ground contact points, generate a third range image incorporating these areas, determine a third area with an object in the third range image, generate vehicle masking data based on the third area; and control autonomous vehicle driving based on a signal indicating the vehicle masking data.