LIDAR Pedestrian Detection via 3D Point Cloud Ground Separation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Pedestrian detection in adverse weather conditions, such as heavy rain or fog, is challenging for LIDAR systems due to reduced precision and increased costs of infrared detection systems, which are often large and power-intensive, limiting their widespread use for automotive and street monitoring applications.
Innovation Solution
A method combining LIDAR with 3D point cloud data, object extraction processing, and 3D and 2D feature extraction, utilizing Random Sample Consensus (RANSAC) for ground separation, object grouping, and machine learning for pedestrian classification, enhances pedestrian detection accuracy without significant hardware upgrades.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If infrared detection system is used for pedestrian detection in nighttime, then detection capability is improved, but device complexity, size and power consumption increase
Solution Approach 1:
The patent replaces the mechanical infrared detection system with an optical LIDAR system that uses laser beams and photosensitive components. This substitution maintains pedestrian detection capability while significantly reducing device complexity, size, and power consumption by using optical components instead of complex infrared machinery
Solution Approach 2:
The patent creates a 3D point cloud chart that copies and represents the spatial distribution of pedestrians and environmental features. This digital model enables detection and analysis without requiring complex physical detection hardware, thereby reducing device complexity while maintaining detection reliability
2Measurement precision
If LIDAR system is used for pedestrian detection, then detection speed and accuracy are improved, but precision deteriorates in heavy rain conditions
Solution Approach 1:
The patent extracts and separates ground information from the 3D point cloud data using ground separation processing. By removing the ground plane and irrelevant background elements, the system focuses computational resources on detecting pedestrians, thereby maintaining detection accuracy even when rain interferes with the raw LIDAR data
Solution Approach 2:
The patent performs preliminary ground separation and object extraction processing before final pedestrian detection. This preliminary processing prepares the data by removing interfering elements (ground, non-pedestrian objects) in advance, so that the detection algorithm can focus on relevant targets even when heavy rain degrades raw LIDAR precision
3Length of moving object
If LIDAR system detects farthest distance, then detection range is extended, but precision linearly declines
Solution Approach 1:
The patent transforms the 3D point cloud data into a 2D point cloud chart through dimensionality reduction. This transformation consolidates spatial information from multiple distances into a 2D representation, allowing the system to maintain precision across extended detection ranges by analyzing spatial patterns in the 2D projection rather than relying on distant 3D point density alone
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
Improves pedestrian detection precision in adverse weather conditions, extending detection ranges and enhancing safety without increasing hardware costs, enabling more effective use in self-driving systems and industrial applications.
Implementation Method 1
Light detection and ranging (LIDAR) is an optical remote sensing technology. A LIDAR system may be divided into three parts, including a laser emitter, scanning and optical components, and photosensitive components, where the laser emitter may emit laser beam with a wavelength in the range of 600 nm-1000 nm
Implementation Method 2
the laser emitter may emit laser beam with a wavelength in the range of 600 nm-1000 nm
Implementation Method 3
the photosensitive components are arranged to detect intensity of reflected light
Data Source
AI summary
A method for performing pedestrian detection with aid of light detection and ranging (LIDAR) is provided. The method includes: obtaining 3-dimensional (3D) point cloud data through the LIDAR; performing ground separation processing on the 3D point cloud data to remove ground information; performing object extraction processing on the 3D point cloud data to obtain 3D point cloud chart that includes pedestrian candidate point cloud group; performing 2-dimensional (2D) mapping processing on the 3D point cloud chart to obtain 2D chart; and extracting 3D feature and 2D feature from the 3D point cloud chart and the 2D chart, respectively, and utilizing the 3D feature and the 2D feature to determine location of the pedestrian. According to the method, image data obtained by the LIDAR may be enhanced, the method may distinguish between pedestrian far away and environment blocks, and pedestrian recognition in nighttime or in bad weather may be improved.