Pedestrian Speed Estimation from Partial LIDAR Point Clouds
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for estimating the speed of pedestrians using LIDAR in autonomous driving vehicles are prone to errors due to the constantly changing body posture of pedestrians, leading to less than ideal minimum bounding boxes and subsequent speed estimation inaccuracies.
Innovation Solution
A method that filters LIDAR points to exclude limb points, projects the remaining points to a horizontal plane, determines a principal axis, and generates a minimum bounding box excluding limb points, allowing for more accurate speed estimation by focusing on the torso and head, which are more stable during movement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If minimum bounding box is constructed for all LIDAR points corresponding to the pedestrian including limbs, then the speed estimation can be obtained, but the accuracy deteriorates due to constantly changing body posture of the pedestrian
Solution Approach 1:
The patent segments the pedestrian's LIDAR points into different body parts (torso, head, limbs) and selectively processes only the stable parts (torso and head) for speed estimation. This segmentation allows the system to exclude limb points that cause bounding box instability while retaining points from body parts that maintain consistent positions, thereby resolving the contradiction between obtaining speed estimates and maintaining bounding box stability.
Solution Approach 2:
The patent applies different processing quality to different parts of the pedestrian's body. Specifically, it applies high-quality filtering to exclude limb points while including torso and head points, creating a localized quality difference in point selection. This local quality approach ensures that only the most reliable points (from stable body parts) contribute to the bounding box, improving both reliability and measurement precision.
2Reliability
If all LIDAR points corresponding to the pedestrian are used for minimum bounding box construction, then complete pedestrian coverage is achieved, but speed estimation accuracy deteriorates due to limb movement variability
Solution Approach 1:
The patent segments the pedestrian detection process into two stages: first, all LIDAR points are detected to ensure complete pedestrian coverage; second, the points are filtered to exclude limbs and retain only torso and head points for speed estimation. This segmentation allows the system to maintain complete detection while improving measurement precision by using only stable body parts for the actual speed calculation.
Solution Approach 2:
The patent extracts and excludes the problematic limb points from the set of all pedestrian points after initial detection. By taking out the limb points that cause variability, the system maintains complete pedestrian detection (all points are initially processed) while improving speed estimation accuracy (only stable torso and head points are used for the final calculation).
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
This approach provides more accurate speed estimates of pedestrians, enabling better trajectory planning and control for autonomous driving vehicles by reducing errors associated with limb movement.
Implementation Method 1
Individual points in the point cloud can be determined by transmitting a laser pulse and detecting a returning pulse, if any, reflected from the object, and determining the distance to the object according to the time delay between the transmitted pulse and the reception of the reflected pulse
Implementation Method 2
determining the distance to the object according to the time delay between the transmitted pulse and the reception of the reflected pulse
Data Source
Figure 1
Figure 2
Figure 3A
AI summary
A method, apparatus, and system for estimating a moving speed of a detected pedestrian at an autonomous driving vehicle (ADV) is disclosed. A pedestrian is detected in a plurality of frames of point clouds generated by a LIDAR device installed at an autonomous driving vehicle (ADV). In each of at least two of the plurality of frames of point clouds, a minimum bounding box enclosing points corresponding to the pedestrian excluding points corresponding to limbs of the pedestrian is generated. A moving speed of the pedestrian is estimated based at least in part on the minimum bounding boxes across the at least two of the plurality of frames of point clouds. A trajectory for the ADV is planned based at least on the moving speed of the pedestrian. Thereafter, control signals are generated to drive the ADV based on the planned trajectory.