LiDAR Object Recognition Using Contour-Box Mismatch Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles and driver assistance systems face challenges in accurately distinguishing between stationary and moving objects using LiDAR sensors, which is crucial for ensuring safety and stability in driving environments.
Innovation Solution
An object recognition apparatus and method that utilizes a processor to analyze contour points from LiDAR data, calculating contour-box distances and determining a degree of mismatch between object boxes and contour points to identify whether an object is stationary or moving, based on threshold values and reliability scores.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional LiDAR-based object detection is used, then basic object position detection is achieved, but the accuracy of distinguishing stationary and moving objects is insufficient
Solution Approach 1:
The patent segments the object detection process into distinct stages: obtaining initial detection results, extracting contour points, calculating contour-box distances, and determining mismatch degrees. This segmentation allows each stage to be optimized independently, improving overall classification accuracy while managing processing complexity through modular computation.
Solution Approach 2:
The patent introduces a new dimensional approach by calculating contour-box distances and mismatch degrees, which are novel metrics not used in traditional LiDAR detection. These additional dimensional measurements enable more precise differentiation between stationary and moving objects by analyzing the spatial relationship between contour points and bounding box boundaries.
2Measurement precision
If contour point analysis is performed to improve classification accuracy, then object classification accuracy is improved, but processing time increases
Solution Approach 1:
The patent applies partial action by selectively analyzing contour points based on their distance relationships. Rather than processing all possible parameters for every object, the system focuses on calculating contour-box distances and mismatch degrees only when necessary for classification, reducing unnecessary computational overhead while maintaining high accuracy.
Solution Approach 2:
The patent changes the parameters being analyzed from traditional object detection metrics to contour-box distance and mismatch degree parameters. This parameter transformation enables more efficient processing by focusing computational resources on the most discriminative features for stationary vs. moving object classification, thereby improving accuracy without proportionally increasing processing time.
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 the accuracy of determining whether an object is stationary or moving, enhancing safety and stability in autonomous driving by effectively classifying objects as capable or incapable of movement.
Implementation Method 1
A distance from a LIDAR to an object may be obtained through an interval between the time if laser is transmitted by the LIDAR and the time if the laser reflected by the object is received
Data Source
AI summary
The present disclosure relates to an object recognition apparatus and method, and the object recognition apparatus. The apparatus may comprise a sensor and a processor, wherein the processor is configured to identify, based on sensing information of the sensor, an object box comprising a plurality of contour points representing an object, determine a plurality of minimum values, wherein each of the plurality of minimum values is a minimum value among distances between one of the plurality of contour points and line segments constituting a boundary of the object box, determine a degree of mismatch between the object box and the plurality of contour points based on at least one of the plurality of minimum values or a number of the plurality of contour points, and output, based on the degree of mismatch, a signal indicating whether the object is a stationary object.