LIDAR Object Recognition for Penetrable Obstacle Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object recognizing devices struggle to accurately distinguish between a nearby vehicle and penetrable objects like exhaust gas or water spray, leading to potential erroneous recognition and increased collision risks due to close proximity or size enlargement.
Innovation Solution
An object recognizing device utilizing LIDAR detection point information to differentiate between vehicle sizes, track vehicles, and determine size enlargement, with additional units to identify penetrable objects by setting specific distances based on vehicle speed, acceleration, and length, allowing for appropriate recognition and collision avoidance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Difficulty of detecting and measuring
If LIDAR detection is used to identify objects near a vehicle, then detection capability is improved, but erroneous recognition of penetrable objects as vehicles occurs
Solution Approach 1:
The patent segments the object recognition process into multiple independent analysis stages: initial detection by LIDAR, size measurement, speed/acceleration analysis, and penetrable object determination. Each stage processes specific parameters separately before integrating results, allowing the system to distinguish between vehicles and penetrable objects through progressive refinement rather than single-step recognition.
Solution Approach 2:
The patent introduces an intermediary analysis layer between LIDAR detection and final object classification. This intermediary process measures object size, calculates speed and acceleration from detection point changes, and uses these intermediate parameters to determine whether detected objects are penetrable (exhaust gas, spray) or solid vehicles, preventing direct erroneous classification.
2Area of stationary object
If the detection range is expanded to include nearby objects, then detection coverage is improved, but false identification of penetrable objects as obstacles increases
Solution Approach 1:
The patent applies dynamic analysis to distinguish penetrable objects from vehicles by examining temporal changes in detection parameters. The system calculates speed and acceleration based on changes in detection point positions over time, recognizing that penetrable objects exhibit different motion characteristics compared to solid vehicles, thereby maintaining high identification reliability across expanded detection ranges.
Solution Approach 2:
The patent changes the analysis parameters from static object classification to dynamic parameter measurement. Instead of immediately classifying detected objects, the system measures size, speed, and acceleration parameters that change over time, using these parameter variations to reliably distinguish between penetrable objects and vehicles even when both are within the expanded detection range.
3Measurement precision
If size measurement is used to differentiate objects, then classification accuracy is improved, but complexity of object analysis increases
Solution Approach 1:
The patent implements partial measurement by focusing on specific size parameters (length, width, height) derived from detection point distributions rather than complete object characterization. This partial action approach provides sufficient classification accuracy for distinguishing vehicles from penetrable objects without requiring exhaustive analysis of all possible object properties, thereby limiting complexity increase.
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
Effectively recognizes penetrable objects near vehicles, preventing erroneous recognition as obstacles and reducing collision risks by generating collision avoidance courses that permit contact with penetrable objects when safe, thus enhancing safety and operational efficiency.
Implementation Method 1
from the fact that a penetrable object (an irregularly shaped object) such as fog or vapor reflects only a part of light emitted from the LIDAR
Implementation Method 2
using a light detection and ranging (LIDAR)
Data Source
AI summary
An object recognizing device includes: an other-vehicle recognizing unit configured to recognize another vehicle traveling near the vehicle based on the detection point information of the LIDAR and to recognize a size and a traveling condition of the other vehicle; an other-vehicle tracking unit configured to track the other vehicle based on the detection point information of the LIDAR; and an enlargement determining unit configured to determine whether the size of the other vehicle which is tracked by the other vehicle tracking unit is enlarged, wherein, when the enlargement determining unit determines that the size of the other vehicle has been enlarged, detection points located near the other vehicle before being enlarged are recognized as an object other than the other vehicle.