LiDAR Object Recognition Using Multi-Surface Grid Map Projection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing LiDAR sensors often incorrectly identify external objects, leading to errors in vehicle control systems, particularly when objects are misclassified as static when they are in motion, causing rapid changes in driving routes.
Innovation Solution
A processor projects LiDAR points onto grid maps using axes parallel and perpendicular to the vehicle's direction to accurately identify object types by comparing these maps with reference grid maps stored in memory, determining matching patterns to classify objects as moving or static, and allocating scores based on lane position.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Difficulty of detecting and measuring
If LiDAR sensors are used to identify external objects, then object detection capability is improved, but object recognition accuracy deteriorates due to incorrect identification and separation of objects
Solution Approach 1:
The patent projects 3D point cloud data from LiDAR onto multiple 2D planes (front, rear, left, right surfaces) to create grid maps. This dimensional transformation allows the system to analyze object characteristics from multiple perspectives simultaneously, improving recognition accuracy by comparing projected points against reference grid maps for each surface type.
Solution Approach 2:
The patent divides the external space into multiple discrete surfaces (front, rear, left, right) and creates separate grid maps for each. By segmenting the recognition task into multiple surface-specific analyses, the system can more accurately identify objects on different sides of the host vehicle and reduce misclassification errors.
2Measurement precision
If grid map projection onto multiple surfaces is performed, then object recognition accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent applies different analysis methods to different surfaces based on their local characteristics. For example, the front surface grid map is analyzed to detect objects in the driving path, while side surfaces are analyzed for lane change decisions. This localized approach allows the system to optimize computational resources by focusing detailed analysis only where needed rather than uniformly processing all surfaces.
Solution Approach 2:
The system projects points onto all four surfaces (front, rear, left, right) to ensure comprehensive coverage, but in practice, not all projections are equally critical for every driving situation. The patent performs the complete set of projections to guarantee safety, accepting the computational cost as a necessary measure for robust object recognition in all scenarios.
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 object recognition accuracy, reducing errors in vehicle control by correctly classifying external objects as moving or static, thereby stabilizing driving routes.
Implementation Method 1
an external object may be identified using a sensor (e.g., a light detection and ranging (LiDAR) sensor)
Data Source
AI summary
An apparatus for recognizing an object and a method thereof are provided. The apparatus may comprise a sensor, a memory configured to store reference grid maps representing an appearance of a vehicle, and a processor configured to project a plurality of points indicating an external object, obtained by the sensor, onto a first surface to obtain a first grid map including a first plurality of grids, project the plurality of points indicating the external object onto a second surface to obtain a second grid map including a second plurality of grids, determine, based on at least one of the reference grid maps, the first grid map, or the second grid map, a type of the external object, and output a signal indicating the type of the external object.