3D Object Detection Using Luminance Clustering for Low-Reflectance Targets
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Detection of objects with low reflectance in three-dimensional point cloud data is challenging, particularly for agricultural implements like tractors, where components such as tires and windows are difficult to identify due to low reflectance.
Innovation Solution
An object detecting device and method that utilizes a combination of image sensors with both visible and infrared sensitivity, along with a clustering process to enhance the detection of low-reflectance objects by using both average and maximum luminance thresholds, and integrating 2D and 3D object information for accurate identification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If only three-dimensional point cloud data is used for object detection, then the detection process is simple, but objects with low reflectance cannot be detected
Solution Approach 1:
The patent combines three-dimensional point cloud data from a distance measurement device with two-dimensional luminance images from an image sensor. This merging of data sources allows the system to detect low-reflectance objects by cross-referencing depth information with luminance information, thereby improving detection accuracy without significantly increasing system complexity.
Solution Approach 2:
The patent introduces a luminance image as an intermediary data source that complements the point cloud data. The luminance image serves as a mediator that provides reflectance information missing from the depth data, enabling the detection of objects with low reflectance that would otherwise be invisible in point cloud data alone.
2Measurement precision
If multiple data types are processed to improve detection accuracy, then detection precision increases, but processing complexity increases
Solution Approach 1:
The patent segments the detection process into distinct stages: first processing three-dimensional point cloud data to identify candidate objects, then using two-dimensional luminance images to verify and refine detection results. This segmentation allows complex multi-data processing to be broken down into manageable steps, reducing overall processing complexity while maintaining high detection accuracy.
Solution Approach 2:
The patent transitions from processing only three-dimensional spatial data to incorporating two-dimensional luminance information. This dimensional complementarity allows the system to leverage the strengths of both data types—depth information for spatial localization and luminance information for reflectance characterization—thereby improving detection accuracy without requiring exponentially increased processing power.
3Measurement precision
If clustering with strict thresholds is used, then detection precision is high, but low-reflectance objects are missed
Solution Approach 1:
The patent changes the detection parameters by incorporating luminance thresholds in addition to spatial clustering criteria. Instead of relying solely on strict spatial proximity thresholds that miss low-reflectance objects, the system uses luminance information to adjust detection sensitivity, thereby maintaining high precision while improving reliability for detecting objects with varying reflectance properties.
Solution Approach 2:
The patent implements dynamic threshold adjustment by adapting detection parameters based on the characteristics of the detected objects and environmental conditions. The system dynamically balances between precision and reliability thresholds, allowing it to maintain high detection precision for high-reflectance objects while becoming more sensitive to low-reflectance objects through luminance-based adjustments.
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
The solution significantly improves the ability to detect low-reflectance objects by enhancing clustering processes and integrating 2D and 3D data, thereby increasing detection accuracy and reliability.
Implementation Method 1
detection of an object having a low reflectance
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
An object detecting device (100) includes: a first luminance obtainer (11) and a distance obtainer (13) that obtain a first luminance image corresponding to an image-capturing area and a depth image corresponding to the image-capturing area of the first luminance image, respectively; a first clustering unit (101) that generates a group that is a set of pixels that is within a certain range and is taken as the same object in the depth image, and determines that the group is a cluster when a total number of pixels included in the group is at least a first threshold; a second clustering unit (102) that determines that the group is a cluster, when (i) the total number of pixels included in the group is less than the first threshold and (ii) luminance of a pixel group that is in the first luminance image and corresponds to the group is at least a second threshold; and a 3D object detector (17) that detects an object in the depth image based on the cluster, and generates 3D object information indicating the object detected.