RGBD Camera Point Cloud Processing for Dark Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Laser-based navigation systems in robots struggle to detect dark or black objects, leading to potential collisions and scratches, especially with vehicles, due to laser absorption.
Innovation Solution
A point cloud data processing method utilizing RGBD cameras to combine color and depth information for enhanced target segmentation, followed by pixel-level denoising and plane fitting to augment point cloud data, improving detection of dark objects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If laser-based navigation is used for obstacle detection, then the robot can navigate autonomously, but dark objects cannot be detected due to laser absorption
Solution Approach 1:
The patent combines laser point cloud data with RGB image data to create a fused point cloud representation. The RGB images provide color information that complements the depth information from laser, enabling the system to detect dark objects that would otherwise be invisible to laser alone. This merging of multiple data sources resolves the contradiction by maintaining reliable navigation while improving dark object detection precision.
Solution Approach 2:
The patent introduces an RGB camera as an intermediary sensor to capture visual information about dark objects. The RGB data acts as a mediator that bridges the gap in laser detection capability, providing color and texture information that helps identify objects with low laser reflectivity. This intermediary approach enables the system to overcome the laser absorption problem while maintaining autonomous navigation reliability.
2Measurement precision
If laser point cloud data is used alone, then the navigation system is simple, but detection accuracy for dark objects is insufficient
Solution Approach 1:
The patent merges laser point cloud data with RGB image data into a unified point cloud representation. This combination improves detection accuracy by incorporating color information that helps distinguish dark objects, while the merging process is designed to be computationally efficient, minimizing the increase in system complexity.
Solution Approach 2:
The patent adds a color dimension to the traditional depth-only point cloud data by integrating RGB information. This dimensional enrichment transforms the detection capability from relying solely on intensity variations to utilizing color characteristics, thereby improving dark object detection accuracy without requiring a completely new detection architecture.
3Measurement precision
If RGBD camera is used to collect image data, then dark objects can be detected through color information, but the system complexity increases
Solution Approach 1:
The patent makes the point cloud data structure multi-functional by incorporating both depth and color information in a unified representation. This universal point cloud structure serves multiple purposes: it maintains compatibility with existing laser-based processing pipelines while simultaneously enabling color-based object identification, thereby improving dark object detection without requiring separate processing systems.
Solution Approach 2:
The patent enables the point cloud data structure to self-enhance by integrating RGB information directly into the point cloud representation. This self-service approach allows the data structure to provide its own color information without requiring external processing systems, reducing overall system complexity while improving detection precision.
Data Source
AI summary
A point cloud data processing method, and a robot and a robot control method using the same are provided. The method includes: obtaining image data including an RGB image and a depth image that is collected through an RGBD camera; obtaining an original mask image by segmenting out targets from the RGB image using a target segmentation mode; obtaining an optimized mask image by performing a pixel-level processing on the mask image; obtaining, based on the optimized mask image and the depth image, a plane equation of each of the targets in the optimized mask image; performing, using the plane equation of each of the targets, a depth value assignment on a plane position of the target in the optimized mask image that is not assigned with the depth value; and obtaining target point cloud data by performing a point cloud conversion on the depth image.