Monitoring Device Using 3D-to-2D Viewpoint Transformation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional object recognition techniques using three-dimensional labeling are computationally intensive due to high computation per pixel, limiting processing time and accuracy, especially when objects are overlapping in three-dimensional space.
Innovation Solution
A monitoring device that uses a three-dimensional laser scanner to obtain distance data, calculates difference values between current and past data, extracts changed regions, and performs two-dimensional labeling to identify objects without the need for three-dimensional labeling, allowing for efficient object recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If three-dimensional labeling is performed to identify objects in a monitoring region, then object identification accuracy is improved, but computation time increases significantly
Solution Approach 1:
The patent transforms the three-dimensional labeling problem into a two-dimensional labeling problem by creating a transformed viewpoint image from a directly-above perspective. This dimensionality change allows the system to identify objects in 3D space using 2D labeling techniques, significantly reducing computation time while maintaining identification accuracy. The transformed viewpoint image represents the monitoring region as seen from directly above, enabling efficient 2D aggregate labeling that corresponds to 3D object structures.
2Measurement precision
If three-dimensional labeling is used to distinguish overlapping objects, then object separation accuracy is improved, but processing complexity increases
Solution Approach 1:
The patent resolves the complexity of separating overlapping objects by viewing the monitoring region from a directly-above perspective and creating a transformed viewpoint image. This approach transforms the complex 3D separation problem into a simpler 2D labeling problem, where adjacent pixels in the transformed image can be easily grouped into aggregates using standard 2D labeling techniques, thereby reducing processing complexity while maintaining separation accuracy.
Solution Approach 2:
The patent creates a transformed viewpoint image that is a 2D representation or copy of the 3D monitoring region viewed from directly above. This copied representation allows the system to perform labeling operations on a simplified 2D structure that mirrors the 3D object arrangements, making the labeling process less complex while preserving the ability to distinguish overlapping objects.
3Productivity
If conventional two-dimensional camera imaging is used, then processing speed is improved, but object recognition accuracy deteriorates when objects overlap in three-dimensional space
Solution Approach 1:
The patent bridges the gap between 2D processing speed and 3D recognition accuracy by creating a transformed viewpoint image that represents the 3D monitoring region from a directly-above perspective. This transformed 2D image retains the speed advantages of 2D processing while encoding 3D spatial relationships that enable accurate recognition of overlapping objects, effectively combining the benefits of both approaches.
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
Enables accurate identification of objects in a monitoring region with reduced computational costs, effectively addressing the limitations of three-dimensional labeling by using two-dimensional labeling and transformed viewpoint images.
Implementation Method 1
a three-dimensional laser scanner 10 measuring a monitoring region
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A monitoring device conducts a process of obtaining pieces of distance data representing distances to a plurality of physical objects present in a monitoring region, from measurement results of the monitoring region, and using the distance data as current distance data; obtaining past distance data from the measurement results, and converting the past distance data into comparison distance data; calculating difference values between the current distance data and the comparison distance data, and extracting changed regions whose difference values are greater than or equal to a threshold; creating an image obtained by transforming a frontal viewpoint image such that a viewpoint of the measurement of the monitoring region is moved, the frontal viewpoint image being based on the current distance data and the changed regions; and identifying the plurality of physical objects present in the monitoring region, on the basis of the frontal viewpoint image and the coordinate transformed image.