Pallet Load Protrusion Measurement Using 3D Image Coordinates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems fail to efficiently measure the protruding lengths of loads on pallets, leading to decreased stacking efficiency due to interference and alignment issues, necessitating a device capable of accurately determining these dimensions.
Innovation Solution
A measurement device utilizing an imager, image processing unit, and calculator that employs machine learning models (Key Point Detection using Mask R-CNN and Instance Segmentation) to extract specific regions and load regions from image data, calculating protruding lengths based on 3D coordinates and standard load form structure information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional measurement methods are used for load protruding lengths, then measurement process is simple, but measurement precision is insufficient and stacking efficiency decreases
Solution Approach 1:
The patent transitions from 2D image capture to 3D coordinate extraction by introducing depth information through machine learning models. The system captures images from multiple viewpoints and reconstructs 3D coordinates of load vertices, enabling precise protruding length measurement that accounts for spatial relationships and pallet orientations.
Solution Approach 2:
The patent replaces traditional mechanical measurement devices with an optical-machining learning system. Instead of using physical contact probes or mechanical gauges, the system uses imagers to capture images and machine learning models to automatically extract coordinates and calculate protruding lengths, eliminating the need for manual measurement operations.
2Productivity
If manual measurement and alignment methods are used, then device complexity is low, but productivity and stacking efficiency are reduced
Solution Approach 1:
The system performs automatic measurement and calculation without human intervention. The machine learning models automatically extract coordinates from images, calculate 3D positions, determine protruding lengths, and provide alignment guidance for optimal stacking, enabling the system to serve itself in the measurement and decision-making process.
Solution Approach 2:
The system performs preliminary measurement and analysis before the actual stacking operation. By capturing images, extracting coordinates, and calculating optimal stacking arrangements in advance, the system prepares all necessary information beforehand, allowing for efficient execution of the stacking process without time-consuming manual measurements during operation.
3Reliability
If simple image capture is used without advanced processing, then device complexity is low, but measurement precision and reliability are insufficient
Solution Approach 1:
The patent divides the measurement process into distinct segments: image capture, coordinate extraction using machine learning, 3D coordinate calculation, protruding length computation, and stacking optimization. Each segment is handled by specialized algorithms or modules, improving overall reliability by ensuring that each step is performed with appropriate precision and validation.
Data Source
AI summary
A measurement device includes: an imager that captures a target including a pallet and a load, the pallet including a reference target; an image processing unit that extracts a specific region in each of the reference target and a region of the load based on image data of the target; and a calculator that acquires first coordinate information indicating 3D coordinates associated with the image data of the target and standard load form structure information indicating a structural dimension of a standard load form of the pallet associated with coordinates of the specific region, and calculates protruding lengths of the load protruding from the pallet. The image processing unit extracts the specific region by inputting the image data of the target into a first machine learning model subjected to training by using, as training data, image data of the pallet and data indicating the specific region.