Neural Identifier Segmentation for Multi-Item Image Scanning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems in fulfillment centers struggle to efficiently process multiple structured identifiers in a single image, requiring separate presentations of each item for accurate scanning, leading to increased operational costs and time inefficiencies.
Innovation Solution
A neural network-based system that uses cameras and computer vision to identify structured identifiers on items being manipulated by humans or robotic arms, allowing scanning without explicit presentation to an optical scanner, by generating binary segmentation masks and decoding structured identifiers from images captured during item movement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If each item is individually presented in front of a scan zone for accurate processing, then scanning accuracy is improved, but operational speed and efficiency deteriorate
Solution Approach 1:
The system segments the image into multiple regions of interest using neural networks, identifying individual objects and their associated structured identifiers within a single captured image. This allows simultaneous processing of multiple items without requiring sequential presentation, thereby maintaining scanning accuracy while improving operational speed.
Solution Approach 2:
The system transitions from sequential temporal processing (one item at a time) to parallel spatial processing (multiple items simultaneously in a single image). By utilizing the spatial dimension of the image data and applying neural network-based object detection and segmentation, the system can identify and process multiple structured identifiers concurrently, resolving the contradiction between accuracy and speed.
2Measurement precision
If each item is individually positioned in the camera's field of view, then identification accuracy is improved, but time consumption and operational costs increase
Solution Approach 1:
The system performs preliminary segmentation and identification of objects and their structured identifiers directly from the captured image before any positioning or manipulation occurs. The neural networks pre-process the image data to identify regions of interest, extract structured identifiers, and determine object locations, eliminating the need for subsequent individual positioning and reducing time consumption while maintaining identification accuracy.
Solution Approach 2:
The system replaces the mechanical positioning system (robotic manipulation to present items sequentially) with a computer vision-based identification system. Instead of mechanically moving each item into position for scanning, the system uses neural networks to automatically identify and locate structured identifiers within the captured image, significantly reducing time consumption while maintaining or improving identification accuracy.
3Measurement precision
If robotic picking and positioning is used for each item, then scanning precision is improved, but operational complexity and costs increase
Solution Approach 1:
The system employs a universal neural network-based processing pipeline that can handle multiple objects with different structured identifiers simultaneously within a single image. This multi-functional approach eliminates the need for separate robotic picking and positioning operations for each item, reducing operational complexity while maintaining scanning precision through consistent algorithmic processing of all identified identifiers.
Solution Approach 2:
The system replaces complex mechanical robotic manipulation systems with a software-based computer vision system. Instead of using robots to pick, position, and present each item sequentially for scanning, the system uses neural networks to automatically detect, segment, and identify structured identifiers directly from images captured during normal item flow, significantly reducing device complexity and operational complexity while maintaining or improving scanning precision.
Data Source
AI summary
Systems and methods are disclosed for identifying manipulated items using neural networks. Systems generate, using a first neural network, indications of an object of interest from the two or more objects in one or more images. Systems identify, using a second neural network, structured identifiers based, at least in part, the one or more images. Then, the system selects a structured identifier associated with the object of interest from the structured identifiers based, at least in part, on the indications to the object of interest.