Bulk Sorting via CNN Segmentation of Hyperspectral Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current sorting technologies are limited in handling bulk quantities due to low capacity and speed, as they require careful feed preparation and are inefficient in processing overlapping objects, making them unsuitable for high-tonnage pre-concentration applications.
Innovation Solution
A bulk sorting system utilizing a convolutional neural network (CNN) with at least two convolutional layers, combined with optical sensors and a mechanical sorter, to analyze multi- or hyperspectral data and sort objects in bulk, enabling the identification and separation of overlapping objects and increasing throughput.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If current sorting technologies process individual particles with careful feed preparation, then identification accuracy is improved, but throughput capacity deteriorates to very low levels (up to 300 tonnes per hour for larger particles)
Solution Approach 1:
The system segments the bulk material stream into detectable individual objects using optical sensors, allowing each object to be identified accurately while maintaining bulk processing speed. The CNN then processes these segmented objects independently to achieve high identification accuracy without requiring careful feed preparation of the entire bulk stream.
Solution Approach 2:
The patent replaces traditional mechanical feed preparation systems with an optical sensing and neural network processing system. Instead of mechanically separating and preparing individual particles before sorting, the system uses optical sensors to detect objects in the bulk stream and a CNN to identify them, achieving both high accuracy and throughput.
2Reliability
If traditional sorters separate individual particles with careful feed preparation, then detection reliability is improved, but sorting speed deteriorates to levels unsuitable for high-tonnage pre-concentration
Solution Approach 1:
The system replaces mechanical feed preparation and detection systems with an optical sensing and neural network processing system. The optical sensors can detect objects at high speeds in the bulk stream, and the CNN processes the data rapidly to maintain both detection reliability and high sorting speed suitable for high-tonnage applications.
Solution Approach 2:
The patent changes the detection parameters from mechanical contact-based detection to optical detection. This allows the system to detect objects at higher speeds without compromising reliability, as optical sensors can capture multiple objects simultaneously in the bulk stream and the CNN can process these detections rapidly.
3Measurement precision
If current sorters require careful feed preparation for individual particle detection, then measurement precision is improved, but device complexity increases due to feed preparation requirements
Solution Approach 1:
The patent replaces complex mechanical feed preparation devices with a simpler optical sensing system. The optical sensors can detect objects directly in the bulk stream without requiring mechanical separation or positioning devices, significantly reducing device complexity while maintaining measurement precision through the CNN processing.
Solution Approach 2:
The optical sensing system serves multiple functions: it detects objects in the bulk stream, provides data for CNN processing, and enables sorting decisions all in one integrated system. This multi-functionality eliminates the need for separate feed preparation devices, reducing overall system complexity while maintaining detection precision.
4Measurement precision
If traditional sorting methods process single streams of objects, then identification accuracy is improved, but productivity deteriorates due to inability to process bulk quantities
Solution Approach 1:
The system segments the bulk material stream into detectable individual objects using optical sensors, allowing each object to be identified accurately while maintaining bulk processing speed. The CNN then processes these segmented objects independently to achieve high identification accuracy without requiring careful feed preparation of the entire bulk stream.
Solution Approach 2:
The patent transitions from processing a single stream of objects to processing bulk quantities by adding spatial dimensions to the detection system. Multiple optical sensors detect objects across the width and depth of the bulk stream simultaneously, and the CNN processes this multi-dimensional data to maintain identification accuracy while achieving high bulk processing capacity.
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 system achieves higher throughput and accuracy in sorting bulk objects, with a speed of 0.4 - 20 m/s and a throughput of 0.5 - 30 tons/hr, with an accuracy level of higher than 80%, effectively addressing the limitations of traditional sorting methods.
Implementation Method 1
at least one optical sensor arranged to capture reflected radiation of the objects and acquire the reflected radiation as multi- or hyperspectral data
Data Source
Figure 1
Figure 2
Figure 3a~3b
AI summary
A bulk sorting system for sorting objects (1) in bulk is provided. The bulk sorting system comprises: at least one radiation source (10) arranged to radiate the objects, at least one optical sensor (12) arranged to capture reflected radiation (22) of the objects and acquire the reflected radiation as multi- or hyperspectral data (24); a processing circuit (16) configured to analyze the reflected radiation of the objects by inputting the multi- or hyperspectral data into a convolutional neural network (CNN) (18) with at least two convolutional layers in order to either detect and classify the objects in the multi- or hyperspectral data and/or semantically segment the multi- or hyperspectral data; and a mechanical sorter (20) configured to sort the objects according to their classification and/or segmentation using the analysis of the processing circuit such that different overlapping and/or stacked objects are separated or treated as a single group of objects.