Multispectral Berry Inspection Without Mechanical Handling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fruit and vegetable inspection machines are bulky, invasive, and unsuitable for delicate small fruits and berries, often causing spoilage due to mechanical handling, and lack space-efficient solutions for quality inspection.
Innovation Solution
A method using a stationary container within a camera obscura setup with multispectral imaging and chlorophyll fluorescence, supported by a conveyor belt, performs quality inspection of small fruits and berries without mechanical agitation, utilizing a multispectral sensor and illuminator system with selectable filters and a deep neural network for image processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional multispectral inspection machines use horizontal separation and sorting mechanisms, then product characterization efficiency is improved, but device complexity and bulkiness increase
Solution Approach 1:
The patent replaces mechanical separation and sorting systems with optical imaging and machine learning algorithms. The multispectral camera captures images of products in the container, and deep learning algorithms automatically identify and characterize individual units without physical manipulation, thereby maintaining high productivity while significantly reducing device complexity and space requirements.
Solution Approach 2:
The system creates digital copies (multispectral images) of the physical products and performs all analysis on these copies. This eliminates the need for physical handling and mechanical separation systems, allowing product characterization to be performed on static containers without complex moving parts or horizontal clearance requirements.
2Productivity
If traditional inspection machines use invasive mechanical handling, then product sorting and characterization are improved, but product quality deteriorates due to spoilage
Solution Approach 1:
The patent replaces all mechanical handling and contact-based inspection methods with non-contact optical imaging and computational analysis. The multispectral camera and deep learning algorithms enable complete product characterization without physical touch, eliminating mechanical damage and spoilage while maintaining sorting and characterization capabilities.
Solution Approach 2:
The system performs multiple functions (identification, sizing, quality assessment, ripeness detection) simultaneously through a single non-contact imaging system, eliminating the need for separate mechanical handling steps that would otherwise be required for each measurement, thereby preventing cumulative damage to delicate products.
3Measurement precision
If traditional inspection systems require horizontal clearance for product separation, then individual unit inspection is improved, but space requirements increase
Solution Approach 1:
The patent transitions from horizontal product separation to vertical stacking of containers, utilizing the vertical dimension for system layout. Multiple containers can be stacked one above another, and the overhead camera captures images looking down through the stack, enabling individual unit inspection without requiring horizontal clearance for product separation.
Solution Approach 2:
The system creates detailed digital representations of each product through multispectral imaging, allowing precise measurement and characterization to be performed on the digital copies rather than requiring physical separation and manipulation of actual products, thereby eliminating the need for horizontal clearance.
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 effective, non-invasive quality inspection of delicate fruits and berries in confined spaces, preserving product integrity and providing maturity and hardness indices through static imaging.
Implementation Method 1
multispectral illuminating means (2), e.g. LED emitters
Implementation Method 2
The product mass thus illuminated in turn reflects and/or emits light radiation 2b which, when acquired by the multispectral sensor means 3, forms multispectral images
Implementation Method 3
Preferably such optical reconstruction means are configured for chlorophyll fluorescence imaging
Implementation Method 4
multispectral sensor means 3, e.g. a multispectral camera
Data Source
Figure 1
Figure 2~3c
AI summary
The quality inspection method for inspecting a horticultural product comprises the following steps: -providing optical reconstruction means including multispectral illuminating means (2) and multispectral sensor means (3); -gathering loose units (6) of said horticultural product in a container (5); -illuminating with the multispectral illuminating means (2) the mass of product present in the container (5) by emitting a sequence of light radiations with different bandwidths; -capturing multispectral (3) images (10', 10") of the product mass present in the container (5) illuminated in this way; - identifying multispectral images (11, 11', 11") of individual units (6) of product by processing the multispectral images (10, 10', 10") of the product mass present in the container (5); - classifying individual (6) bulk units with quality indices associated with the spectrum of their multispectral images (11, 11', 11") thus identified; where the process of identifying the multispectral images (11, 11', 11") of the individual units (6) of product bulk is carried out by means of an artificial neural network trained to segment the images (10, 10', 10") of the product bulk in the container (5).