Produce Detection Using Concatenated Neural Network Features

Solution Overview

Problem

Current manual inspection methods for produce quality, such as fruit and vegetables, are prone to human errors and biases, leading to delays in processing and reduced shelf life due to the time-consuming nature of counting and defect identification.

Innovation Solution

Combining feature sets from multiple pre-trained deep learning architectures, such as Inception, Densenet, and Xception, through feature detection and concatenation, followed by additional analysis using convolution and dense layers, to enhance object detection and defect classification accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual inspection by QC associates is used, then quality control can be performed, but inspection time is excessive (50% of time spent on counting and defect identification) and human errors occur

Engineering Contradiction:
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent replaces the manual mechanical inspection process with an automated computer vision system using deep learning models (ResNet, VGG, Inception) that process images to detect defects and classify produce quality, eliminating the need for human QC associates to manually inspect each item

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system creates digital copies (images) of produce items and analyzes these copies using neural networks, allowing repeated inspection without physically handling the produce, thus reducing inspection time while maintaining accuracy

Inventive Principle:
Principle #26Copying

2Reliability

If manual inspection is used, then quality can be assessed, but the process is prone to human errors and biased inspection results

Engineering Contradiction:
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent replaces human inspectors with an automated AI-based inspection system that uses deep learning models to consistently analyze produce quality, eliminating variability caused by human factors such as experience, perspective, and training differences

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system performs self-service inspection by automatically capturing images and analyzing them through neural networks without requiring human intervention, thereby reducing processing delays and ensuring consistent, unbiased results

Inventive Principle:
Principle #25Self-service

3Manufacturing precision

If complete manual inspection is executed, then all produce can be inspected, but GDC processing time is prolonged

Engineering Contradiction:
Core Design Contradiction:
Manufacturing precisionVSProductivity

Solution Approach 1:

The patent replaces slow manual inspection with high-speed automated image processing using deep learning models that can rapidly analyze multiple produce items simultaneously, dramatically increasing processing speed while maintaining comprehensive quality control

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system enables continuous inspection by automatically processing images through neural networks without interruption, allowing inspecting to occur continuously rather than in discrete manual cycles, thus improving productivity while maintaining completeness

Inventive Principle:
Principle #20Continuity of useful action

Data Source

PatentUS11734813B2System and method for produce detection and classification
Publication Date: 2023.08.22
  • US11734813B2 patent drawing
  • US11734813B2 patent drawing
  • US11734813B2 patent drawing

AI summary

Systems, methods, and computer-readable storage media for object detection and classification, and particularly produce detection and classification. A system configured according to this disclosure can receiving, at a processor, an image of an item. The system can then perform, across multiple pre-trained neural networks, feature detection on the image, resulting in feature maps of the image. These feature maps can be concatenated and combined, then input into an additional neural network for feature detection on the combined feature map, resulting in tiered neural network features. The system then classifies, via the processor, the item based on the tiered neural network features.