Machine Vision Fruit Picking Using Neural Network Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current fruit and vegetable picking robots are limited in their ability to identify and pick multiple types of fruits and vegetables, requiring different robots and end picking apparatus for each type, and struggle to determine the optimal grabbing point due to variations in type, shape, and position, affecting picking accuracy.
Innovation Solution
A fruit and vegetable picking method based on machine vision using a pre-trained neural network model, specifically a Mask r-cnn network, to identify the type of fruit or vegetable and determine the cutting point, allowing a single device to pick multiple types by preprocessing images, extracting features, and controlling an end picking apparatus to cut the stalk without damaging the pulp.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If different robot arms and end picking apparatus are used for each type of fruit, then picking accuracy for each specific fruit type is improved, but device complexity and processing cost greatly increase
Solution Approach 1:
The patent implements a universal picking robot system that can identify and pick multiple types of fruits and vegetables using a single robot arm and end picking apparatus. The system uses a neural network model trained on diverse fruit and vegetable data to achieve multi-functional capability, eliminating the need for dedicated picking devices for each fruit type while maintaining high picking accuracy through accurate identification and optimal grabbing point determination.
2Device complexity
If a single picking device is used for multiple fruit types, then device complexity is reduced, but the ability to accurately determine grabbing points for different fruit types and positions deteriorates
Solution Approach 1:
The patent employs a neural network model that learns and adapts to different fruit and vegetable parameters including type, shape, size, and position. The model takes diverse input parameters from images and outputs optimized grabbing points for each specific fruit instance, enabling a single device to accurately determine grabbing points across multiple fruit types by dynamically adjusting its decision parameters based on learned patterns.
Solution Approach 2:
The system uses image processing to create a digital representation (copy) of the fruit or vegetable, including its shape, position, and characteristics. The neural network analyzes this copied representation to determine the optimal grabbing point, allowing the system to make accurate picking decisions for any fruit type without physical contact until the final picking action.
3Device complexity
If traditional vision systems are used for fruit identification, then system simplicity is maintained, but the ability to identify multiple types of fruits and vegetables deteriorates
Solution Approach 1:
The patent replaces traditional mechanical or rule-based vision systems with an intelligent neural network model. This substitution enables the system to automatically learn and identify multiple types of fruits and vegetables from images without requiring complex manual programming or multiple specialized sensors, achieving high versatility while maintaining relatively simple system architecture through software-based intelligence.
Data Source
AI summary
A fruit and vegetable picking method and device based on machine vision and storage medium that includes: acquiring fruit image of fruit or vegetable to be picked currently, calling pre-trained neural network model to identify fruit image, and determining type of fruit or vegetable; acquiring type of fruit or vegetable, and determining cuttable area on stalk of fruit or vegetable according to type of fruit or vegetable, locating cutting point; controlling end picking apparatus to cut off stalk of fruit or vegetable, according to cutting point being determined. By using pre-trained neural network model to detect fruit and cut stalk for picking, one device may be able to pick multiple fruits and vegetables, have high versatility and picking accuracy, avoid hurting pulp, and have improved harvest efficiency and quality of fruits and vegetables picked.