RGB Camera Crop Property Sensing Using AI
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current agricultural harvesting machines rely on infrared sensors for moisture and dry matter content determination, which require specific wavelengths and are costly, while existing spectral reflection-based systems cannot accurately measure these properties within the 600-900 nm range.
Innovation Solution
A self-propelled agricultural machine equipped with an RGB camera and artificial intelligence, specifically a neural network, determines moisture and dry matter content from RGB images, eliminating the need for infrared devices by using convolutional neural networks to analyze visible light spectra and associate RGB images with crop flow scalar values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If infrared sensors operating at 1350-1550 nm wavelengths are used to detect moisture content, then measurement accuracy is improved, but device cost and complexity increase
Solution Approach 1:
The patent changes the measurement parameter from infrared wavelength (1350-1550 nm) to visible light wavelengths (400-700 nm). By training a neural network to analyze RGB images captured in the visible spectrum, the system achieves moisture content measurement without requiring complex infrared sensors, thus reducing device complexity while maintaining measurement capability
Solution Approach 2:
The patent creates a computational model (neural network) that copies the moisture detection functionality of infrared sensors. Instead of physically measuring infrared reflectance, the neural network learns to infer moisture content from visible light RGB patterns, replacing expensive hardware with a software-based solution
2Measurement precision
If infrared sensors are installed in harvesting machines, then crop moisture content can be determined, but the overall weight of the machine increases
Solution Approach 1:
The patent replaces physical infrared sensing hardware with a computational copy of the measurement function. The neural network processes standard RGB camera images to infer dry matter content, eliminating the need for heavy infrared sensor assemblies while preserving the measurement capability
Solution Approach 2:
The patent substitutes a mechanical/optical measurement system (infrared sensors) with an information processing system (neural network analyzing visible light images). This replacement reduces the mechanical weight of the harvesting machine while achieving the same measurement objective
3Adaptability or versatility
If multiple specialized sensors are used for different crop properties, then measurement coverage is improved, but manufacturing costs increase
Solution Approach 1:
The patent makes the RGB camera device universal by training a single neural network model that can determine multiple crop properties (moisture content, dry matter content, nitrogen content) from the same visible light images. This multi-functional approach eliminates the need for multiple specialized sensors, reducing manufacturing costs while maintaining comprehensive measurement coverage
Solution Approach 2:
The patent merges the functionality of multiple specialized sensors (infrared moisture sensors, dry matter sensors, nitrogen sensors) into a single RGB camera system with a multi-purpose neural network. By combining these measurement capabilities into one device, the system reduces overall manufacturing cost while maintaining adaptability across different crop property measurements
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present invention relates to a self-propelled agricultural machine (3) with an artificial intelligence for evaluating an RGB image (8) and a method for generating a data set (1) for training this artificial intelligence. The present invention is based on the general idea that an artificial intelligence is designed and configured to determine a crop property from an RGB image (8).