Wheat Yield Prediction Using Deep Learning Image Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for measuring wheat yield per unit area are time-consuming, laborious, and prone to subjective errors due to manual counting, and existing digital image processing methods suffer from image distortion and variability issues, limiting their applicability for rapid field data acquisition.
Innovation Solution
A method utilizing computer vision and deep learning technology to acquire and correct wheat spike images, recognize spikes and grains using trained models, and predict yield based on camera parameters, climatic conditions, and neural networks, enabling efficient and accurate calculation of spikes and grains per unit area.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual counting method is used to detect number of effective spikes and grains, then measurement can be performed, but time consumption and labor are excessive
Solution Approach 1:
The patent replaces manual mechanical counting with automated image processing technology. A camera captures images of wheat spikes, and computer vision algorithms automatically count spikes and grains, eliminating manual labor while maintaining measurement accuracy. The system processes multiple images to calculate yield metrics automatically.
Solution Approach 2:
The patent creates digital copies (images) of wheat spikes and grains, then processes these copies through image analysis algorithms to determine counts and yield. This allows multiple measurements to be taken from image copies without physically handling or destroying the actual wheat samples.
2Productivity
If manual counting method is used, then yield measurement can be obtained, but subjective factors affect counting accuracy
Solution Approach 1:
The patent replaces human subjective judgment with objective computer vision algorithms. The system uses standardized image processing criteria to count spikes and grains consistently, eliminating variability introduced by different operators' subjective interpretations of what constitutes an effective spike or grain.
3Productivity
If digital image processing technology is used to obtain spike characteristics, then counting speed improves, but image distortion causes measurement error
Solution Approach 1:
The patent applies distortion correction algorithms as a preliminary step before measuring spike characteristics. The system pre-processes images to compensate for expected lens distortion and perspective effects, thereby preventing measurement errors before they occur in the subsequent analysis steps.
4Extent of automation
If relationship between grain number and spike characteristics is established, then grain counting is enabled, but variety influence causes counting error
Solution Approach 1:
The patent applies different processing strategies tailored to specific wheat varieties. The system can identify variety characteristics and adjust counting parameters or thresholds accordingly, allowing optimized accuracy for each variety rather than using a single universal approach that may fail for certain types.
5Reliability
If indoor image acquisition device is used, then controlled imaging is achieved, but rapid field data acquisition is not possible
Solution Approach 1:
The patent designs an imaging system that functions reliably in multiple environments. The camera and image processing algorithms are configured to handle both controlled indoor conditions and variable outdoor field conditions, allowing the same system to be used for rapid field data acquisition while maintaining acceptable image quality through adaptive processing.
Data Source
AI summary
A method for observing a per unit yield of wheat based on computer vision and deep learning technology, including: acquiring wheat spike images and coordinate position data; calculating a camera parameter, and performing a distortion correction and a cropping on the images; performing a wheat spike recognition on the wheat spike image by using a deep learning target recognition model; performing a wheat spike recognition on the wheat spike image by using a trained deep learning target recognition model 1 and cropping a wheat spike from the wheat spike image; performing a wheat grain recognition on the wheat spike image by using a trained deep neural network target recognition model 2; calculating a number of spikes per unit area in a same wheat field by using a corrected wheat spike image and a recognized wheat spike; calculating a number of effective grains of spikes in same wheat field by using the corrected wheat spike image and a recognized wheat grain; predicting thousand-grain weight according to the number of spikes per unit area, climate condition and deep neural network model of thousand-grain weight; and calculating per mu yield according to predicted thousand-grain weight and the number of effective grains of spikes per unit area.