Farming Machine Obstruction Detection With Unified CNN Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous farming machines struggle to efficiently recognize a diverse range of obstructions in a field, including both common and uncommon obstacles, due to the time-consuming process of data collection and manual labeling required for training obstruction detection models, which can lead to unsafe operations and potential damage.
Innovation Solution
A farming machine equipped with sensors that capture image data, using a convolutional neural network trained on labeled images of obstructions with prescribed actions, to identify and navigate around obstructions, enhancing safety by modifying farming objectives in real-time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional multiple models are used to detect obstructions, then detection coverage is improved, but data collection time and manual labeling effort increase significantly
Solution Approach 1:
The patent applies universality by creating a single unified obstruction detection model that can identify multiple types of obstructions (workers, rocks, hay bales, equipment) across various field conditions. This single model replaces the traditional approach of using multiple separate models, thereby maintaining comprehensive detection coverage while significantly reducing data collection and labeling requirements.
Solution Approach 2:
The patent employs parameter changes by modifying the training approach from collecting extensive labeled data for each obstruction type to using a reduced dataset with augmented samples. The system changes the parameters of data representation and model training to achieve reliable detection with less manual labeling effort.
2Reliability
If more obstruction types are detected, then safety is improved, but the complexity of the detection system increases
Solution Approach 1:
The unified detection model achieves multi-functionality by detecting various obstruction types (workers, rocks, hay bales, equipment) within a single system architecture. This approach maintains comprehensive safety coverage while avoiding the complexity of integrating and coordinating multiple separate detection models.
Solution Approach 2:
The patent merges multiple detection capabilities into a single unified model. By combining the detection functions for different obstruction types into one model, the system reduces overall complexity while maintaining the ability to detect all relevant obstruction types for safety.
3Measurement precision
If manual labeling of obstruction images is performed, then model training accuracy is improved, but productivity decreases due to time-consuming manual effort
Solution Approach 1:
The patent applies preliminary action by pre-augmenting the training dataset with synthetic obstruction samples before model training. This preliminary preparation of diverse obstruction examples enables the model to achieve high accuracy without requiring extensive manual labeling of real-world images, thereby improving productivity.
Solution Approach 2:
The system uses copying by creating synthetic copies of obstruction images through data augmentation techniques. These copied and transformed images (with variations in position, scale, rotation, and lighting) serve as training data, reducing the need for manual collection and labeling of unique real-world obstruction images while maintaining training accuracy.
Data Source
AI summary
A control system of a farming machine is configured to identify obstruction in a field from image data of the field. The control system accesses an obstruction model configured to identify obstructions in a field from image data of the field. The obstruction model is generated by accessing image data of obstructions in a training field, each obstruction corresponding to a prescribed action occurring at a prescribed time, labelling the image data of the obstructions, and training the obstruction model based on the labelled image data. The control system captures image data of the field including an obstruction and inputs the image data into the obstruction model. Responsive to identifying the obstruction in the field, the control system modifies treatment instructions of the farming machine such that the farming machine performs a implements a farming objective while avoiding the obstruction in the field.