Machine Vision System for Robotic Harvesting
Automated Mushroom Detection and Maturity Assessment System
The Challenge:
Mushroom farms in the United States face critical labor shortages, particularly in harvesting tasks. The selective harvesting of white button mushrooms is labor-intensive and relies heavily on skilled workers to make judgment calls based on the size, cap shape, and coloration of the mushrooms. These challenges highlight the need for innovative automation to improve efficiency and consistency in mushroom harvesting.
Our Solution:
At the Smart Agriculture Lab, we have developed a cutting-edge computer vision system that automates mushroom detection and maturity estimation using depth images. This technology aims to reduce labor dependency and optimize the harvesting process, ensuring mushrooms are picked at their peak maturity for extended shelf life.
How It Works:
Depth Image Acquisition: We utilize a depth camera to capture detailed 3D images of mushroom beds, allowing precise detection of individual mushrooms at various growth stages.
Mushroom Detection and Segmentation: A novel image processing method combines depth and RGB images to accurately detect and segment mushrooms. The system is enhanced with YOLOv8, a deep learning model, to compare its performance against traditional methods.
Maturity Assessment: Using the depth images, our system analyzes the size and shape of mushroom caps. A Support Vector Machine (SVM) classifier then categorizes each mushroom as mature or immature based on extracted features, such as cap diameter and slope.
Figure: Image Analysis for Mushroom Detection and Maturity Estimation
Left and Center: Depth images showcasing detected mushrooms at various growth stages with segmentation markers indicating maturity levels. Right: Heatmap visualization of depth values providing insights into the spatial distribution and density of mushrooms, aiding in precise harvesting decisions. These images demonstrate the capability of our computer vision system to accurately identify and categorize mushrooms based on size and developmental stage.
Key Benefits:
Enhanced Efficiency: Automates the detection and maturity assessment process, significantly reducing the need for manual labor and subjective judgment.
Increased Accuracy: Our advanced machine learning and image processing techniques provide over 95% accuracy in classifying mushroom maturity, leading to more consistent harvest quality.
Reduced Waste: By precisely determining the optimal time for harvest, our system helps reduce waste and increases the overall yield and quality of mushroom crops.
Future Goals:
We are committed to further refining the technology by enhancing the boundary identification in mushrooms using depth images and integrating more sophisticated machine learning models. Our aim is to fully automate the mushroom harvesting process, making it more efficient, profitable, and sustainable
