??AI for Cucumbers: Nutrient Stress Detection in CEA ??

If you wish to get regular "AI in Agriculture" newsletter to your mailbox, please, add your email to my mailing list.

Join more than 8,500 readers who get weekly emails about AI in Agriculture

What causes nutrient deficiency in cucumbers?

Nutrient deficiency in cucumbers occurs when the plants do not receive adequate amounts of essential nutrients, such as nitrogen, potassium, or calcium, which are crucial for their growth and development.

This can result from poor soil quality, imbalanced fertilization, water stress, or suboptimal environmental conditions, leading to symptoms like stunted growth, yellowing leaves, and reduced fruit quality.

Inadequate nutrient supply hampers the plant's ability to carry out vital physiological processes, ultimately affecting overall yield and health.

In the figure above: (A) Plant debris was mixed with sterile soil (CC). (B) Plant debris was mixed into the soil and covered with polyethylene film (CC-PE). (C) Plant debris was mixed into the soil, and CaCN2(CC-CaCN2) was added. (D) Natural soil without any treatment (CK).

Nutrient Stress Symptom Detection in Cucumber Seedlings Using Segmented Regression and a Mask Region-Based Convolutional Neural Network Model

Country: Republic of Korea ????

Published: 17 August 2024

This study aimed to detect the early signs of nutrient stress in cucumber seedlings using a combination of segmented regression and Mask R-CNN (convolutional neural network) models, focusing on morphological and textural features extracted from images.

The main question that the researchers wanted to answer was: How can early signs of nutrient stress in cucumber seedlings be detected accurately using computer vision and machine learning techniques before human visual inspection?        

The researchers employed a combination of statistical and deep learning methods to detect nutrient stress in cucumber seedlings. Segmented regression analysis was applied to identify changes in plant morphology and texture over time, using features such as canopy area and leaf texture variations.

The Mask R-CNN model, utilizing a ResNet-101 backbone, was trained with annotated images to segment and classify seedlings into stressed and non-stressed groups.

In the figure above: (a) original image with stress conditions (0 dSm?1), (b) annotated image with stress conditions (0 dSm?1), (c) original image with stress conditions (6 dSm?1), and (d) annotated image with stress conditions (6 dSm?1).

To achieve this, the researchers captured images of seedlings at different stages and used textural analysis (e.g., energy, entropy, homogeneity) to detect subtle signs of nutrient deficiency. Transfer learning was applied to train the model efficiently with a smaller dataset.

Use case of AI application for nutrient deficiency detection

AI was central to this research, particularly through the application of Mask R-CNN, a deep learning model, to segment and classify images of cucumber seedlings.         

1. The model identified the initiation and progression of nutrient stress by analyzing plant morphology and textural changes over time.
2. The integration of ResNet-101 enabled accurate feature extraction, and transfer learning reduced training time, enhancing the model's effectiveness in detecting stress early.

Key findings of the research

The key findings of the study demonstrated that the method could detect nutrient stress symptoms in cucumber seedlings 1.5 days earlier than visual inspection by humans.

The Mask R-CNN model achieved high performance with an F1 score of 93.4%, precision of 93%, and recall of 94%.

Notably, features such as the top projected canopy area, energy, entropy, and homogeneity were identified as reliable indicators of nutrient stress. This early detection method is crucial for improving plant health and productivity by allowing timely intervention.

Agricultural professionals, particularly those working in precision farming, can apply these results to monitor and manage plant health more effectively, especially in controlled environments like greenhouses.

Technologies used:

• Segmented regression analysis
• Mask R-CNN model
• ResNet-101 (pre-trained on ImageNet)
• Raspberry Pi 4 & Raspberry Pi Camera
• SAS software (for statistical analysis)

In the figure above: The outcomes of segmented regression for the TPCA and three textural characteristics for the detection of the initiation of stress caused by a nutrient deficit (0.0 dSm?1) in cucumber seedlings; (a) TPCA, (b) entropy, (c) energy, and (d) homogeneity. Black dots represent the change points for average seedling parameters at time ti.

In the figure above: (a) annotated image in the model for the nutrient deficit (0.0 dSm?1) stress dataset, (b) annotated image in the model for the excess nutrient (6.0 dSm?1) stress dataset, (c) nutrient stress detection and segmentation on the nutrient deficit (0.0 dSm?1) dataset, and (d) nutrient stress detection and segmentation on the excess nutrient (6.0 dSm?1) dataset.

References for "AI for Cucumbers: Nutrient Stress Detection in CEA ??"

For your interest (FYI)