Thesis Topics
Computer vision for high throughput phenotyping of maize ears
1. Thema / Topic: Computer vision for high throughput phenotyping of maize ears
2. Type of Thesis: M.Sc.
Supervisor: Prof. Dr. Joachim Müller
Contact Person: Khandoker Ahammad, Leon Oehme
3. Beschreibung der Arbeit / Description and objective:
Maize (Zea mays) is one of the most important crops in the world being widely grown for bioenergy, animal nutrition and human consumption. As the maize ear is the most nutritious part of the plant, yield parameters such as ear length and kernel number are highly relevant for breeding superior varieties. These traits are usually measured destructively by threshing the ears (e.g. for counting kernels) or they are measured manually by human workers, which is tiresome, expensive and can be inconsistent.
Therefore, we developed a low-cost scanner for maize ears which is capable of rapidly and accurately measuring the maize ear phenotype. The system includes both capturing of RGB images and real-time image analysis based on modern AI-optimized hardware. In the future, existing algorithms should be improved and new analysis algorithms should be embedded. Additionally, new spectral sensors could be implemented.
Your tasks could include:
· Literature research
· Writing code for image analysis and training machine learning models
· Manual measurements of maize ears
· Application of the scanner
4. Anforderungen / Requirements for the applicants:
Skills in any programming languages are beneficial but being willing to learn basic knowledge in Python and Linux is sufficient.
5. Kontakt / How to apply:
Please send a mail to: khandoker.ahammad@uni-hohenheim.de
6. Bewerbungsfrist / Application deadline:
30th of June, 2025
7. Dauer der Arbeit / Duration:
3 - 6 months
Early Detection of Fusarium Head Blight in Winter Wheat
Early Detection of Fusarium Head Blight in Winter Wheat
1. Thema / Topic:
Early Detection of Fusarium Head Blight in Winter Wheat Using Thermal and Hyperspectral Cameras with Deep Learning
2. Type of Thesis: M.Sc.
Supervisor: Prof. Dr. Joachim Müller
Contact Person : Dr. Shamaila Zia Khan, Khandoker Ahammad
3. Beschreibung der Arbeit / Description and objective:
Fusarium Head Blight (FHB), caused by Fusarium graminearum and Fusarium culmorum, is a significant threat to global wheat production [1]. Early detection of FHB is essential for minimizing crop losses and ensuring food security. This study explores the use of advanced imaging technologies, including infrared thermography (IRT) and hyperspectral imaging (HSI) or RGB imaging, for early detection of FHB in wheat plants. The effectiveness of hyperspectral imaging, particularly in the visible (VIS) and near-infrared (NIR) regions (400–900 nm), in identifying early FHB symptoms is well-documented, with HSI showing high accuracy in detecting infections during the early stages of the medium milk phase. Furthermore, this research integrates Deep learning classification techniques for classifying healthy and infected wheat spikes based on data from these imaging modalities. These techniques have been shown to enhance classification accuracy and provide valuable insights into disease progression. The application of machine learning, including Convolutional Neural Networks (CNNs), to process imaging data for real-time, automated FHB detection is aligned with current research in precision agriculture and integrated disease management. The findings from this study contribute to the growing potential of imaging and machine learning in improving disease management practices and optimizing crop health monitoring. The objectives of the study are:
• Collect thermal and hyperspectral data from infected wheat plants.
• Analyze temperature variations and pigment-specific changes related to FHB.
• Develop and optimize a deep learning model for early detection.
• Validate the model, assess its applicability, and report findings.
4. Anforderungen / Requirements for the applicants
• Basic understanding of plant biology or agricultural sciences.
• Interest in imaging technologies (thermal or hyperspectral) and their applications in agriculture.
• Eager to learn and work with deep learning models (experience a plus but not required).
• Familiarity with data analysis tools (e.g., Python) is beneficial but not required.
• Enthusiasm for hands-on fieldwork and analyzing image data.
5. Kontakt / How to apply:
Please send your CV and your academic transcripts to:
Dr. Shamaila Zia Khan Email: shamaila.ziakhan@uni-hohenheim.de
Khandoker Ahammad Email: khandoker.ahammad@uni-hohenheim.de
6. Bewerbungsfrist / Application deadline: 30th April. 2025
7. Dauer der Arbeit / Duration: 1st of May until 30th of November 2025