Project Description
Addressing the persistent threat of Maize Streak Disease in Sub-Saharan Africa, which causes up to 100% yield loss and annual economic damages of USD 480 million annually, this system integrates a motor-controlled trapeze system with a Convolutional Neural Network and an overhead camera to scan maize crops, detecting early signs of MSD by analyzing color changes in leaves and producing a visual map of infected areas, thereby overcoming the limitations of resistant maize varieties, unsustainable pest management, and labor-intensive manual inspections.
Problem
Maize Streak Disease (MSD), caused by the Maize Streak Virus, severely impacts maize crops in Sub-Saharan Africa, leading to up to 100% yield loss and significant economic damage.
Current control measures, including resistant varieties and manual inspections, are inadequate, highlighting the need for more effective solutions.
MSD manifests early in the crop lifecycle with symptoms like leaf chlorosis—yellow streaks on leaves—that are difficult to detect manually in fields with up to 17,000 plants per acre. This challenge necessitates a scalable, automated detection system to facilitate early intervention and prevent extensive crop damage.
Typically, symptoms such as streaks on the leaves can become visible as early as 10 to 14 days after emergence, especially if the seedlings are infected soon after germination. This early appearance of symptoms is critical as it marks a vital period for taking measures to manage the disease before it can severely impact the plant's development and the overall yield of the crop.
MSD spreads quickly leading to about 100% yield losses
leaf chlorosis
MSD usually beings at the V5 stage
Solution Concept Model
Step 1: Setup of the Trapeze System
Step 2: Image Capture
Step 3: Image Analysis Using AI
Step 4: Mapping Infected Areas Using Heat Map Visualization
Step 5: Early Intervention and Management- Utilize the map to strategize targeted interventions,
Prototype & Proof of Concept(PoC)
Role
- AI/ML & Embedded Systems Engineer
Constructed and programmed a robotic trapeze overhead system to survey farming field for early crop disease detection.
Built a Convolutional neural network model for the detection of Maize Streak Disease by through the analysis of images of maize leaves taken by robot.
Teammates
Jingfei Huang (3D modelling and Design), Dan Kim (System Mapping and Project Managment)
