← All research topics

Computer Vision in Crop Monitoring

Tomato detection, ripeness, yield estimation and the effect of illumination.

Background

Machine vision supports harvesting decisions, yield forecasting and plant health monitoring in tomato production, but results depend strongly on imaging conditions and on choosing the right algorithm.

Objectives

• Map which data image processing can deliver for precision crop production and how it supports yield and harvest-time forecasting.
• Quantify how illumination affects colour-based detection of ripe fruit.
• Review the latest image processing methods for tomato detection and cultivation.

Methodology

• Literature review of machine vision, CNN and vision-transformer methods for tomato classification, counting, yield estimation, disease, pest and weed detection, navigation and harvesting robots.
• Controlled experiment: HSV colour segmentation of ripe tomatoes at 15 artificial illumination levels.

Results

• Colour-based fruit detection is stable at about 3,000 lx and above; at lower light the detected fruit looks smaller, and near 1,600 lx only about half of the real size is measured.
• The review summarises the image processing algorithms and results of the last two years and identifies open problems.
• Ongoing work extends this line to field-grown tomato phenotyping with depth data and segmentation models.

Related publications