Developing image processing and machine learning tools for automatic identification and quantification of cabbage stem flea beetle pressure.

Background

Cabbage stem flea beetle (Psylliodes chrysocephala, CSFB) is one of the main pests of oilseed rape (Brassica napus, OSR) in Europe. Adult beetles cause considerable feeding damage to cotyledons and young leaves, threating crop establishment. Their larvae damage plant petioles and stems causing reduced vigour and increased risk of frost damage and disease, reducing overwintering survival1. Traditional methods of assessing CSFB damage are labour intensive, subjective, and often inaccurate, resulting in inadequate pest management. Using image processing, machine learning / deep learning techniques for accurate, automatic damage assessment offers a viable way to improve monitoring for farmers and researchers.

The student will produce images of feeding damage under controlled conditions and from field experiments (using the OCH farm network). They will be trained in field assessment and invertebrate behavioural ecology methods to enable this. The student will investigate different approaches for image acquisition, will develop methods to integrate data of different quality from different sources and will compare and develop a variety of AI methods for automatic identification and quantification of cabbage stem flea beetle pressure in field and laboratory environments.  

Objectives

1. Develop image processing and machine learning tools to accurately identify OSR plants in the field 
2. Develop image processing and machine learning tools to identify (shot holing) and quantify (leaf area removed) adult CSFB feeding damage and discriminate between similar leaf damage (e.g. by slugs and pigeon).
3. Develop image processing methods and machine learning tools to identify and count CSFB adults and their natural enemies (the parasitoids Microctonus brassicae and Tersilochus microgaster) from yellow water/sticky traps.
4. Develop image processing for automatic solutions to identify count CSFB larvae (and discriminate the three larval stages).

Novelty

Image processing and machine learning tools are novel tools that have just started to be explored for agricultural pest management. To our knowledge there is no tool to automatically identify and quantify CSFB adult or larval damage. 

Timeliness

The growing pressure of CSFB and the reduction in OSR area due to pest pressure2 highlights the need for sustainable pest management techniques. Monitoring pest damage is a key principle of Integrated Pest Management, but current assessments of CSFB damage are time-consuming and prone to human error. Automatic assessment of CSFB damage and the presence of their natural enemies will be revolutionary tools that will help growers and researchers to quantify pest damage more rapidly and precisely, enabling timely and targeted interventions while reducing the use of chemical insecticides and increasing reliance on biocontrol. These tools will also help to quantify impacts of crop/habitat management methods on pest control and functional biodiversity. 

Research environment

The student will benefit from a highly collaborative environment across the international partner institutions in the UK and Denmark. This partnership brings together diverse expertise, perspectives, and technical capabilities to improve CSFB monitoring from a multidisciplinary approach. The supervisory team includes experts in entomology, ecology, image analysis, and computational modelling who are committed to support the student’s professional development – preparing them for a successful career in academia, industry or the public sector.

1 https://doi.org/10.1111/gcbb.12918

2 https://doi.org/10.1111/gcbb.12922