Software that classifies wheat aphids gets new version

Whether to facilitate decision-making or offer more precision in crop diagnoses, technology is increasingly advancing as an ally in the agribusiness area. Since 2018, one of the tools that has been working with the purpose of helping to improve productivity in the field is AphidCV, a software aimed at classifying and recognizing aphids, developed through a partnership between the Postgraduate Program in Applied Computing at University of Passo Fundo (UPF) and Embrapa Trigo. Five years after launch, the program has a third version, which offers expanded species detection, improvements in handling and more practical access.

AphidCV was developed to count and verify the morphometry of aphids, separating them into categories (apterous adults, winged adults and nymphs). “The measurement of insect populations in plants is necessary in several studies, such as defining the pest limit for intervention with chemical control, evaluating the level of genetic resistance of plants, evaluating factors that affect population growth and the reproductive capacity of insects. . These works are usually carried out manually, with the support of magnifying glasses in a tiring process, time consuming and subject to human errors that affect precision”, explains Embrapa Trigo researcher, Douglas Lau. In practice, the AphidCV software works on the image of a glass plate containing the insects, instantly processing the total number and size of each individual. “With the support of computer vision, it is possible to reduce analysis time and increase the number of samples in the period”, says Lau.

The new version called "AphidCV 3.0 Web" implements a fast and automated method for counting and classifying aphids (aphids), using image processing, computer vision and deep learning techniques. The software is intended for use by agricultural researchers, entomologists and biologists. “The first version was born from a request from Embrapa, with the objective of creating an image processing and computer vision tool for identifying aphids and only classifying and recognizing one species of aphid. Now, the third edition is advancing and has support for identifying individuals of four species”, explains UPF professor Rafael Rieder.

Another differentiator of the new version is the practicality of access. In the first two versions, the software was installed on the computer, while the latest version is web-based, so access can be done using a username and password over the internet. “The program is now integrated with TrapSystem, an online platform for managing data collected from insect traps, with different tools that facilitate the analysis of aphid population fluctuation and the organization of historical series. Our idea is to continue updating these intelligent insect detection and classification models and also offer them to other species, as soon as we have a larger database”, says Rieder.

Research involves dozens of students and professionals

Activities related to AphidCV are part of the daily research carried out at the Computational Visualization and Modeling Center. This laboratory, installed next to the UPF Scientific and Technological Park, develops solutions in the areas of Virtual Reality, Augmented Reality, Mixed Reality, Machine Learning, Image Processing and Computer Vision. It has the participation of undergraduate and postgraduate students from UPF computing courses: Computer Science, Computer Engineering, Systems Analysis and Development and Masters in Applied Computing.

In the case of AphidCV, the development of the platform also involves institutions that work in an integrated and collaborative way, creating solutions that allow the different necessary steps to be carried out. In addition to Embrapa Trigo, there is a partnership with researchers from the Federal Institute of Education, Science and Technology Sul-Riograndense (IFSul) – Campus Passo Fundo. Currently, around 25 students and professionals from the three institutions are involved in the research. The objective is centered on the development of new software and evaluation with new users, aiming to form a chain of technologies to meet different agricultural needs.

The research gained notoriety and, recently, Professor Rafael was awarded a Technological Productivity and Innovative Extension Grant from CNPq. The proposal "Computer Vision Solutions for managing epidemics caused by viruses transmitted by insects" aims to develop, improve and validate Computer Vision and Machine Learning solutions, integrating them with data management systems and platform simulation models web, for monitoring agricultural crops and making decisions on managing epidemics caused by viruses transmitted by insects. “Basically, it concerns the evolution of the AphidCV, InsectCV, and other computing technologies applied to agriculture”, highlights the professor, who also maintains contact with foreign institutions that have shown interest in the research.