Researchers from HSE University and Tambov State Technical University (TSTU) have developed intelligent robotic systems for the monitoring and quality control of fruits and vegetables. The system allows gardeners to carry out agrotechnical and protective activity, ensuring the highest possible quality of the crop.
Between 2017 and 2021, the area given over to apple trees in Russia increased by 27.2% from 171,600 to 218,200 hectares. The amount of state support for horticulture for the period from 2013 to 2018 totalled 13.7 billion roubles, which made it possible to allocate 78,400 hectares for new gardens, including 51,500 hectares for intensive planting. This increase in gardens has lead to the development of agricultural technologies that help protect crops from damage and diseases.
Researchers from HSE University and Tambov State Technical University have developed a robotic platform for proximal sensing (measuring soil condition), monitoring diseases and growing plants in conditions of intensive gardening as part of the Mirror Laboratories project.
Intensive gardening is a modern way of growing fruit crops, providing a short payback period for investments. Intensive planting implies close planting of trees, a fast fruiting period, and relatively small adult plants.
HSE and TSTU researchers have developed algorithms that can be used to determine both visible and invisible damage to fruit in real time. To assess the prediction of fruit quality, the researchers used hyperspectral control, which is produced by a special robotic platform on wheels. The platform collects visual information and magnetometer and gyroscope readings. The data is then transferred to the cloud and processed on a computer.
The platform consists of a mobile robot that moves around an orchard of small apple trees and analyses their condition. It can move autonomously along the aisles, scanning foliage and fruits, and transmitting this information to the operator at the base station. The data is then used to analyse the condition of vegetation and fruits, determining diseases or crop quality.
During the study, the experts used hyperspectral images to detect mechanical damage to the skin of fruit trees. The accuracy of damage detection was 92%.
The hyperspectral method allows growers to obtain data that cannot be detected by visual observation. Detection is carried out at the subpixel level or by combining data.
Researchers note that the results obtained can already be used in decision support systems in precision gardening. For example, gardeners can carry out agrotechnical and protective measures in good time, ensuring the highest possible quality of the crop. HSE experts plan to develop algorithms that will make it possible to determine the internal quality of fruits, including maturity, firmness, and sugar content.