The objective of the project is to develop a solution that combines machine learning and image analysis in order to determine the amounts of pre-determinated objects in image. In this case objects are are the flowers, raw berries and ripe berries. Cloud computing services will be applied, since they often have the features supporting machine learning capabilities already to a certain extent. This cloud-based machine learning system will be referred as back end.
The ultimate goal is to be able to determine the amount of berries in a location by a photo created by ordinary mobile device. For this new mobile application a prototype will be produced, which allows us to create testing material with a varying range of mobile phones from different price ranges. A test plan will also be created in the project, which allows us to perform tests on mobile phones and their features. The purpose is to research whether it is possible to achieve valid measurements by using mobile phone cameras. The validity of the source material will also be tested – i.e. whether a single photo or a video source is better alternative for analysis.
Berry harvest observations and forecasts have been going on since 90´s in Finland. The observations have been based on network of monitoring forests, in every forest there are five (5) one (1) square meter observation squares. Flowers, raw berries and ripe berries are calculated during the growing season and forecasts are based on these calculations. Nibio is currently establishing similar approach in Norway. On 2017 Natural Resources Finland (Luke) started to apply citizen science concept in berry observations. In the concept, everyone can establish a monitoring forest with observation squares and make the calculations. The results are stored to marjahavainnot.fi -website, in which the results of observations are freely available. Although the observation procedure is simple and straightforward, calculations are time-consuming and require meticulousness and accuracy. New approaches and techniques are needed for the observations. One possibility is to determine the number of flowers, raw berries and ripe berries by machine vision. In this approach, observer takes a digital picture of the observation square and computer algorithm estimates the number of flowers, raw berries or ripe berries.
Berry machine is being financed by the Regional Council of Lapland. The overall budget is 184 431€ ( IR 20 000€ and public financing 70 551€). The project is being carried out during 18.1.2021 – 30.9.2022.