Image pattern classification for the identification of disease causing agents in plants
Computers and Electronics in Agriculture
Computers and Electronics in Agriculture
International Journal of Information Management: The Journal for Information Professionals
A comparison of methods for multiclass support vector machines
IEEE Transactions on Neural Networks
| Hi-index | 0.00 |
Crop disorders are a serious threat to food security of inhabitants of remote areas in developing countries. While farmers in developed countries have frequently access to various expert resources that help them to identify the onset of a disease, farmers in developing countries usually do not have such support. However, their access to the Internet and thus to the web has rapidly improved during the last few years. This provides a new opportunity to communicate crop pathology information to remote places. We have developed the ''Information system for the assessment of plant disorders'' (Isacrodi) to support farmers in protecting their crop. Farmers are guided to use a controlled but extensible set of attributes to describe the state of their crop. On this basis, Isacrodi provides suggestions which disorders may affect the crop, and which measures would be effective against these disorders. Experts provide Isacrodi with descriptions of actual incidents where they have identified the disorder. Isacrodi uses a computational classifier to provide suggestions to users autonomously. The classifier is constructed based on expert's inputs. Suggestions of disorders and countermeasures are presented as ranked lists, leaving the final identification of the disorder and decisions of countermeasures to the user, as they may have additional information beyond the attributes used by Isacrodi. The performance of the classifier was evaluated by generating data that reflects the envisaged usage of the Isacrodi system. Data on crop disorders provided by experts was used to train the classifier and data that simulated the growers wishing to find out which disorder affects their crop was used to test the classifier. The results show that with limited expert input and errors in data provided by users, the classifier is capable of identifying disorders with reasonable accuracy, particularly when the user considers the three top scoring disorders rather than just the top one. Human experts will attain a much better accuracy than the Isacrodi classifier, particularly when provided with samples from the affected crop. However, where such expertise is not available, Isacrodi can provide valuable support to farmers.