Modern approaches to image analysis in plant disease identification systems with the use of fast detector

Abstract

In terms of doing the exact economy with the aim of ensuring the effectiveness of management decisions, frantic competition and reduction of fertile areas, the question arises about the development of an intelligent system of management of agrarian sector. The greatest danger to the farmers, today, are the risks plant diseases, which greatly reduce the yield. Therefore, farmers are forced to find new means of identification of plant diseases.

The use of FAST detector in pattern recognition of plant diseases, can be implemented in combination with the detectors of colors, allowing you to select the area of the disease, which can then be the subject of research. First of all, it is connected with possibilities of errors, not enough clear photos and unstable lighting. However, the definition of singular points in the image are not completely solves the problem of identification of plant diseases, as in the noise they can also be incorrect and insufficiently identified. So it all depends on conditions, diagnostics, and pattern recognition of the disease. The vast majority of examples are based on the set of singular points, but research conducted in the field of machine learning based on neural networks. Of course, that the respective tasks require significant computing power, in terms of keeping accurate farming, which is implemented in the field to provide quite difficult. Thus, it is necessary simple and time efficient solution which will provide use cheap tools which are implemented the algorithms of pattern recognition.

The use of FAST detector showed the possibility to install a significant number of singular points in the image, with the result that certain patterns are defined between the descriptors of such points, however, the accuracy and practical application of this method in real conditions require further study.