Detection of palm oil bud rot employing artificial vision

D. Montero, C. Rueda

    Producción científica: Contribución a una revistaPonencia publicada en las memorias del evento con ISSNrevisión exhaustiva

    11 Citas (Scopus)

    Resumen

    This paper describes the development of a system capable of capturing, processing, and analysing images of palm oil plantations in order to detect and identify bud rot disease. The process begins with the capturing of images using the DJI Phantom 4 UAV, which is first configured for a flight plan of the desired area and altitude. The resulting images are processed by photogrammetry software to create orthomosaics. The developed algorithm uses the grayscale of the generated orthomosaic and identifies the palms affected by bud rot. This is accomplished using the sliding-window method, by which smaller samples of the image are used and evaluated independently. Each sample is then extracted with a ULBP feature vector that numerically represents the texture of the image and is classified by means of a previously trained logistic regression model, allowing the recognition of positive cases of the disease. Possible positive cases are further distinguished using the non-maximum suppression algorithm. The system was tested with different images than the images used for training and for establishing the set point. As a result, the system showed a 92% precision and a 96% of sensitivity for bud rot disease detection. These results are satisfactory in terms of detecting bud rot disease using a low-cost system.

    Idioma originalInglés
    Número de artículo012004
    PublicaciónIOP Conference Series: Materials Science and Engineering
    Volumen437
    N.º1
    DOI
    EstadoPublicada - 25 oct. 2018
    Evento3rd International Congress of Mechanical Engineering and Agricultural Science, CIIMCA 2017 - Bucaramanga, Colombia
    Duración: 2 oct. 20176 oct. 2017

    Nota bibliográfica

    Publisher Copyright:
    © Published under licence by IOP Publishing Ltd.

    Huella

    Profundice en los temas de investigación de 'Detection of palm oil bud rot employing artificial vision'. En conjunto forman una huella única.

    Citar esto