A leaf detection method using image sequences and leaf movement

J. Hemming, E.J. van Henten, B.A.J. van Tuijl, J. Bontsema

    Research output: Contribution to journalArticleAcademicpeer-review

    4 Citations (Scopus)


    Besides harvesting the fruits, a very time demanding task is removing old leaves from cucumber and tomato plants grown in greenhouses. To be able to automate this process by a robot, a leaf detection method is required. One possibility for the detection is to exploit the different dynamic behaviour of leaves and stems in case they are excited by mechanical impulse such as an airstream. The objective of this paper is to investigate this technique using a camera and image analysis. For the experiments carried out, images of a tomato crop were used. Images were taken with a digital camera from different distances. To generate an air stream, different radial blowers were used. A sequence of 5 digital images was recorded with a sample rate of 1 second with different settings. To find the differences between the images, the intensity of the images were mathematically subtracted. The resulting binary images were further processed with the aim to extract the leaf positions. These steps included automatic thresh-holding, noise reduction and morphological image analysis operations. It was found that moving objects in the images could be detected using the method described above. Budging objects could only successfully be traced if colour, intensity or structure of the moving object differed with the colour, intensity or structure of the background. The blowing direction had a strong influence on the movements. It is concluded that with this technique not only leaves can be detected but also other parts as e.g. fruits, stems and branches can be discriminated.
    Original languageEnglish
    Pages (from-to)765-772
    JournalActa Horticulturae
    Publication statusPublished - 2005


    Dive into the research topics of 'A leaf detection method using image sequences and leaf movement'. Together they form a unique fingerprint.

    Cite this