Читать книгу Agricultural Informatics - Группа авторов - Страница 40

The continuous image of the field is captured by the camera in every millisecond and gets processed for any unwilling pattern from known pattern. HOG (Histogram Oriented Gradient) image processing strategy is applied here to distinguish between the known and unknown pattern. The known image pattern like crop leaf are treated as training image set which is input to the image processing algorithm. Then by image comparing technique, classifier classifies the image as known or unknown. If any unknown pattern is observed moveable then it will be treated as a weed and unknown pattern is observed not movable then it will be treated as pest.

Figure 2.4 Proposed image processing method to detect pest and weed.

The same image processing method is used in the integrated Agro-IoT system to detect weed and pest and finally store the data in image database. Figure 2.4 illustrates the internal image process method used in the proposed integrated Agro-IoT system.