DBPapers
DOI: 10.5593/sgem2017/23/S11.080

RESEARCH INTO THE USE OF IMAGE PROCESSING AND AUTOMATIC MAPPING OF WEEDS WITHIN GIS SYSTEMS

M. Gidea, C. Constantinescu, A. Trif, A.Boasca, M. Burcea
Thursday 14 September 2017 by Libadmin2017

References: 17th International Multidisciplinary Scientific GeoConference SGEM 2017, www.sgem.org, SGEM2017 Conference Proceedings, ISBN 978-619-7408-03-4 / ISSN 1314-2704, 29 June - 5 July, 2017, Vol. 17, Issue 23, 641-648 pp, DOI: 10.5593/sgem2017/23/S11.080

ABSTRACT

With a cultivated area of over 2x108ha, wheat is the largest crop worldwide by cultivated surface. For wheat crops, the production suffers loss caused by weeds, and depending on the weeding level, the fertilization level can surpass 50% (1). Weed management is just one of the technological components present in any crop technology. Given the biological particularities, wheat crop is mainly weeded with dicotyledonous species. The objective of this paper is the comparative evaluation of the classic method for determining the weeding, and the automatic mapping method which uses a mobile platform to acquire geo-referential images. Through the classic method of mapping weeds, the weeding spectrum was determined – characterized by the average level of weeding, participation of each species and the weeding constant. Regarding the automatic evaluation of wheat crop weeding level, a mobile platform equipped with a video cam was set up, which works independent of the moving speed, at a frequency of 1 picture/meter. The obtained images are stored on an HDD, in image+ GPS-RTK coordinates packages. The discrimination of dicotyledonous weeds is performed through direct processing and analysis of images obtained by the video cam. The primary factor of image processing is applying color filters so that the image content would be clean of the background and objects not depicting plants. Following this filtering, the useful scene is obtained. Recognizing the shapes and discriminating the plants is performed afterwards by determining the shape/cut-outs and identifying the types of plants through comparing determinations, in real time, on image analysis of leaf seizes with sets of shape vectors (reference vectors) which were previously stored. On the basis of the reports drafted, the weeded surface can be estimated, with dicotyledonous species compared to the total analyzed surface. The equipment has been tested in field conditions. From the result, it can be noticed that although the weeding shows a uniform distribution, the surface covered only by dicotyledonous was 22.2% of the total analyzed surface. Regarding the differences between the two methods, they were not ensured statistically. In parallel, the results of the automatic image processing were performed manually, with a validation index of 88.3%. Following our research, it was established that the equipment can be integrated in a distribution system with differentiated sprayer ON / OFF, in real time, depending on the presence or absence of dicotyledonous species. The herbicides / distribution precision - reported to the area covered with dicotyledonous species, depends in this case on the print of the herbicides’ jet sprayed at soil, thus correlating with the type and particularities of the nozzles used.

Keywords: wheat, weed, image processing, GPS, orthophoto