Abstract:
Abstract - In recent years, machine vision and optical sensors, which can be used in an autonomous weed killing equipment, are being used extensively to detect weeds from crops. In this smdy, seven types of weeds that grow in most of the sugar beet fields in Iran. especially in Fars province, were considered in real outdoor conditions. Several color feature extraction algorithms have also been investigated to separate soil from the plants as well as weeds from the sugar beets. The performance of the proposed algorithm was evaluated by determining correct classi fication rates (CCR) and misclassification rates (MCR) of the results. The findings revealed that the proposed method could successfully detect five of the seven types of the weeds, including Chenopudium album L., Amaranthiis reirofelexus L., Physalis alkekengi L., Convolvulus arvens'.s L., Setaria vertidis L. Beauv and Echiitochloa cms-gali (L.) Beauv.
Machine summary:
COLOR SEGMENTATION SCHEME FOR CLASSIFYING WEEDS FROM SUGAR BEET USING MACHINE VISION A.
Several color feature extraction algorithms have also been investigated to separate soil from the plants as well as weeds from the sugar beets.
The main idea in this work is to segment weeds from sugar beet based on the color features.
The problem with this work is that in sugar beet field both plants and weeds are green objects.
(View the image of this page) Figure 1: Weed segmentation by means of machine vision.
January/ Ju oe, 2006 Iranian Journal of Information Science & Technology, Volume 4, Number 1 (View the image of this page) Figure 6: Light intensity changes in RGB space.
Similar to the procedure used for soil, several parts of sugar beets and weeds from different images and conditions were collected.
In order to eliminate different light intensity January/ June, 2006 Iranian Journal of Information Science & Technology, Volume 4, Number I antl *hadows on leaves, chromaticity components of all images wcre determined.
January/ June, 2006 Iranian Journal of Information Science & Technology, Volume 4, Number I (View the image of this page) Figure 8: a) ma.
It is important for sugar beet plants not to be segmented as weeds.
Iranian Journal of Information Science & Technology, Volume 4, Number 1 January / June, 2006 Table 2: CCR values for different weed species.
Results in Table (2) show that this method can be used for segmentation of seven weed species but it is weak for recognition of Portulaca, because its color is very close to sugar beet.