Hadi Orak; S. Abdanan-Mehdizade
Abstract
One of the objectives of precision agriculture is to minimize the volume of herbicide application by using weed management systems. To achieve this goal, a system based on image processing techniques was developed to detect weeds. In the proposed method, HSV color space was used to discriminate ...
Read More
One of the objectives of precision agriculture is to minimize the volume of herbicide application by using weed management systems. To achieve this goal, a system based on image processing techniques was developed to detect weeds. In the proposed method, HSV color space was used to discriminate between vegetation and background, and between crops and weeds. In this research, the H component of HSV color space along with suitable erosion and dilation were used to automatically segment background (soil) and foreground (weed). According to what was obtained, the algorithm could identify weed from sugar beet plants with the accuracy of 94%. The intelligent sprayer system, in the field trials, in comparison with conventional sprayers (Buferagri sprayer), reduced 67.86% of volume of herbicide due to application of computer vision. The use of this method, as an intelligent sprayer system in sugar beet fields, is recommended.
H. Orak; S. Abdanan-Mehdizadeh
Abstract
A large amount of herbicide is being used for controlling weeds in agricultural, lawns, sport fields on yearly basis. This causes environmental pollution and economic concerns. To reduce the use of herbicides, hand labor may be the best way of removing weeds. It is, however, costly and time consuming. ...
Read More
A large amount of herbicide is being used for controlling weeds in agricultural, lawns, sport fields on yearly basis. This causes environmental pollution and economic concerns. To reduce the use of herbicides, hand labor may be the best way of removing weeds. It is, however, costly and time consuming. In this paper, two methods of distinguishing weeds from lawns using computer vision techniques are proposed. Due to the fact that the weeds have different colors and identification of them from grass is not possible; therefore, an algorithm was developed based on the assumption that the grass area should contain more edges while the weed area is smoother than the grass area. For identification of weed/grass two methods were used, namely: Bayesian Classifier (BO) and morphology (MO. Results indicated that correct weed identification rates for MO and BO methods were 89.58% and 80.42% respectively. Furthermore, from results obtained it can be concluded that herbicide usage was reduced more than 70%, which from economical point of view as well as reduction of environmental pollution is of great importance.