Application of generalized Hough transform for detecting sugar beet plant from weed using machine vision method

Document Type : Research Article

Authors

• A. Bakhshipour Ziaratgahi ¹
• A. A. Jafari ¹
• Y. Emam ²
• S. M. Nassiri ¹
• S. Kamgar ¹
• D. Zare ¹

¹ Department of Biosystems Engineering, Shiraz University, Shiraz, Iran

² Department of Crop Production and Plant Breeding, Shiraz University, Shiraz, Iran

https://doi.org/10.22067/jam.v7i1.49959

Abstract

Introduction
Sugar beet (Beta vulgaris L.) as the second most important world’s sugar source after sugarcane is one of the major industrial crops. The presence of weeds in sugar beet fields, especially at early growth stages, results in a substantial decrease in the crop yield. It is very important to efficiently eliminate weeds at early growing stages. The first step of precision weed control is accurate detection of weeds location in the field. This operation can be performed by machine vision techniques.
Hough transform is one of the shape feature extraction methods for object tracking in image processing which is basically used to identify lines or other geometrical shapes in an image. Generalized Hough transform (GHT) is a modified version of the Hough transform used not only for geometrical forms, but also for detecting any arbitrary shape. This method is based on a pattern matching principle that uses a set of vectors of feature points (usually object edge points) to a reference point to construct a pattern. By comparing this pattern with a set pattern, the desired shape is detected. The aim of this study was to identify the sugar beet plant from some common weeds in a field using the GHT.
Materials and Methods
Images required for this study were taken at the four-leaf stage of sugar beet as the beginning of the critical period of weed control. A shelter was used to avoid direct sunlight and prevent leaf shadows on each other. The obtained images were then introduced to the Image Processing Toolbox of MATLAB programming software for further processing.
Green and Red color components were extracted from primary RGB images. In the first step, binary images were obtained by applying the optimal threshold on the G-R images.
A comprehensive study of several sugar beet images revealed that there is a unique feature in sugar beet leaves which makes them differentiable from the weeds. The feature observed in all sugar beet plants at the four-leaf stage was a stretched S-shaped curve at the junction of the leaf and petiole. This unique shape characteristic was used as the pattern for sugar beet detection using GHT. To implement the Hough transform in the images, a 50-member group of samples was prepared from S-shaped curve to build appropriate patterns. Desired features for the Hough transformation were extracted from the patterns. In the next step, the attempts were made to find the images for the shapes similar to each of the patterns.
Results and Discussion
Plants were thoroughly separated from soil and residues. The accuracy of segmentation algorithm was achieved by almost 100%.
The accuracy of the generalized Hough algorithm was evaluated in two stages. In the first stage, the algorithm accuracy was assessed in detecting patterns in the images. Results showed that the accuracy of the algorithm was 96.21%. In the second stage, the algorithm was evaluated for some other test images, whereas the algorithm achieved an overall accuracy of 91.65%. In some cases, the presence of a large overlap between objects in the image reduced the detection accuracy. This was because of two main reasons; 1) high interference and ambiguity in the object edges, so that Hough transform was not able to detect the predefined patterns in the objects and, 2) weeds highly overlapped with sugar beet plants and thereby they were wrongly detected as sugar beet. However, since there is no or little interference between plants at the four-leaf stage, this interference can be eliminated by morphological operations. Due to this fact, it can be said that the results of GHT algorithm are acceptable for the detection of sugar beet in the plants close to four-leaf stage.
Conclusion
A special feature in the shape of sugar beet leaves was used as a criterion to distinguish between sugar beet and weeds. The results showed that by quantifying this special feature, which is an S-shaped curve near the petioles connection of beet leaves, sugar beet can be discriminated from weeds with an accuracy of 91.65 %. Recalled that this feature is a shape characteristic, therefore, the generalized Hough algorithm must be applied prior to plant canopy development, which is consistent with the critical period of weed control in sugar beet fields.

Keywords

• Generalized Hough
• Shape processing
• Sugarbeet
• Visible machine vision
• Weed

20.1001.1.22286829.1396.7.1.7.2

References

1. Ahmed, F., H. A. Al-Mamun, A. S. M. Hossain Bari, E. Hossain, and P. Kwan. 2012. Classification of crops and weeds from digital images: A support vector machine approach. Crop Protection 40: 90-104.

2. Arribas, J. I., G. V. Sanchez-Ferrero, G. Ruiz-Ruiz, and J. Gomez-Gil. 2011. Leaf classification in sunflower crops by computer vision and neural networks. Computers and Electronics in Agriculture 78 (1): 9-18.

3. Ballard, D. H. 1981. Generalizing the Hough Transform to Detect Arbitrary Shapes. Pattern Recognition 13 (2): 111-122.

4. Blasco, J., N. Aleixos, J. Roger, E. Rabatel, and E. Molto. 2002. Robotic weed control using machine vision. Biosystems Engineering 83 (2): 149-157.

5. Cooke, D. A., and R. K. Scott. 1993. The Sugar Beet Crop. Chapman and Hall, Publishers. pp: 675.

6. Cussans, G. W. 1986. The potential for integrated weed management (IWM) control. 49th winter congress. International Institute for Sugar Beet Research 253-262.

7. Gee, C., J. Bossu, G. Jones, and F. Truchetet. 2008. Crop/weed Discrimination in Perspective Agronomic Images. Computers and Electronics in Agriculture 60 (1):49-59.

8. Ghadiri, H. 1996. Concept and application of critical period of weed control. Collections of full papers of 4th Iranian crop production and breeding congress Isfahan 257-265.

9. Gillott, I. 2001. Critical herbicide uses in minor crops- an agronomist’s view. Proceedings of the BCPC Conference-Weeds, Brighton, UK, 799-802.

10. Hakimi, M., and J. Gohari. 1993. Determination of the most suitable row distance in sugar beet cultivation. Publication of Iran sugarbeet seed institute.

11. Hemming, J., and T. Rath. 2001. Computer vision-based weed identification under field conditions using controlled lighting. Journal of Agricultural Engineering Research 78 (3): 233-243.

12. Jafari, A. 2005. Developing a Suitable Algorithm for Weeds Segmentation from Sugar Beet Crop Using Machine Vision and Artificial Neural Networks. PhD Thesis, Department of Agricultural Machinery, Faculty of Agriculture, Tehran University.

13. Jafari, A., S. S. Mohtasebi, H. E. Jahromi, and M. Omid. 2006. Weed detection in sugar beet fields using machine vision. International Journal of Agriculture and Biology 8 (5): 602-605.

14. Jahadakbar, M. R., R. Tabatabai, and H. R. Ebrahimian. 2004. Critical period of weed competition with sugar beet in Kabotarabad-Esfahan. Journal of Sugar Beet 20 (1): 73-92.

15. Kavdır, I. 2004. Discrimination of sunflower, weed and soil by artificial neural networks. Computers and Electronics in Agriculture 44 (2): 153-160.

16. Kaya, R., and S. Buzluk. 2006. Integrated weed control in sugar beet through combinations of tractor hoeing and reduced dosage of herbicide mixture. Turkish Journal of Agriculture and Forestry 30: 137-144.

17. Kolivand, M. 1995. Study of sugar beet growth pattern in Kermanshah. Journal of Sugarbeet 11 (1): 1-19.

18. Leemans, V., and M. F. Destain. 2006. Application of the Hough Transform for Seed Row Localization using Machine Vision. Biosystems Engineering 94 (3): 325-336.

19. Morishita, D. W., and M. J. Wille. 2001. Broadleaf weed control in sugar beet with soilapplied and sequential post emergence herbicides compared to micro herbicide rates. Available from: www.uidaho.edu/sugar beet/weed/00-12.htm.

20. Moshashai, K., M. Almasi, S. Minaei, and A. M. Borghei. 2008. Identification of sugarcane nodes using image processing and machine vision technology. International Journal of Agricultural Researches 3: 357-364.

21. Perez, A. J., F. Lopez, J. V. Benlloch, and S. Christensen. 2000. Colour and shape analysis techniques for weed detection in cereal fields. Computers and Electronics in Agriculture 25: 197-212.

22. Shahbazi, H. A., and M. Abdollahian-Noghabi. 2000. Critical period of weed competition in sugar beet in Mashhad, 16 (1): 58-74.

23. Shapiro, L. G., and G. C. Stockman. 2001. Computer Vision, Prentice-Hall Inc., Upper. Saddle River, New Jersey, pp. 41.

24. Sogaard, H. T. 2005. Weed Classification by Active Shape Models. Biosystems Engineering 91 (3): 271-281.

25. Sonka, M., V. Hlavac, and R. Boyle. 1993. Image Processing, Analysis, and Machine Vision. Brooks/Cole Publishing Company.

26. Tellaeche, A., X. P. BurgosArtizzu, G. Pajares, A. Ribeiro, and C. Fernandez-Quintanilla. 2008. A new vision-based approach to differential spraying in precision agriculture. Computers and Electronics in Agriculture 60 (2): 144-155.

27. Terawaki, M., T. Kataoka, H. Okamoto, and S. Hata. 2002. Distinction between sugar beet and weeds for development of automatic thinner and weeding machine of sugar beet. Proceeding of the Automation Technology for Off-Road Equipment Conference (Chicago, Illinois, USA).