Designing and Evaluating an Ultrasonic System for Identification of Weed Species

Document Type : Research Article

Authors

• danial gandomzadeh ¹
• Abbas Rohani ¹
• Mohammad Hossein Abbaspour-Fard ¹
• Mahmood Reza Golzarian ²
• A. Mohammadzadeh ³
• ebrahim izadi

¹ Ferdowsi University of Mashhad

² Ferdowsi University Of Mashhad

³ Bahar Institute of Higher Education

10.22067/jpp.v30i2.43278

Abstract

Introduction: Considering the importance of healthy and inexpensive agricultural production, it is necessary to seek ways for precisely discrimination of weeds in the field to minimize the use of herbicides. In this research the feasibility of weed detection due to the reflected ultrasonic waves from some common weeds including Portulacaceae, Chenopodium album L, Tribulus terrestris L, Amaranthus retroflexus L and Salsola iberica, was investigated.
Materials and Methods: An electronic circuit with several parts such as a microcontroller, a power supply (5 DC volts), a RS-232 output port, and an ultrasonic wave generator and detector was constructed. It emits a 40 KHz ultrasonic wave and receives the recursive wave which is reflected from the weed canopy. It can be mounted on an adjustable tripod that is aligned along the three main directions (X, Y, and Z) and can also be turned around the X axis. The data acquisition was accomplished in the research field of the College of Agriculture, Ferdowsi University of Mashhad. The experiments were performed by mounting the system at constant height of 4 cm from the crop canopy. To avoid interfering of the recursive wave with the emitted wave, the generator and the detector were placed far apart. For each experiment the temperature and the relative humidity were recorded in a check list. For the Neural Network the so called BDLRF algorithm was used for training the network and started with a relatively constant large step size of learning rate and momentum term . Before destabilizing the network or when the convergence is slowed down, these values are decreased monotonically (22). In this study Double Sequential Classification Method was used for weed discrimination. This classification method can better simulate the human procedure for classification of different objects, from each other. The human being at the first stage, and based on some distinguishable criteria classifies the things into some main groups and then sorts each group to some other distinguishable subgroups and this procedure will continue up to all things to be classified from each other. Therefore, if a feature can separate more class from others, it is selected as optimum feature. But the optimum feature can only separate the limited numbers of groups in each stage. Other groups are separated with other optimum feature in some sequence stages. In this study the Double Sequential Classification Method is employed for the calcification of the weed species.
Results and Discussion: Results showed that due to different surface and morphological characteristics of the crop canopy of the weed species under study, the pattern and the amplitude of the reflected ultrasonic waves are significantly different. The comparison of means of statistical features extracted from the reflected ultrasonic waves confirmed these differences. A Multi-Layers Perceptron (MLP) neural network, which was trained with a reduction learning rate, was developed and evaluated. The simultaneous separation of the five weed species showed that the error of detection during the training phase of Chenopodium album L was the highest among other species and was 16.67 percent, while the system was able to detect other species completely. However, the detection error for all species in the evaluation phase was more than 50 percent. Hence, a double sequential classification method was used through four sequential stages. In this method 11 neural networks were designed and finally four neural networks were selected. Results showed that Tribulus terrestris L was identified and separated completely from other species in the first stage, subsequently, Amaranthus retroflexus L in the second stage and Portulacaceae in the third stage, respectively. The remaining two species including, Chenopodium album L and Salsola iberica were successfully discriminated in the fourth stage.
Conclusion: Results showed that this method can be a promising technique for real time identification and discrimination of different weed species in the field. It can be replaced with the conventional, laborious and expensive methods to reduce the final costs of agricultural production. Besides, it can reduce the consumption of herbicides in the fields. However, some efforts are required to implement the technique on the existing herbicide applicators or as a new machine for precision agriculture.

Keywords

• Classification
• Electronic circuit
• Neural Network
• Statistics features

20.1001.1.20084749.1395.30.2.4.6

---

References

1- Ahmed F., Al-Mamun H. A., Bari A. S. M. H., Hossain E., and Kwan P. 2012. Classification of crops and weeds from digital images: A support vector machine approach. Crop Protection, 40: 98-104.

2- Alchanatis V., Ridel L., Hetzroni A., and Yaroslavsky L. 2005. Weed detection in multi-spectral images of cotton fields. Computers and Electronics in Agriculture, 47: 243-260.

3- Andújar D., Weis M., and Gerhards R. 2012. An Ultrasonic System for Weed Detection in Cereal Crops. Sensors, 12: 17343-17357.

4- Andujar D., Escol ÀA., Dorado J., and Fernandez-Quintanilla C. 2011. Weed discrimination using ultrasonic sensors. Weed Research, 51: 543-547.

5- Burgos-Artizzu X. P., Ribeiro A., Guijarro M., and Pajares G. 2011. Real-time image processing for crop/weed discrimination in maize fields. Computers and Electronics in Agriculture, 75: 337-346.

6- Burgos-Artizzu X. P., Ribeiro A., Tellaeche A., Pajares G., and Fernandez-Quintanilla C. 2010. Analysis of natural images processing for the extraction of agricultural elements. Image and Vision Computing, 28: 138-149.

7- Escolà A., Andújar D., Dorado J., Fernandez-Quintanilla C., and Rosell-Polo J. R. Weed detection and discrimination in maize fields using ultrasonic and lidar sensors.

8- Fricke T., Richter F., and Wachendorf M. 2011. Assessment of forage mass from grassland swards by height measurement using an ultrasonic sensor. Computers and Electronics in Agriculture, 79: 142-152.

9- Gee C., Bossu J., Jones G., and Truchetet F. 2008. Crop/weed discrimination in perspective agronomic images. Computers and Electronics in Agriculture, 60: 49-59.

10- Harper N., and McKerrow P. 2001. Recognising plants with ultrasonic sensing for mobile robot navigation. Robotics and Autonomous Systems, 34: 71-82.

11- Harper N. L., and McKerrow P. J. 1997. Recognition of Plants with CTFM Ultrasonic Range Data using a Neural Network. International Conference on Robotics and Automation,

12- Henten E. J. v., Asselt C. J. v., Bakker T., Blaauw S. K., Govers M. H. A. M., J.W. Hofstee1, Jansen R. M. C., Nieuwenhuizen A. T., Speetjens S. L., Stigter J. D., Straten G. v., and Willigenburg L. G. V. WURKing: a small sized autonomous robot for the Farm of the Future.

13- Jones G., Gee C., and Truchetet F. 2009. Assessment of an inter-row weed infestation rate on simulated agronomic images. Computers and Electronics in Agriculture, 67: 43-50.

14- Kay L. 1974. A sonar aid to enhance spatial perception of the blind: engineering design and evaluation. Radio and Electronic Engineer, 44: 605-627.

15- Maeyama S., Ohya A., and Yuta S. i. 1994. Positioning by Tree Detection Sensor and Dead Reckoning for Outdoor Navigation of a Mobile Robot. International Conference on Multisensor Fusion and Integration for Intelligent Systems,

16- McKerrow P., and Yoong K. K. 2007. Classifying still faces with ultrasonic sensing. Robotics and Autonomous Systems, 55: 702-710.

17- Mizrach A. 2008. Ultrasonic technology for quality evaluation of fresh fruit and vegetables in pre- and postharvest processes. Postharvest Biology and Technology, 48: 315-330.

18- Montalvo M., Guerrero J. M., Romeo J., Emmi L., Guijarro M., and Pajares G. 2013. Automatic expert system for weeds/crops identification in images from maize fields. Expert Systems with Applications, 40: 75-82.

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

20- Wu X., Xu W., Song Y., and Cai M. 2011. A Detection Method of Weed in Wheat Field on Machine Vision. Procedia Engineering, 15: 1998-2003.

21- Zamahn Q. U., and Salyani M. 2004. Effects of foliage density and ground speed on ultrasonic measurement of citrus tree volume. American Society of Agricultural Engineers,

22- Zarifneshat S., Rohani A., Ghassemzadeh H. R., Sadeghi M., Ahmadi E., and Zarifneshat M. 2012. Predictions of apple bruise volume using artificial neural network. Computers and Electronics in Agriculture, 82: 75-86.