Deep Learning-Based Weed Detection and Removal Robot Using Raspberry Pi for Smart Agriculture

Abstract

Weed infestation in agricultural fields significantly reduces crop productivity by competing for essential resources such as nutrients, water, and sunlight. Traditional weed control methods, which includes manual weeding and chemical herbicide spraying, are becoming less effective due to increasing labour costs, environmental concerns, and inconsistent efficiency in large-scale farming operations. This research presents the design and development of an autonomous Deep Learning-Based Weed Detection and Removal Robot using Raspberry Pi for smart agriculture applications. The proposed system combines computer vision, deep learning, ultrasonic sensing, and a mechanical weed removal mechanism to enable precise and chemical-free weed management. The robot autonomously navigates through agricultural fields using a DC motor-driven chassis. An HC-SR04 ultrasonic sensor continuously monitors the path ahead, and whenever a plant or obstacle is detected within a predefined distance of 20 cm, the robot stops automatically. A Raspberry Pi Camera Module captures an image of the detected plant, which is then processed using a MobileNetV2-based Convolutional Neural Network (CNN) model deployed on Raspberry Pi. The trained model classifies the captured image into two categories: crop or weed. If the detected plant is identified as a crop, the robot resumes movement without any action. However, if the plant is classified as a weed, a motor-controlled cutter mechanism is activated to physically remove the weed without affecting nearby crops. A dataset containing 4,200 labelled crop and weed images was used for training and validation. The proposed deep learning model achieved a training accuracy of 98.3% and a validation accuracy of 96.7%, demonstrating reliable classification performance under varying field conditions. Experimental results indicate that the robot reduces weed removal time by approximately 73% compared to conventional manual methods while completely eliminating the use of chemical herbicides and manual labour. The system provides a low-cost, energy-efficient, and environmentally sustainable solution for precision farming and smart agricultural automation.

Introduction

Agriculture is a vital sector in India, but weed infestation remains a major challenge, causing crop yield losses of 20–80%. Traditional weed control methods, such as manual weeding and chemical herbicides, are either labor-intensive or environmentally harmful, leading to soil degradation, water contamination, and herbicide-resistant weeds. Recent advances in artificial intelligence (AI), deep learning, computer vision, and embedded systems have enabled the development of intelligent, low-cost agricultural robots for precision weed management.

This study proposes an autonomous weed detection and removal robot that combines an HC-SR04 ultrasonic sensor, Raspberry Pi Camera Module, Raspberry Pi 4B, and a MobileNetV2 deep learning model to distinguish crops from weeds in real time. Once a plant is detected, the robot captures its image, classifies it onboard using TensorFlow Lite without cloud connectivity, and activates a motor-driven cutter to remove weeds while leaving crops untouched. The system follows a sense–classify–act workflow and uses a confidence threshold of 0.85 to minimize incorrect actions.

The hardware consists of a Raspberry Pi, camera, ultrasonic sensor, L298N motor driver, DC motors, rechargeable battery, and a four-wheel chassis. The software is implemented in Python 3.9 using libraries such as OpenCV, TensorFlow, TensorFlow Lite, NumPy, Pandas, and scikit-learn. A balanced dataset of 4,200 crop and weed images was created from PlantVillage, Weed-AI, field images, and augmented data. Images were preprocessed through resizing, normalization, and augmentation techniques such as rotation, flipping, zooming, and brightness adjustments to improve model robustness.

The proposed robot provides an affordable, eco-friendly alternative to chemical herbicides by combining high-accuracy deep learning, edge computing, and mechanical weed removal. It is specifically designed for smallholder farmers, offering reduced labor requirements, real-time autonomous operation, and improved precision agriculture while promoting sustainable farming practices.

Conclusion

The proposed weed detection and removal robot successfully combines deep learning and robotics for smart farming applications. The system uses a MobileNetV2 model to identify crops and weeds and a cutter mechanism to remove weeds automatically. Experimental results showed that the model achieved a validation accuracy of 96.7%, demonstrating its ability to classify plants reliably under different conditions. The robot helps reduce manual labor, saves time, and eliminates the need for chemical herbicides. The use of low-cost hardware such as Raspberry Pi makes the system affordable and suitable for small and medium-scale farms. Overall, the project demonstrates that artificial intelligence can be effectively applied in agriculture to improve productivity and support sustainable farming practices. With further improvements in detection capability, navigation, and battery performance, the system has strong potential for future agricultural applications.

References

[1] Y. Gharde, P. K. Singh, R. P. Dubey, and P. K. Gupta, \"Assessment of Yield and Economic Losses in Agriculture Due to Weeds in India,\" Crop Protection, vol. 107, pp. 12–18, May 2018. DOI: 10.1016/j.cropro.2018.01.007 [2] K. Jabran, G. Mahajan, V. Sardana, and B. S. Chauhan, \"Allelopathy for Weed Control in Agricultural Systems,\" Crop Protection, vol. 72, pp. 57–65, Jun. 2015. DOI: 10.1016/j.cropro.2015.03.004 (covers herbicide-resistance, ecological harm, and need for alternatives) [3] A. Krizhevsky, I. Sutskever, and G. E. Hinton, \"ImageNet Classification with Deep Convolutional Neural Networks,\" Advances in Neural Information Processing Systems (NeurIPS), vol. 25, pp. 1097–1105, 2012. [4] K. L. Joshitha, S. Nishanthini, B. J. Harshini, and T. Sadhana, \"Raspberry Pi Based Crop and Weed Classification Using Machine Learning Algorithm,\" IOP Conference Series: Earth and Environmental Science, vol. 980, article 012029, 2022. DOI: 10.1088/1755-1315/980/1/012029 [5] X. Fan, X. Chai, J. Zhou, and T. Sun, \"Deep Learning Based Weed Detection and Target Spraying Robot System at Seedling Stage of Cotton Field,\" Computers and Electronics in Agriculture, vol. 214, article 108350, Nov. 2023. DOI: 10.1016/j.compag.2023.108350 [6] S. P. Mohanty, D. P. Hughes, and M. Salathé, \"Using Deep Learning for Image-Based Plant Disease Detection,\" Frontiers in Plant Science, vol. 7, article 1419, Sep. 2016. DOI: 10.3389/fpls.2016.01419 [7] A. Kamilaris and F. X. Prenafeta-Boldú, \"Deep Learning in Agriculture: A Survey,\" Computers and Electronics in Agriculture, vol. 147, pp. 70–90, Apr. 2018. DOI: 10.1016/j.compag.2018.02.016 [8] K. He, X. Zhang, S. Ren, and J. Sun, \"Deep Residual Learning for Image Recognition,\" in Proc. IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, USA, Jun. 2016, pp. 770–778. DOI: 10.1109/CVPR.2016.90 [9] A. Olsen, D. A. Konovalov, B. Philippa, P. Ridd, J. C. Wood, J. Johns, W. Banks, B. Girgenti, O. Kenny, J. Whinney, B. Calvert, M. Rahimi Azghadi, and Q. N. White, \"DeepWeeds: A Multiclass Weed Species Image Dataset for Deep Learning,\" Scientific Reports, vol. 9, article 2058, Feb. 2019. DOI: 10.1038/s41598-018-38343-3 [10] M. Sandler, A. Howard, M. Zhu, A. Zhmoginov, and L.-C. Chen, \"MobileNetV2: Inverted Residuals and Linear Bottlenecks,\" in Proc. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Salt Lake City, UT, USA, Jun. 2018, pp. 4510–4520. DOI: 10.1109/CVPR.2018.00474 [11] D. C. Slaughter, D. K. Giles, and D. Downey, \"Autonomous Robotic Weed Control Systems: A Review,\" Computers and Electronics in Agriculture, vol. 61, no. 1, pp. 63–78, Apr. 2008. DOI: 10.1016/j.compag.2007.05.008 [12] I. Sa, Z. Chen, M. Popovic, R. Khanna, F. Liebisch, J. Nieto, and R. Siegwart, \"weedNet: Dense Semantic Weed Classification Using Multispectral Images and MAV for Smart Farming,\" IEEE Robotics and Automation Letters, vol. 3, no. 1, pp. 588–595, Jan. 2018. DOI: 10.1109/LRA.2017.2774979 [13] F. Dang, D. Chen, Y. Lu, and Z. Li, \"YOLOWeeds: A Novel Benchmark of YOLO Object Detectors for Multi-Class Weed Detection in Cotton Production Systems,\" Computers and Electronics in Agriculture, vol. 205, article 107655, Feb. 2023. DOI: 10.1016/j.compag.2023.107655 [14] T. Duckett, S. Pearson, S. Blackmore, and B. Grieve, \"Agricultural Robotics: The Future of Robotic Agriculture,\" UK-RAS Network White Paper, Sheffield, UK, 2018. [Online]. Available: https://arxiv.org/abs/1806.06762 [15] G. Olsen, G. Goswami, and M. Haugen, \"A System for Weeds and Crops Identification — Reaching over 10 FPS on Raspberry Pi with the Usage of MobileNets, DenseNet and Custom Modifications,\" Sensors, vol. 19, no. 18, article 4051, Sep. 2019. DOI: 10.3390/s19184051 [16] N. Bah, A. Hafiane, and R. Canals, \"Deep Learning with Unsupervised Data Augmentation for Weed Detection in Line Crops,\" IEEE Access, vol. 8, pp. 69448–69468, 2020. DOI: 10.1109/ACCESS.2020.2986399 [17] R. Gerhards and S. Christensen, \"Real-Time Weed Detection, Decision Making and Patch Spraying in Maize, Sugarbeet, Winter Wheat and Winter Barley,\" Weed Research, vol. 43, no. 6, pp. 385–393, Dec. 2003. DOI: 10.1046/j.1365-3180.2003.00349.x [18] Y. LeCun, Y. Bengio, and G. Hinton, \"Deep Learning,\" Nature, vol. 521, no. 7553, pp. 436–444, May 2015. DOI: 10.1038/nature14539 [19] A. Howard, M. Sandler, B. Chen, W. Wang, L.-C. Chen, M. Tan, G. Chu, V. Vasudevan, Y. Zhu, R. Pang, H. Adam, and Q. Le, \"Searching for MobileNetV3,\" in Proc. IEEE/CVF International Conference on Computer Vision (ICCV), Seoul, South Korea, Oct. 2019, pp. 1314–1324. DOI: 10.1109/ICCV.2019.00140 [20] E.-C. Oerke, \"Crop Losses to Pests,\" Journal of Agricultural Science, vol. 144, no. 1, pp. 31–43, Feb. 2006. DOI: 10.1017/S0021859605005708

Copyright

Copyright © 2026 Vaibhav Prasad, Preet Sambare , Rohit Sharma, Mr. Vinayak Bembrekar. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.