A Review on A Computer Vision System for Automatic Crop-Weed Detection

Author: Madhusudan B.S., Ramineni Harsha Nag, Prajwal R., Aruna T.N. and Adarsha Gopala Krishna Bhat

Download PDF

Abstract

Weed control is a significant factor that could affect crop productivity. With the advancement in technology, computer vision becomes one of the meticulous methods for instantaneously detecting crop-weeds and providing vital data for spot-specific weed supervision. Computer vision is a technology that employs a computer and a camera, rather than relying on the sensory visuals of an individual, to distinguish, trace, and evaluate the target for a better picture through image processing. This review summarizes the advances and challenges in spot-specific crop-weed detection over the past four years using computer vision technology. The summary of this study discusses conventional methods in weed management, which aid in the development of automatic crop-weed detection for on-field real-time weed control. There are still major challenges for crop-weed classification, such as the overlapping of crop plant foliage and varying illumination levels, leading to the failure of detection algorithms. To achieve universal acceptance of the technology, it is necessary to establish a broader crop dataset. In the upcoming days, through thorough investigation, computer vision techniques will be better applied in autonomous crop-weed detection. With the advancements in computer vision technology, the efficacy and accuracy of crop-weed detection are further enhanced. It also focuses on providing better understanding to laymen for decision support, which aids in the rapid growth of agricultural automation.

Keywords

Image Processing, spot-specific weed supervision, Sensory Visuals, Overlapping, Decision Support

Conclusion

The review provides an overview of the application of computer vision technology in the field of automatic crop-weed detection. Specifically, the paper focuses on summarizing studies that highlight both the advances and challenges in crop-weed segregation for spot-specific weed supervision over the past four years. From the review, it can be concluded that previous work has significantly contributed to the advancement of automatic crop-weed detection, offering the benefits of affordability, high accuracy, and efficiency. However, considering the current scenario, we must also acknowledge the challenges that computer vision technology will encounter in crop-weed detection. Firstly, the task of crop-weed classification in complex scenarios, where crop plant vegetation overlaps and is obstructed by weeds, poses an extremely challenging problem that needs to be addressed promptly. Secondly, variations in illumination levels lead to differences in colors, shadows, noise levels, saturation, reflection, and glare in the same scene, causing detection and segmentation algorithms to fail. Therefore, the need for different color space models to adapt to varying illumination levels is essential. Lastly, in order to achieve universal acceptance of this technology, it is crucial to establish large-scale datasets. In light of the above discussion, it can be inferred that computer vision technology will find more effective applications in autonomous crop-weed detection in the future. With the availability of large-scale datasets, computer vision technology is likely to gain universal acceptance in the field of crop-weed detection. In the upcoming days, as computer vision technology continues to advance, it will enhance the efficacy and accuracy of crop-weed detection, providing valuable insights to agriculturalists for decision support and contributing to the rapid growth of agricultural automation.

References

-

How to cite this article

Madhusudan B.S., Ramineni Harsha Nag, Prajwal R., Aruna T.N. and Adarsha Gopala Krishna Bhat (2023). A Review on A Computer Vision System for Automatic Crop-Weed Detection. Biological Forum – An International Journal, 15(10): 255-262.