Developing an extreme learning machine based approach to weed segmentation in pastures

• DOI: 10.1016/j.atech.2023.100288
• Publication Date: 2023-07-25T17:47:01Z
• Authors: Jonathan Ford; Edmund Sadgrove; David Paul
• Keywords: Overfitting

Effective weed management in pastures is critical for maintaining the productivity of grazing land. Autonomous ground vehicles (AGVs) are increasingly being considered localization and treatment agricultural Weeds, however, can be difficult to distinguish from background plants, due similarities colour, shape texture. While deep learning approaches used solve issue, they computationally expensive, require a large volume training images order combat overfitting. In this paper we present novel Extreme Learning Machine based network segmenting weeds pasture. The proposed method utilizes combination LBP, HOG colour features, tested on four small datasets, achieving high mean Intersection over Union 87.1, 79.5, 81.6 87.6 Bathurst burr, horehound, thistle serrated tussock respectively.