Answers are grounded in this paper and cite their sources.
Background Goji berry is an important specialty crop in Ningxia, China, where large-scale cultivation has increased the demand for efficient orchard management. However, the intelligence level of weeding equipment in goji orchards remains limited, particularly in the basal region of the plant, where improper operation may cause root damage and incomplete weed removal. Methods To address this problem, this study developed a DeepLabV3+-based perception framework for goji berry basal-region recognition in orchard environments. An autonomous mobile testing platform was established according to the planting pattern of the Bairuiyuan plantation in the Helan Mountains, and a binocular camera was used for image acquisition. Based on the collected orchard images, a semantic segmentation model was constructed to identify the basal regions of goji berry plants. The proposed framework was evaluated under different weed conditions through static image-based testing and continuous row-level field testing. Results In static image-based testing, the group-level missed detection rate and false detection rate were 12.5% ± 7.6% and 2.2% ± 1.9%, respectively. In continuous row-level field testing with the mobile platform, the corresponding rates were 2.0% ± 2.0% and 2.7% ± 2.7%, respectively. Conclusions The results indicate that the proposed framework can provide stable basal-region perception during continuous field operation, despite variations in weed conditions. This study provides practical support for the development of intelligent intra-row weeding systems in goji orchards.
No takes yet. Share an insight, caveat, or question.
Li et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: