To enable efficient and low-cost automated apple harvesting, this study presented a multi-class instance segmentation model, SCAL (Star-CAA-LADH), which utilizes a single RGB sensor for image acquisition. The model achieves accurate segmentation of fruits, fruit-bearing branches, and main branches using only a single RGB image, providing comprehensive visual inputs for robotic harvesting. A Star-CAA module was proposed by integrating Star operation with a Context-Anchored Attention mechanism (CAA), enhancing directional sensitivity and multi-scale feature perception. The Backbone and Neck networks were equipped with hierarchically structured SCA-T/F modules to improve the fusion of high- and low-level features, resulting in more continuous masks and sharper boundaries. In the Head network, a Segment_LADH module was employed to optimize classification, bounding box regression, and mask generation, thereby improving segmentation accuracy for small and adherent targets. To enhance robustness in adverse weather conditions, a Chain-of-Thought Prompted Adaptive Enhancer (CPA) module was integrated, thereby increasing model resilience in degraded environments. Experimental results demonstrate that SCAL achieves 94.9% AP_M and 95.1% mAP_M, outperforming YOLOv11s by 6.6% and 4.6%, respectively. Under multi-weather testing conditions, the CPA-SCAL variant consistently outperforms other comparison models in accuracy. After INT8 quantization, the model size was reduced to 14.5 MB, with an inference speed of 47.2 frames per second (fps) on the NVIDIA Jetson AGX Xavier. Experiments conducted in simulated orchard environments validate the effectiveness and generalization capabilities of the SCAL model, demonstrating its suitability as an efficient and comprehensive visual solution for intelligent harvesting in complex agricultural settings.

Over the past 30 years, there has been continuous progress in global science and technology. However, many agricultural products still heavily rely on traditional methods of manual and mechanical harvesting, facing challenges such as high costs and low efficiency. To address these challenges, researchers have developed various harvesting robots to handle diverse tasks in complex farm environments. This study analyzed pertinent papers on harvesting robots retrieved from the Web of Science (WOS) core database and the China National Knowledge Infrastructure (CNKI) database, spanning the years 1993 to 2022. Using specialized software such as CiteSpace and VOSviewer, a bibliometric analysis was conducted to examine the research progress and hotspots in the field of harvesting robots. The analysis of 517 English papers indicated a continuous expansion in the research scale of harvesting robots. Furthermore, the research history can be divided into three distinct periods. Currently, research on harvesting robots is experiencing a rapid growth phase, with the number of related papers steadily increasing each year. In the year 2022 alone, 151 English papers were published. This growth is attributed to close collaborations among different countries/regions, institutions, and authors. China, the United States, and Japan play crucial roles in the research of harvesting robots. Notably, China has published 326 English papers, ranking first globally. Through analysis, it was also found that Chinese papers focused on harvesting robots earlier, thereby promoting the development of agricultural robots. Additionally, bibliometric analysis revealed that the research hotspots of harvesting robots mainly include system and structure design, object recognition and localization, and multi-robot coordination, among others. In the future, development trends of harvesting robots will focus on: 1) diversifying robot types, 2) expanding application scenarios,3) enhancing overall performance to reduce losses, and 4) reducing manufacturing costs. In conclusion, through a comprehensive bibliometric analysis, this study has provided valuable insights to advance the automation of harvesting.