This review provides an overview of apple detection and localization using image analysis and artificial intelligence techniques for enabling robotic fruit harvesting in orchard environments. Classic methods for detecting and localizing infield apples are discussed along with more advanced approaches using deep learning algorithms that have emerged in the past few years. Challenges faced in apple detection and localization such as occlusions, varying illumination conditions, and clustered apples are highlighted, as well as the impact of environmental factors such as light changes on the performance of these algorithms. Potential future research perspectives are identified through a comprehensive literature analysis. These include combining cutting-edge deep learning and multi-vision and multi-modal sensors to potentially apply them in real-time for apple harvesting robots. Additionally, utilizing 3D vision for a thorough analysis of complex and dynamic orchard environments, and precise determination of fruit locations using point cloud data and depth information are presented. The outcome of this review paper will assist researchers and engineers in the development of advanced detection and localization mechanisms for infield apples. The anticipated result is the facilitation of progress toward commercial apple harvest robots.

中文翻译：

---

用于苹果检测和定位的图像处理和人工智能：综合综述


这篇综述概述了使用图像分析和人工智能技术在果园环境中实现机器人水果收获的苹果检测和定位。讨论了检测和定位内田苹果的经典方法，以及使用过去几年出现的深度学习算法的更高级方法。重点介绍了苹果检测和定位面临的挑战，例如遮挡、不同的照明条件和聚集的苹果，以及环境因素（如光线变化）对这些算法性能的影响。通过全面的文献分析确定潜在的未来研究前景。其中包括将尖端的深度学习与多视觉和多模态传感器相结合，以潜在地将它们实时应用于苹果收获机器人。此外，利用 3D 视觉对复杂和动态的果园环境进行全面分析，并使用点云数据和深度信息精确确定水果位置。本综述论文的结果将帮助研究人员和工程师开发先进的内田苹果检测和定位机制。预期的结果是促进了商用苹果收获机器人的进展。