Accurate Fruit and Vegetable Detection Method for Space Station Cargo Bay
-
摘要: 机器视觉技术在空间站货舱管理的实际应用方面处于初步阶段，为了解决空间站货舱的狭小空间、遮挡和光照等问题导致的检测精度不足，提出一种基于YOLO11的空间站货舱果蔬检测改进算法：LEBR-YOLO。该方法把卷积改进为结合空间信息和边缘信息的高效输入特征提取干层，同时添加双层注意力机制，提高了提取特征的能力。引入改进的轻量级共享可变形检测模块，提高了遮挡情况下的检测能力。使用迁移学习作为优化模型的方法，弥补数据集的不足，提高泛化能力。实验表明，该方法在自制果蔬类数据集上达到了95.3%的准确率，88.6%召回率和93.9%的mAP@0.5，同时依然保持较低的模型复杂度，满足轻舟货运飞船在轨运行的需要。该方法可以有效的用于空间站水果蔬菜类物品检测，提高了检测精度，有效减少了误检、漏检。
Abstract:- The practical application of machine vision technology in the management of space station cargo bay is in the preliminary stage, in order to solve the lack of detection accuracy due to the narrow space, occlusion and light problems in the space station cargo bay, an improved algorithm for the detection of fruits and vegetables in the space station cargo bay based on YOLO11 is proposed: LEBR-YOLO. the method improves convolution to an efficient input feature extraction stem layer combining spatial information and edge information. A two-layer attention mechanism is added to improve the feature extraction capability. An improved lightweight shared deformable detection module is introduced to improve the detection ability under occlusion. Migration learning is also used as a method to optimize the model to compensate for the lack of dataset and improve the generalization ability. Experiments show that the method achieves 95.3% accuracy, 88.6% recall and 93.9% mAP@0.5 on the homemade fruit and vegetable dataset, while still maintaining a low model complexity to meet the needs of the Lightship cargo spacecraft in orbit. The method can be effectively used for the detection of fruit and vegetable items in the space station, which improves the detection accuracy and effectively reduces misdetection and omission.
-
-
计量
- 文章访问数: 12
- HTML全文浏览量: 1
- PDF下载量: 1
-
被引次数:
0(来源:Crossref)
0(来源:其他)
下载:
