摘要
由一名新闻记者-机器人与机器学习的工作人员新闻编辑每日新闻-关于机器人的最新研究结果已经发表。据《盐城消息》报道，NewsRx记者称，“智能苹果机器人可以显著提高苹果采摘的效率，实现苹果快速准确的识别和定位是采摘机器人操作的前提和基础。现有的苹果识别和定位方法主要集中在目标检测和语义识别技术上。”新闻记者从盐城师范大学的研究中得到一句话:“然而,这些方法在面对遮挡和重叠问题时往往存在定位错误,而且少数实例分割方法效率低,严重依赖于检测结果。”摘要:提出了一种基于RG B-d的苹果识别定位方法和一种改进的SOLOv2实例分割方法。为了提高实例分割网络的效率,采用EfficientNetV2作为特征提取网络,该网络具有较高的参数效率。为了提高苹果被遮挡或重叠时的分割精度,摘要:提出了一种轻量级的sp注意模块。该模块提高了模型的位置敏感性,使得位置特征能够在语义特征相似的情况下区分重叠对象。为了准确地确定apple-pi的关键点,提出了一种基于rgb-d的apple定位方法。通过对比实验分析,改进的SOLOv2实例分割方法表现出了显著的性能。与SOLOv2相比,F1评分、mAP、Apple实例分割数据集上的mIoU和mIoU分别提高了2.4%、3.6%和3.8%,Params和FLOP s分别降低了1.94M和31 Gflop。
Abstract
By a News Reporter-Staff News Editor at Robotics & Machine Learning Daily News Daily News-Current study results on robotics have been published. According to news originating from Yancheng, People's Republic of China, by NewsRx correspondents, research stated, "Intelligent applepicking robots can significantly improve the efficiency of apple picking, and the realiz ation of fast and accurate recognition and localization of apples is the prerequ isite and foundation for the operation of picking robots. Existing apple recogni tion and localization methods primarily focus on object detection and semantic s egmentation techniques." The news reporters obtained a quote from the research from Yancheng Teachers Uni versity: "However, these methods often suffer from localization errors when faci ng occlusion and overlapping issues. Furthermore, the few instance segmentation methods are also inefficient and heavily dependent on detection results. Therefo re, this paper proposes an apple recognition and localization method based on RG B-D and an improved SOLOv2 instance segmentation approach. To improve the effici ency of the instance segmentation network, the EfficientNetV2 is employed as the feature extraction network, known for its high parameter efficiency. To enhance segmentation accuracy when apples are occluded or overlapping, a lightweight sp atial attention module is proposed. This module improves the model position sens itivity so that positional features can differentiate between overlapping object s when their semantic features are similar. To accurately determine the apple-pi cking points, an RGB-D-based apple localization method is introduced. Through co mparative experimental analysis, the improved SOLOv2 instance segmentation metho d has demonstrated remarkable performance. Compared to SOLOv2, the F1 score, mAP , and mIoU on the apple instance segmentation dataset have increased by 2.4, 3.6 , and 3.8%, respectively. Additionally, the model's Params and FLOP s have decreased by 1.94M and 31 GFLOPs, respectively."
NETL
NSTL