[1]王鹏鹰,靳晟,李永可,等.IDSC-YOLOv8-seg:轻量级梯形渠道水尺分割算法[J].计算机技术与发展,2025,(04):127-134.[doi:10.20165/j.cnki.ISSN1673-629X.2024.0363]
　WANG Peng-ying,JIN Sheng,LI Yong-ke,et al.IDSC-YOLOv8-seg:A Lightweight Algorithm for Trapezoidal Channel Water Gauge Segmentation[J].,2025,(04):127-134.[doi:10.20165/j.cnki.ISSN1673-629X.2024.0363]
点击复制

IDSC-YOLOv8-seg:轻量级梯形渠道水尺分割算法

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
期数:
2025年04期
页码:
127-134
栏目:
人工智能
出版日期:
2025-04-10

文章信息/Info

Title:
IDSC-YOLOv8-seg:A Lightweight Algorithm for Trapezoidal Channel Water Gauge Segmentation
文章编号:
1673-629X(2025)04-0127-08
作者:
王鹏鹰1; 2; 靳晟1; 2; 李永可1; 2; 韩博1; 2
1. 新疆农业大学 计算机与信息工程学院,新疆 乌鲁木齐 830052; 2. 新疆农业信息化工程技术研究中心,新疆 乌鲁木齐 830052
Author(s):
WANG Peng-ying1; 2; JIN Sheng1; 2; LI Yong-ke1; 2; HAN Bo1; 2
1. School of Computer and Information Engineering,Xinjiang Agricultural University,Urumqi 830052,China; 2. Xinjiang Agricultural Informatization Engineering Technology Research Centre,Urumqi 830052,China
关键词:
水尺分割; 实例分割; 机器视觉; 轻量化; YOLO算法
Keywords:
water gauge segmentation; instance segmentation; machine vision; lightweight; YOLO algorithm
分类号:
TP391.41
DOI:
10.20165/j.cnki.ISSN1673-629X.2024.0363
摘要:
水尺分割在水位检测极其重要,针对目前梯形渠道水位识别中对复杂环境下的水尺分割精度低,模型计算量大、难以部署等难题,提出一种轻量化的 IDSC-YOLOv8-seg 梯形水尺分割算法。 首先使用轻量级 MobileNet 作为特征提取网络,并使用 GhostConv 卷积模块替换特征融合网络中的 Conv;其次设计了 ISDC-GhostC2f 模块,结合了水尺高长宽比的结构特点,充分利用多尺度深度可分离卷积模块的优势,降低了计算成本,提高了推理速度和效率;再引入高效通道注意力机制(ECA),增强对水尺多尺度细节特征的获取能力,以提升模型对复杂环境下水尺的分割能力,并将 MPDIoU 作为网络损失函数,解决 CIoU 损失函数的局限性,提升了网络收敛速度和精度;最后使用新的数据增强技术,以提高模型的稳定性和泛化性。 结果表明,改进后 IDSC-YOLOv8-seg 算法平均精度均值 mAP@ 0. 5 和mAP@ 0. 5:0. 95 相较于原模型分别提高了 1. 3% 和 0. 9% ,模型的参数量和大小分别降低 46. 6% 和 44. 1% 。 综合说明,改进后的模型在精度满足需求的同时明显降低了参数量和模型大小,为后期水位计算提供技术支撑。
Abstract:
Water gauge segmentation is extremely important in water level detection. Aiming at the current problems of water gauge seg-mentation in trapezoidal channel water level identification in complex environments, such as low accuracy, large amount of model calculation and difficulty in deployment,a lightweight IDSC-YOLOv8-seg trapezoidal water gauge segmentation algorithm is proposed.Firstly,the lightweight MobileNet is used as feature extraction network,and GhostConv convolutional module is used to replace Conv in feature fusion network. Secondly,the ISDC-GhostC2f module is designed,which combines the structural characteristics of the height aspect ratio of the water scale,makes full use of the advantages of the multi-scale depth -separable convolution module,reduces the calculation cost,and improves the reasoning speed and efficiency. The efficient channel attention mechanism (ECA) is introduced to enhance the ability to acquire multi-scale detail features of the water gauge,so as to improve the model's ability to segment the water gauge in complex environments. MPDIoU is used as a network loss function to solve the limitations of CIoU loss function,so as to improve the convergence speed and accuracy of the network. Finally, new data enhancement techniques are used to improve model stability and generalization. The results show that the average accuracy of the improved IDSC-YOLOv8-seg algorithm mAP@ 0. 5 and mAP@ 0. 5:0. 95 are 1. 3% and 0. 9% higher than those of the original model,respectively,and the parameter number and size of the model are reduced by 46. 6% and 44. 1% ,respectively. It is shown that the improved model can meet the requirement of accuracy and reduce the number of parameters and the size of the model obviously,which provides technical support for the later stage calculation.

相似文献/References:

[1]姜世浩,齐苏敏,王来花,等.基于 Mask R-CNN 和多特征融合的实例分割[J].计算机技术与发展,2020,30(09):65.[doi:10. 3969 / j. issn. 1673-629X. 2020. 09. 012]
　JIANG Shi-hao,QI Su-min,WANG Lai-hua,et al.Instance Segmentation Modal Based on Mask R-CNN and Multi-feature Fusion[J].,2020,30(04):65.[doi:10. 3969 / j. issn. 1673-629X. 2020. 09. 012]
[2]张婧婧.基于深度学习和图像处理的马体尺测量设计[J].计算机技术与发展,2020,30(11):180.[doi:10. 3969 / j. issn. 1673-629X. 2020. 11. 033]
　ZHANG Jing-jing.Measurement and Design of Horse爷s Body Size Based on Deep Learning and Image Processing Technology[J].,2020,30(04):180.[doi:10. 3969 / j. issn. 1673-629X. 2020. 11. 033]
[3]曾浩文,汪慧兰*,赵 侃,等.基于 SOLOV2 改进的实例分割算法研究[J].计算机技术与发展,2023,33(09):45.[doi:10. 3969 / j. issn. 1673-629X. 2023. 09. 007]
　ZENG Hao-wen,WANG Hui-lan*,ZHAO Kan,et al.Research on Improved Instance Segmentation Algorithm Based on SOLOV2[J].,2023,33(04):45.[doi:10. 3969 / j. issn. 1673-629X. 2023. 09. 007]
[4]贾丛晨,肖建.动态环境下融合实例分割的视觉SLAM优化方法[J].计算机技术与发展,2024,34(06):89.[doi:10.20165/j.cnki.ISSN1673-629X.2024.0113]
　JIA Cong-chen,XIAO Jian.Optimization of Visual SLAM Algorithm for Fusion Instance Segmentation in Dynamic Environment[J].,2024,34(04):89.[doi:10.20165/j.cnki.ISSN1673-629X.2024.0113]

更新日期/Last Update: 2025-04-10