Surface fault location of conveyor belt based on saliency and deep convolution neural network
-
摘要: 提出了一种基于显著性和深度卷积神经网络的输送带表面故障定位方法。该方法在输送带上、下表面的边缘烙上数字，利用图像处理技术检测输送带图像中的数字来间接定位输送带表面故障。首先，将采集的输送带图像进行高斯滤波、灰度线性变换等预处理，以提高图像质量、增强背景与目标的对比度；然后根据谱残差理论对预处理后的图像进行视觉显著性处理，获取包含数字区域的视觉显著图；最后，运用卷积神经网络对显著图进行分类，以区分出数字区域和非数字区域。实验结果表明，该方法可以很好地检测输送带图像上的数字，进而实现对输送带表面故障的定位。Abstract: A surface fault location of conveyor belt based on saliency and deep convolution neural network was proposed. The method imprints figures on the edge of upper and lower surfaces of conveyor belt, and uses image processing technology to detect the number in belt image, so as to indirectly locate surface fault of the conveyor belt. Firstly, the acquired image of the conveyor belt is preprocessed by Gaussian filtering and gray-scale linear transformation to improve image quality and enhance contrast between the background and the target. Then, visual saliency treatment is conducted to the preprocessed image according to spectral residual theory, and a visual saliency map containing numeric regions is obtained. Finally, saliency map is classified by using the convolution neural network to distinguish digital region from non-digital region. The experimental results show that the method can detect number of conveyor belt image and realize surface fault location of conveyor belt.
-
Keywords:
- conveyor belt /
- fault location /
- neural network /
- visual saliency /
- image processing
-
-
期刊类型引用(8)
1. 张新磊，谢翠娟，许俊锋. 机器人自主探索规划方法和系统设计. 计算机与数字工程. 2024(10): 2908-2913+2959 . 
百度学术
2. 朱雷，闯喜宏，李磊，梁秋祥，魏铭毅. 输煤廊道巡检机器人智能避障控制方法设计. 自动化与仪器仪表. 2023(07): 210-213+218 . 百度学术
3. 郭爱军，刘建宇，郑安. 基于多传感器信息融合灭火机器人避障策略. 煤炭技术. 2022(04): 201-203 . 百度学术
4. 张燕. 基于模糊控制的智能移动机器人最优避障路线生成系统. 机械设计与制造工程. 2022(04): 86-90 . 百度学术
5. 杨春雨，张鑫. 煤矿机器人环境感知与路径规划关键技术. 煤炭学报. 2022(07): 2844-2872 . 百度学术
6. 岳仍瑞. 基于概率论的机器人避障方案规划. 信息记录材料. 2021(01): 178-179 . 百度学术
7. 刘曦，柳文轩，阿丽米热·阿布来提，麦吾兰·伊力哈木. 双目避障系统在桥式起重机上的应用. 现代制造技术与装备. 2021(10): 95-97 . 百度学术
8. 袁晓明，郝明锐. 煤矿无轨辅助运输无人驾驶关键技术与发展趋势研究. 智能矿山. 2020(01): 89-97 . 百度学术
其他类型引用(1)
计量
- 文章访问数: 48
- HTML全文浏览量: 11
- PDF下载量: 9
- 被引次数: 9
下载:
