The time-domain calibration coefficient of a D-Dot sensor should be identical across various transverse electromagnetic (TEM) cells to comply with the IEEE Std 1309. However, in our previous calibration experiments, p...The time-domain calibration coefficient of a D-Dot sensor should be identical across various transverse electromagnetic (TEM) cells to comply with the IEEE Std 1309. However, in our previous calibration experiments, poor consistency was observed. The size of D-Dot sensors relative to TEM cells is considered the main reason for this poor consistency. Therefore, this study aims at determining the calibration coefficient of a D-Dot sensor. We calculate the theoretical coefficient as a reference. Practical calibration experiments involve the processing of TEM cells with three different sizes. To observe the response more clearly, corresponding models are constructed and numerical simulations are performed. The numerical simulations and experimental calibration are in good agreement. To determine the calibration accuracy, we quantify the accuracy using the relative error of the calibration coefficient. By comparing the coefficients obtained, it can be concluded that the perturbation error is about 15% when the relative size is over 1/3. Further, the relative size should be less than 1/5 to obtain a relative error below 10%.展开更多
煤矿掘进巷道锚护位置的精准识别与定位是钻锚机器人实现智能永久支护亟需突破的关键技术。笔者提出一种基于视觉图像与激光点云融合的巷道锚护孔位智能识别定位方法,包括图像目标识别、点云图像特征融合和定位坐标提取3个步骤:①针对...煤矿掘进巷道锚护位置的精准识别与定位是钻锚机器人实现智能永久支护亟需突破的关键技术。笔者提出一种基于视觉图像与激光点云融合的巷道锚护孔位智能识别定位方法,包括图像目标识别、点云图像特征融合和定位坐标提取3个步骤:①针对煤矿井下低照度、水雾和粉尘等环境因素导致的锚孔轮廓成像模糊的问题,采用IA(Image-Adaptive)-SimAM-YOLOv7-tiny网络对巷道待锚护孔位进行视觉识别,该网络能够自适应地增强图像亮度和对比度,恢复锚孔边缘的高频信息,并使模型重点关注锚孔特征,提高锚孔检测的成功率;②求解激光雷达和工业相机联合标定的外参矩阵,将图像检测的锚孔边界框通过透视投影关系生成锥形感兴趣区域(Region Of Interest,ROI),获得对应的目标点云团簇;③采用点云处理算法提取锚护孔位边界点云,获得孔位中心坐标及其法向量,并通过坐标深度差比较判断锚孔识别的正确性。文中搭建了锚杆台车机械臂钻孔定位系统,对算法自主定位的精度以及准确度进行验证,试验结果表明:IA-SimAM-YOLOv7-tiny模型的平均精度均值(Mean Average Precision,mAP)为87.3%,较YOLOv7-tiny模型提高了4.6%;提出的融合算法定位误差为3 mm,单锚孔情况下系统平均识别时间为0.77 s,与单一视觉方法相比,采用激光与视觉多源融合不仅可以降低环境和小样本训练对定位性能的影响,而且可以获得锚护孔位的法向量,为机械臂调整钻孔位姿实现精准锚固提供依据。展开更多
文摘The time-domain calibration coefficient of a D-Dot sensor should be identical across various transverse electromagnetic (TEM) cells to comply with the IEEE Std 1309. However, in our previous calibration experiments, poor consistency was observed. The size of D-Dot sensors relative to TEM cells is considered the main reason for this poor consistency. Therefore, this study aims at determining the calibration coefficient of a D-Dot sensor. We calculate the theoretical coefficient as a reference. Practical calibration experiments involve the processing of TEM cells with three different sizes. To observe the response more clearly, corresponding models are constructed and numerical simulations are performed. The numerical simulations and experimental calibration are in good agreement. To determine the calibration accuracy, we quantify the accuracy using the relative error of the calibration coefficient. By comparing the coefficients obtained, it can be concluded that the perturbation error is about 15% when the relative size is over 1/3. Further, the relative size should be less than 1/5 to obtain a relative error below 10%.
文摘煤矿掘进巷道锚护位置的精准识别与定位是钻锚机器人实现智能永久支护亟需突破的关键技术。笔者提出一种基于视觉图像与激光点云融合的巷道锚护孔位智能识别定位方法,包括图像目标识别、点云图像特征融合和定位坐标提取3个步骤:①针对煤矿井下低照度、水雾和粉尘等环境因素导致的锚孔轮廓成像模糊的问题,采用IA(Image-Adaptive)-SimAM-YOLOv7-tiny网络对巷道待锚护孔位进行视觉识别,该网络能够自适应地增强图像亮度和对比度,恢复锚孔边缘的高频信息,并使模型重点关注锚孔特征,提高锚孔检测的成功率;②求解激光雷达和工业相机联合标定的外参矩阵,将图像检测的锚孔边界框通过透视投影关系生成锥形感兴趣区域(Region Of Interest,ROI),获得对应的目标点云团簇;③采用点云处理算法提取锚护孔位边界点云,获得孔位中心坐标及其法向量,并通过坐标深度差比较判断锚孔识别的正确性。文中搭建了锚杆台车机械臂钻孔定位系统,对算法自主定位的精度以及准确度进行验证,试验结果表明:IA-SimAM-YOLOv7-tiny模型的平均精度均值(Mean Average Precision,mAP)为87.3%,较YOLOv7-tiny模型提高了4.6%;提出的融合算法定位误差为3 mm,单锚孔情况下系统平均识别时间为0.77 s,与单一视觉方法相比,采用激光与视觉多源融合不仅可以降低环境和小样本训练对定位性能的影响,而且可以获得锚护孔位的法向量,为机械臂调整钻孔位姿实现精准锚固提供依据。
