A Pyramidal Morphology Algorithm is developed for speckle reduction of SARimages in this paper. For reducing the loss of information in the pyramidal algorithm for morphologyprocessing, in this modified algorithm, the...A Pyramidal Morphology Algorithm is developed for speckle reduction of SARimages in this paper. For reducing the loss of information in the pyramidal algorithm for morphologyprocessing, in this modified algorithm, the sub-images are processed parallel in the downsamplingoperation and the sub-images are reconstructed in the upsampling operation. It can be applied toimage filtering parallel. After analysis the computer simulations show that these two kinds offilters are both effective in speckle reduction of SAR images. The modified parallel algorithm doesbetter than the original algorithm and Lee filter on some characteristics.展开更多
针对车辆漆面缺陷检测精度低、检测算法参数量大、难易样本不均匀等问题,提出一种基于改进YOLOv8的车辆漆面检测算法。首先,为了提升划痕状缺陷检测能力并降低模型规模,将DAT(Deformable Attention Transformer)注意力机制引入主干网络...针对车辆漆面缺陷检测精度低、检测算法参数量大、难易样本不均匀等问题,提出一种基于改进YOLOv8的车辆漆面检测算法。首先,为了提升划痕状缺陷检测能力并降低模型规模,将DAT(Deformable Attention Transformer)注意力机制引入主干网络来增强长距离特征依赖关系,同时使用幻影卷积(GhostConv)替换网络中的卷积(Conv)模块。然后,为了提升特征提取能力并进一步降低模型规模,结合FasterBlock模块与高效多尺度注意力(EMA)机制提出C2f-E(C2f Based on EMA)模块。接着,为了提高小目标检测性能,基于双向特征金字塔网络(BiFPN)进行设计,并增加小目标检测头与多尺度特征融合支路,提出BiFPN-D(BiFPN with Small Object Detection Head)颈部金字塔结构。最后,为了解决难易样本的平衡问题并提高针对小目标缺陷的检测性能,使用WIoUv3(Wise-Intersection over Union version 3)作为训练网络的损失函数。在自建的车辆漆面缺陷数据集上进行训练并开展对比实验。实验结果表明,相较于YOLOv8n,改进模型的均值平均精度(mAP@0.5)提高了5.5百分点、规模减小了1.4×106。展开更多
基金the National Natural Science Foundation of Jiangsu Province.China.( No.BK2 0 0 10 47)
文摘A Pyramidal Morphology Algorithm is developed for speckle reduction of SARimages in this paper. For reducing the loss of information in the pyramidal algorithm for morphologyprocessing, in this modified algorithm, the sub-images are processed parallel in the downsamplingoperation and the sub-images are reconstructed in the upsampling operation. It can be applied toimage filtering parallel. After analysis the computer simulations show that these two kinds offilters are both effective in speckle reduction of SAR images. The modified parallel algorithm doesbetter than the original algorithm and Lee filter on some characteristics.
文摘针对车辆漆面缺陷检测精度低、检测算法参数量大、难易样本不均匀等问题,提出一种基于改进YOLOv8的车辆漆面检测算法。首先,为了提升划痕状缺陷检测能力并降低模型规模,将DAT(Deformable Attention Transformer)注意力机制引入主干网络来增强长距离特征依赖关系,同时使用幻影卷积(GhostConv)替换网络中的卷积(Conv)模块。然后,为了提升特征提取能力并进一步降低模型规模,结合FasterBlock模块与高效多尺度注意力(EMA)机制提出C2f-E(C2f Based on EMA)模块。接着,为了提高小目标检测性能,基于双向特征金字塔网络(BiFPN)进行设计,并增加小目标检测头与多尺度特征融合支路,提出BiFPN-D(BiFPN with Small Object Detection Head)颈部金字塔结构。最后,为了解决难易样本的平衡问题并提高针对小目标缺陷的检测性能,使用WIoUv3(Wise-Intersection over Union version 3)作为训练网络的损失函数。在自建的车辆漆面缺陷数据集上进行训练并开展对比实验。实验结果表明,相较于YOLOv8n,改进模型的均值平均精度(mAP@0.5)提高了5.5百分点、规模减小了1.4×106。
