针对YOLO v5l(you only look once version 5 large)算法对于小目标、少样本且背景复杂的排水管道缺陷图像检测的精度低、误检和漏检率较高等问题,提出了一种基于YOLO v5l-Im算法的排水管道缺陷检测改进方法。做了三点改进:首先提出了Fo...针对YOLO v5l(you only look once version 5 large)算法对于小目标、少样本且背景复杂的排水管道缺陷图像检测的精度低、误检和漏检率较高等问题,提出了一种基于YOLO v5l-Im算法的排水管道缺陷检测改进方法。做了三点改进:首先提出了Focal-EIoU(focal embedding intersection over union)损失函数,有效提升了检测模型的性能;其次为增强检测模型对小目标缺陷的检测效果,减少缺陷误检和漏检的概率,将骨干网络中浅层特征图融合到双向特征金字塔网络(bidirectional feature pyramid network,BiFPN)中,增加针对小目标的预测层;最后在YOLO v5l中引入坐标注意力机制(coordinate attention,CA),提高模型对图像中感兴趣区域的敏感程度,减少冗余背景信息的干扰。3种改进对平均检测准确率(mean average precision,mAP)的提升分别为2.0、2.9、5.9个百分点。将三种有效改进融合到一起,检测结果表明:本文提出的YOLO v5l-Im模型的mAP达到了92.1%,较原模型的85.5%提升了6.5个百分点。由此可见,所做的改进有效增强了YOLO v5l对排水管道缺陷的检测能力。展开更多
To distinguish the taxonomy of Yihuang wild rice found in Yihuang County, Jiangxi Province, China in 1999, the botanical characteristics of Yihuang wild rice were retrieved by genus, and species; at the same time, the...To distinguish the taxonomy of Yihuang wild rice found in Yihuang County, Jiangxi Province, China in 1999, the botanical characteristics of Yihuang wild rice were retrieved by genus, and species; at the same time, the tubercle characters on lemma in Yihuang wild rice, O. rufipogon, 0. officinalis, 0. meyeriana, and O. granulata, were also observed by scanning electron microscopy. The retrieval result by genus showed that Yihuang wild rice is Leersia Soland. Ex Swartz., not Oryza L., Hygroryza Nees and Chikusichloa Koidz. The retrieval result by species showed that Yihuang wild rice is L. sayanuka Ohwi, not L. hexandra Swartz., L. japonica (Makino) Honda and L. oryzoides (L.) Swartz. The results of the SEM observation on tubercles on lemma showed that the tubercle characters on lemma of Yihuang wild rice belongs to "inlaid- silica-peak" type, which is different from the three types of wild rice existing in China, O. rufipogon, O. officinalis, O. Meyeriana, and also O. granulata, but is similar to L. tisseranti, thereby gives a proof of the above retrieval results. It could be concluded that Yihuang wild rice does not belong to Oryza L., but to Leersia sayanuka Ohwi of Leersia Soland. Ex Swartz.展开更多
分别在800℃、825℃、850℃焊接温度、30 m in保温时间,3 MPa焊接压力下,进行Ti-6A l-4V钛合金板与304L不锈钢网的真空扩散焊接。对接头组织结构与化学元素扩散进行了扫描电镜与能谱分析,并测试了接头的剪切强度。结果表明:不添加中间...分别在800℃、825℃、850℃焊接温度、30 m in保温时间,3 MPa焊接压力下,进行Ti-6A l-4V钛合金板与304L不锈钢网的真空扩散焊接。对接头组织结构与化学元素扩散进行了扫描电镜与能谱分析,并测试了接头的剪切强度。结果表明:不添加中间过渡层金属,可以成功地实现钛合金板与不锈钢网的扩散焊接,并使接头的剪切强度达到90 MPa以上。不锈钢网中的Fe、N、iCr扩散并固溶到钛合金中,稳定了β相,使钛合金在一定深度上,其组织由原来的α+β双相结构转变为单相的β相。不锈钢中的Cr,由于钛合金中Ti的扩散进入,而在界面发生了上坡扩散现象。这种Cr在不锈钢一定深度内的富集,形成窄长的富Cr区域,冷却后转变为硬脆的σ相。但在焊接接头中没有发现明显其它的金属间化合物或氧化物相的生成,使得接头的机械性能得到了很好的保证。展开更多
文摘针对YOLO v5l(you only look once version 5 large)算法对于小目标、少样本且背景复杂的排水管道缺陷图像检测的精度低、误检和漏检率较高等问题,提出了一种基于YOLO v5l-Im算法的排水管道缺陷检测改进方法。做了三点改进:首先提出了Focal-EIoU(focal embedding intersection over union)损失函数,有效提升了检测模型的性能;其次为增强检测模型对小目标缺陷的检测效果,减少缺陷误检和漏检的概率,将骨干网络中浅层特征图融合到双向特征金字塔网络(bidirectional feature pyramid network,BiFPN)中,增加针对小目标的预测层;最后在YOLO v5l中引入坐标注意力机制(coordinate attention,CA),提高模型对图像中感兴趣区域的敏感程度,减少冗余背景信息的干扰。3种改进对平均检测准确率(mean average precision,mAP)的提升分别为2.0、2.9、5.9个百分点。将三种有效改进融合到一起,检测结果表明:本文提出的YOLO v5l-Im模型的mAP达到了92.1%,较原模型的85.5%提升了6.5个百分点。由此可见,所做的改进有效增强了YOLO v5l对排水管道缺陷的检测能力。
基金supported by the National Natural Science Foundation of China(30560073)the Key Technologies R&D Program of Jiangxi Province,China(200210100210).
文摘To distinguish the taxonomy of Yihuang wild rice found in Yihuang County, Jiangxi Province, China in 1999, the botanical characteristics of Yihuang wild rice were retrieved by genus, and species; at the same time, the tubercle characters on lemma in Yihuang wild rice, O. rufipogon, 0. officinalis, 0. meyeriana, and O. granulata, were also observed by scanning electron microscopy. The retrieval result by genus showed that Yihuang wild rice is Leersia Soland. Ex Swartz., not Oryza L., Hygroryza Nees and Chikusichloa Koidz. The retrieval result by species showed that Yihuang wild rice is L. sayanuka Ohwi, not L. hexandra Swartz., L. japonica (Makino) Honda and L. oryzoides (L.) Swartz. The results of the SEM observation on tubercles on lemma showed that the tubercle characters on lemma of Yihuang wild rice belongs to "inlaid- silica-peak" type, which is different from the three types of wild rice existing in China, O. rufipogon, O. officinalis, O. Meyeriana, and also O. granulata, but is similar to L. tisseranti, thereby gives a proof of the above retrieval results. It could be concluded that Yihuang wild rice does not belong to Oryza L., but to Leersia sayanuka Ohwi of Leersia Soland. Ex Swartz.