针对已有的目标检测方法在复杂场景中对鱼类目标检测效果不理想的问题,提出了一种基于膨胀卷积和参数重构的鱼类目标实时检测方法.先设计了一种四分支融合卷积结构,在引入少量参数量的情况下,扩大了目标检测的感受野,提升了目标检测的效...针对已有的目标检测方法在复杂场景中对鱼类目标检测效果不理想的问题,提出了一种基于膨胀卷积和参数重构的鱼类目标实时检测方法.先设计了一种四分支融合卷积结构,在引入少量参数量的情况下,扩大了目标检测的感受野,提升了目标检测的效果.再引入了RepVGG(重构VGG)并联辅助分支思想,在训练过程中使用复杂模型进行特征学习,而在推理过程中对模型中的BN(Batch Normalization)层以及1×1的辅助分支中的参数进行融合,利用参数重构对训练过程的冗余参量进行合并,保证了模型的低参数量和实时推理.基于YOLOv5s进行实验,相比原始的YOLOv5s获得了更高的检测精度和召回率,平均精度(mean Average Precision,mAP)达到83.1%,超越了目前主流的目标检测算法.提出的算法在检测速度上相比原始模型无明显降低,处理速度上达到100FPS,在实现高精度检测的前提下保证了鱼类目标的实时检测,为基于视觉的鱼类检测方案提供了有效的技术支持.展开更多
鱼类目标检测对渔业精准养殖、生产自动化、资源调查及鱼行为的研究等具有重要的意义。为了能快速准确地得到鱼类目标的位置和所属类别,提出了一种改进YOLO v4模型的鱼类目标检测方法,在CIoU(Complete Intersection over Union)损失函...鱼类目标检测对渔业精准养殖、生产自动化、资源调查及鱼行为的研究等具有重要的意义。为了能快速准确地得到鱼类目标的位置和所属类别,提出了一种改进YOLO v4模型的鱼类目标检测方法,在CIoU(Complete Intersection over Union)损失函数基础上构建了新的损失项,改进的损失函数使真实框与相交框呈相同宽高比进行回归,同时通过设置多锚点框模式,增强在特定尺寸面积上的检测效果。结果显示:改进YOLO v4模型的mAP(mean Average Precision)比原模型有较大提升,在自建数据集、Fish4Knowledge数据集和NCFM数据集上的mAP分别达到了94.22%、99.52%、92.16%。研究表明,改进YOLO v4模型可以快速准确地检测到鱼的位置和类别,检测速度满足实时的要求,可以为渔业精准养殖等提供参考。展开更多
在鱼类识别检测技术中,采用anchor-free算法中的Centernet算法对鱼类进行识别检测时,低层特征信息容易丢失,导致识别精度和识别效率降低。为此,提出了一种基于Feature fusion Module and Loss function optimization of Centernet(FMLCe...在鱼类识别检测技术中,采用anchor-free算法中的Centernet算法对鱼类进行识别检测时,低层特征信息容易丢失,导致识别精度和识别效率降低。为此,提出了一种基于Feature fusion Module and Loss function optimization of Centernet(FMLCenternet)算法的鱼类识别检测算法。在Centernet算法网络结构中引入特征融合模块将低层特征信息和高层特征信息融合,输出更加完整的特征图,提高识别检测精度;设置参数调节正负样本的损失比例,使得网络模型的损失函数得到优化,提高整个模型的识别检测效率。在PASCALVOC数据集中对所提算法进行有效性的验证,并对网络结构的性能进行分析。收集大量的目标数据集以及标注数据集信息,训练优化的网络结构并与不同的模型进行对比分析。实验结果表明,FMLCenternet算法对鱼类进行识别检测时,识别平均精度(AP50)可以达到85%以上,平均检测时间低于100 ms。所提算法不仅识别检测精度较高,而且识别检测效率也得到了提升。展开更多
The spatial distribution pattern of organisms is a basic issue in understanding the mechanisms of community assembly. Although the spatial distributions of animals and plants have been well studied,those of microorgan...The spatial distribution pattern of organisms is a basic issue in understanding the mechanisms of community assembly. Although the spatial distributions of animals and plants have been well studied,those of microorganisms are still being debated. In this study, we used a fi sh gut microecosystem to detect the spatial pattern of microbes, because it can provide a relatively unifi ed and stable environment. Results suggest that the turnover of intestinal bacterial assemblages showed a weak but highly signifi cant negative correlation between similarity and distances in the microbial community, in respect of both grass carp intestinal loci distances and the geographical distance between fi sh sampling sites. Our results also suggest that intestinal bacterial assemblages responded to differences within the external environment and within different parts of the fi sh themselves. These results show that some, or possibly all, microbes are restricted in their distribution and that environmental factors are also important infl uences on the structure of intestinal bacterial assemblages. The fi sh gut microecosystem is useful in promoting study of the spatial distribution patterns of microorganisms.展开更多
We isolated a strain of lymphocystis disease virus （LCDV） from Japanese flounder （Paralichthys olivaceus） cultured in northern China. Based on published sequences of major capsid protein （MCP） gene of LCDV-cn （...We isolated a strain of lymphocystis disease virus （LCDV） from Japanese flounder （Paralichthys olivaceus） cultured in northern China. Based on published sequences of major capsid protein （MCP） gene of LCDV-cn （GenBank： AF126405）, we designed two primer sets P1/P2 and P3/P4. We then used one-step or nested PCR and in-situ hybridization （ISFI） to detect LCDV and identify the target tissues or cells in infected Japanese flounder. The PCR products were positive in purified viral supematant, skin nodules, gut, gill, kidney, spleen, stomach, heart, and liver of Japanese flounder. We compared the DNA sequence with 14 MCP nucleotide sequences from GenBank, including Megalocytivirus （OFIV and RSIV）, lridovirus （CzlV and W/V）, Ranavirus （TFV and FV3）, and Lymphocystivirus （8 LCDV）. Based on the alignment, we confirmed the PCR product was from Lymphocystivirus （GenBank accession number DQ279090 （LCDV-HD））. Using ISH, we noted the presence of LCDV in the skin nodules, gut, gill, spleen, stomach, and heart of spontaneously infected Japanese flounders. We successfully amplified LCDV fragments from Schlegel＇s black rockfish （Sebastes schlegeli Higendorf）, redwing sea robin （Lepidotrigla microptera Gtinther） and turbot （Scophthalmus maximus） using the one-step and nested PCR, suggesting the target genes can be widely detected in fish using this method.展开更多
Iridovirus infection often causes death and considerable economic losses in the aquaculture industry. This research applies the co-agglutination method that is fast, cheap and accurate in confirming the diagnosis of t...Iridovirus infection often causes death and considerable economic losses in the aquaculture industry. This research applies the co-agglutination method that is fast, cheap and accurate in confirming the diagnosis of the cause of an outbreak of illness caused by iridovirus in the field, so that remedial action can be taken quickly and appropriately to minimize the impact of wider losses. Samples were taken from the grouper and pomfret star farms that are experiencing outbreaks of infectious diseases in the months from May to August 2015, Tanjungpinang, Indonesia. The sick and allegedly attacked by iridovirus samples showed abnormal swimming clinical symptoms, weakness and the swollen spleen. The swollen spleen of sick fish created suspension in phosphate buffered saline （PBS） with pH 7.2, and then centrifuged at 8,000 rpm for I0 rain. The supernatant after centrifuge was used as the test sample. On a clean object glass, 50 μL of the supematant was treated with 50 μL kit co-agglutination pre-prepared. The positive results were shown by the agglutination reaction after 10-15 rain, while as a negative control, PBS was reacted with co-agglutination kit that looked homogeneous （no agglutination）. It was showed that the grouper （Epinepkelus sp.） on four farms and pomfret star （Thracinotus blochii） on one farm that experienced an outbreak of infectious disease in Tanjungpinang showed positively infected iridovirus. The same positive iridovirus result was also demonstrated by examination using polymerase chain reaction （PCR） at 570 bp. So, the causative agent of plague on grouper and pomfret star was iridovirus. In addition, the co-agglutination test based on serology is more quick, cheap and accurate.展开更多
文摘针对已有的目标检测方法在复杂场景中对鱼类目标检测效果不理想的问题,提出了一种基于膨胀卷积和参数重构的鱼类目标实时检测方法.先设计了一种四分支融合卷积结构,在引入少量参数量的情况下,扩大了目标检测的感受野,提升了目标检测的效果.再引入了RepVGG(重构VGG)并联辅助分支思想,在训练过程中使用复杂模型进行特征学习,而在推理过程中对模型中的BN(Batch Normalization)层以及1×1的辅助分支中的参数进行融合,利用参数重构对训练过程的冗余参量进行合并,保证了模型的低参数量和实时推理.基于YOLOv5s进行实验,相比原始的YOLOv5s获得了更高的检测精度和召回率,平均精度(mean Average Precision,mAP)达到83.1%,超越了目前主流的目标检测算法.提出的算法在检测速度上相比原始模型无明显降低,处理速度上达到100FPS,在实现高精度检测的前提下保证了鱼类目标的实时检测,为基于视觉的鱼类检测方案提供了有效的技术支持.
文摘鱼类目标检测对渔业精准养殖、生产自动化、资源调查及鱼行为的研究等具有重要的意义。为了能快速准确地得到鱼类目标的位置和所属类别,提出了一种改进YOLO v4模型的鱼类目标检测方法,在CIoU(Complete Intersection over Union)损失函数基础上构建了新的损失项,改进的损失函数使真实框与相交框呈相同宽高比进行回归,同时通过设置多锚点框模式,增强在特定尺寸面积上的检测效果。结果显示:改进YOLO v4模型的mAP(mean Average Precision)比原模型有较大提升,在自建数据集、Fish4Knowledge数据集和NCFM数据集上的mAP分别达到了94.22%、99.52%、92.16%。研究表明,改进YOLO v4模型可以快速准确地检测到鱼的位置和类别,检测速度满足实时的要求,可以为渔业精准养殖等提供参考。
文摘在鱼类识别检测技术中,采用anchor-free算法中的Centernet算法对鱼类进行识别检测时,低层特征信息容易丢失,导致识别精度和识别效率降低。为此,提出了一种基于Feature fusion Module and Loss function optimization of Centernet(FMLCenternet)算法的鱼类识别检测算法。在Centernet算法网络结构中引入特征融合模块将低层特征信息和高层特征信息融合,输出更加完整的特征图,提高识别检测精度;设置参数调节正负样本的损失比例,使得网络模型的损失函数得到优化,提高整个模型的识别检测效率。在PASCALVOC数据集中对所提算法进行有效性的验证,并对网络结构的性能进行分析。收集大量的目标数据集以及标注数据集信息,训练优化的网络结构并与不同的模型进行对比分析。实验结果表明,FMLCenternet算法对鱼类进行识别检测时,识别平均精度(AP50)可以达到85%以上,平均检测时间低于100 ms。所提算法不仅识别检测精度较高,而且识别检测效率也得到了提升。
基金Supported by the National Basic Research Program of China(973 Program)(No.2009CB118705)the National Natural Science Foundation of China(Nos.30970358,31071896)the Youth Innovation Promotion Association,CAS
文摘The spatial distribution pattern of organisms is a basic issue in understanding the mechanisms of community assembly. Although the spatial distributions of animals and plants have been well studied,those of microorganisms are still being debated. In this study, we used a fi sh gut microecosystem to detect the spatial pattern of microbes, because it can provide a relatively unifi ed and stable environment. Results suggest that the turnover of intestinal bacterial assemblages showed a weak but highly signifi cant negative correlation between similarity and distances in the microbial community, in respect of both grass carp intestinal loci distances and the geographical distance between fi sh sampling sites. Our results also suggest that intestinal bacterial assemblages responded to differences within the external environment and within different parts of the fi sh themselves. These results show that some, or possibly all, microbes are restricted in their distribution and that environmental factors are also important infl uences on the structure of intestinal bacterial assemblages. The fi sh gut microecosystem is useful in promoting study of the spatial distribution patterns of microorganisms.
基金Supported by the National Natural Science Foundation of China (No 30771648)the National High Technology Research and Development Program of China (863 Program) (No 2006AA100306)
文摘We isolated a strain of lymphocystis disease virus （LCDV） from Japanese flounder （Paralichthys olivaceus） cultured in northern China. Based on published sequences of major capsid protein （MCP） gene of LCDV-cn （GenBank： AF126405）, we designed two primer sets P1/P2 and P3/P4. We then used one-step or nested PCR and in-situ hybridization （ISFI） to detect LCDV and identify the target tissues or cells in infected Japanese flounder. The PCR products were positive in purified viral supematant, skin nodules, gut, gill, kidney, spleen, stomach, heart, and liver of Japanese flounder. We compared the DNA sequence with 14 MCP nucleotide sequences from GenBank, including Megalocytivirus （OFIV and RSIV）, lridovirus （CzlV and W/V）, Ranavirus （TFV and FV3）, and Lymphocystivirus （8 LCDV）. Based on the alignment, we confirmed the PCR product was from Lymphocystivirus （GenBank accession number DQ279090 （LCDV-HD））. Using ISH, we noted the presence of LCDV in the skin nodules, gut, gill, spleen, stomach, and heart of spontaneously infected Japanese flounders. We successfully amplified LCDV fragments from Schlegel＇s black rockfish （Sebastes schlegeli Higendorf）, redwing sea robin （Lepidotrigla microptera Gtinther） and turbot （Scophthalmus maximus） using the one-step and nested PCR, suggesting the target genes can be widely detected in fish using this method.
文摘Iridovirus infection often causes death and considerable economic losses in the aquaculture industry. This research applies the co-agglutination method that is fast, cheap and accurate in confirming the diagnosis of the cause of an outbreak of illness caused by iridovirus in the field, so that remedial action can be taken quickly and appropriately to minimize the impact of wider losses. Samples were taken from the grouper and pomfret star farms that are experiencing outbreaks of infectious diseases in the months from May to August 2015, Tanjungpinang, Indonesia. The sick and allegedly attacked by iridovirus samples showed abnormal swimming clinical symptoms, weakness and the swollen spleen. The swollen spleen of sick fish created suspension in phosphate buffered saline （PBS） with pH 7.2, and then centrifuged at 8,000 rpm for I0 rain. The supernatant after centrifuge was used as the test sample. On a clean object glass, 50 μL of the supematant was treated with 50 μL kit co-agglutination pre-prepared. The positive results were shown by the agglutination reaction after 10-15 rain, while as a negative control, PBS was reacted with co-agglutination kit that looked homogeneous （no agglutination）. It was showed that the grouper （Epinepkelus sp.） on four farms and pomfret star （Thracinotus blochii） on one farm that experienced an outbreak of infectious disease in Tanjungpinang showed positively infected iridovirus. The same positive iridovirus result was also demonstrated by examination using polymerase chain reaction （PCR） at 570 bp. So, the causative agent of plague on grouper and pomfret star was iridovirus. In addition, the co-agglutination test based on serology is more quick, cheap and accurate.
