随着深度学习技术的发展,水下图像检测近年来受到广泛的关注,为了克服在复杂水下环境下传统小鱼群的误检、漏检和识别准确率低等问题,提出一种改进YOLOv5的目标检测方法(INV-YOLOv5)。该方法包括将YOLOv5m中的Focus模块替换为卷积模块,...随着深度学习技术的发展,水下图像检测近年来受到广泛的关注,为了克服在复杂水下环境下传统小鱼群的误检、漏检和识别准确率低等问题,提出一种改进YOLOv5的目标检测方法(INV-YOLOv5)。该方法包括将YOLOv5m中的Focus模块替换为卷积模块,提高网络精度;在主干网络(Backbone)中添加多头自注意力机制,增大网络特征提取视野;最后,在网络中引入了内卷算子和加权的特征融合,降低网络的参数量,提高检测精度。在实验阶段,使用Labeled Fishes in the Wild数据集和WildFish数据集验证,该方法的平均精度(mAP)分别为81.7%和83.6%,与YOLOv5m网络相比分别提升了6%和14.5%,不仅拥有较高的识别率并且更加轻量化,而且模型大小与YOLOv5m网络相比减少了6 M(Mega)左右,验证了所提出的改进方法具有较好的效果。展开更多
针对夜间场景光线微弱、能见度低导致夜视图像信噪比低、成像质量差的问题,提出了基于深度学习的双通道夜视图像复原方法。首先,用两种基于全连接多尺度残差学习分块(FMRB)的卷积神经网络(CNN)分别对红外夜视图像和微光夜视图像进行多...针对夜间场景光线微弱、能见度低导致夜视图像信噪比低、成像质量差的问题,提出了基于深度学习的双通道夜视图像复原方法。首先,用两种基于全连接多尺度残差学习分块(FMRB)的卷积神经网络(CNN)分别对红外夜视图像和微光夜视图像进行多尺度特征提取和层次特征融合,从而得到重建的红外图像和增强的微光图像;然后,两种处理后的图像通过自适应加权平均算法进行融合,并根据场景的不同自适应地凸显两个图像中具有更高显著性的有效信息;最后,得到分辨率高且视觉效果好的夜视复原图像。使用基于FMRB的深度学习网络得到的红外夜视重建图像,相较于卷积神经网络超分辨率(SRCNN)重建算法得到的在峰值信噪比(PSNR)和结构相似性(SSIM)的平均值上分别提升了3.56 d B和0.091 2;相较于MSRCR,得到的微光夜视增强图像在PSNR和SSIM的平均值上分别提升了6.82 d B和0.132 1。实验结果表明,所提方法得到的重建图像的清晰度明显得到改善,获得的增强图像的亮度也明显得到提升,而且前二者的融合图像的视觉效果较好,可见所提方法能有效改善夜视图像的复原效果。展开更多
To screen chemicals for controlling potato blight effectively, randomized block experiment was adopted, the control efficiency of five chemicals for Carbendazim, Metalaxyl, Dithane mancozeb, Curzate, Cymoxanil mancoze...To screen chemicals for controlling potato blight effectively, randomized block experiment was adopted, the control efficiency of five chemicals for Carbendazim, Metalaxyl, Dithane mancozeb, Curzate, Cymoxanil mancozeb was explored. The results showed that the efficiency of Metalaxyl was much better than that of other conventional chemicals in controlling potato blight, up to 70.59% ; the mean control efficiency of Curzate and Cymoxanil mancozeb was 68.76% and 67.05% ; the control efficiency of Carbendazim was lower, 55.76%. Big and middle tuber rate, tuber weight per hole, plot yield, unit area yield of potatoes in Metalaxyl treatment were the highest, 78.05%, 437.24 g/hole, 34.0 kg/21.6 m2 and 15 748.61 kg/hm2, 49.12% higher than the control yield. The indexes of potatoes in Cymoxanil mancozeb treatment were the second highest, 73.33%, 385.31 g/hole, 32.8 kg/21.6 m2 and 15 192.78 kg/hm2, 43.86% higher than the control yield.展开更多
文摘随着深度学习技术的发展,水下图像检测近年来受到广泛的关注,为了克服在复杂水下环境下传统小鱼群的误检、漏检和识别准确率低等问题,提出一种改进YOLOv5的目标检测方法(INV-YOLOv5)。该方法包括将YOLOv5m中的Focus模块替换为卷积模块,提高网络精度;在主干网络(Backbone)中添加多头自注意力机制,增大网络特征提取视野;最后,在网络中引入了内卷算子和加权的特征融合,降低网络的参数量,提高检测精度。在实验阶段,使用Labeled Fishes in the Wild数据集和WildFish数据集验证,该方法的平均精度(mAP)分别为81.7%和83.6%,与YOLOv5m网络相比分别提升了6%和14.5%,不仅拥有较高的识别率并且更加轻量化,而且模型大小与YOLOv5m网络相比减少了6 M(Mega)左右,验证了所提出的改进方法具有较好的效果。
文摘针对夜间场景光线微弱、能见度低导致夜视图像信噪比低、成像质量差的问题,提出了基于深度学习的双通道夜视图像复原方法。首先,用两种基于全连接多尺度残差学习分块(FMRB)的卷积神经网络(CNN)分别对红外夜视图像和微光夜视图像进行多尺度特征提取和层次特征融合,从而得到重建的红外图像和增强的微光图像;然后,两种处理后的图像通过自适应加权平均算法进行融合,并根据场景的不同自适应地凸显两个图像中具有更高显著性的有效信息;最后,得到分辨率高且视觉效果好的夜视复原图像。使用基于FMRB的深度学习网络得到的红外夜视重建图像,相较于卷积神经网络超分辨率(SRCNN)重建算法得到的在峰值信噪比(PSNR)和结构相似性(SSIM)的平均值上分别提升了3.56 d B和0.091 2;相较于MSRCR,得到的微光夜视增强图像在PSNR和SSIM的平均值上分别提升了6.82 d B和0.132 1。实验结果表明,所提方法得到的重建图像的清晰度明显得到改善,获得的增强图像的亮度也明显得到提升,而且前二者的融合图像的视觉效果较好,可见所提方法能有效改善夜视图像的复原效果。
基金Sponsored by Special Fund for the Construction of Guizhou Provincial Modern Agricultural Industrial Technological System(GZCYTX2013)Potato Sci-tech Base Construction of Qixingguan District,Bijie City(2012NO.5)
文摘To screen chemicals for controlling potato blight effectively, randomized block experiment was adopted, the control efficiency of five chemicals for Carbendazim, Metalaxyl, Dithane mancozeb, Curzate, Cymoxanil mancozeb was explored. The results showed that the efficiency of Metalaxyl was much better than that of other conventional chemicals in controlling potato blight, up to 70.59% ; the mean control efficiency of Curzate and Cymoxanil mancozeb was 68.76% and 67.05% ; the control efficiency of Carbendazim was lower, 55.76%. Big and middle tuber rate, tuber weight per hole, plot yield, unit area yield of potatoes in Metalaxyl treatment were the highest, 78.05%, 437.24 g/hole, 34.0 kg/21.6 m2 and 15 748.61 kg/hm2, 49.12% higher than the control yield. The indexes of potatoes in Cymoxanil mancozeb treatment were the second highest, 73.33%, 385.31 g/hole, 32.8 kg/21.6 m2 and 15 192.78 kg/hm2, 43.86% higher than the control yield.
