A novel deep learning method for maize disease identification based on small sample-size and complex background datasets

Maize diseases are a major source of yield loss, but due to the lack of human experience and limitations of traditional image-recognition technology, obtaining satisfactory large-scale identification results of maize diseases are difficult. Fortunately, the advancement of deep learning-based technology makes it possible to automatically identify diseases. However, it still faces issues caused by small sample sizes and complex field background, which affect the accuracy of disease identification. To address these issues, a deep learning-based method was proposed for maize disease identification in this paper. DenseNet121 was used as the main extraction network and a multi-dilated-CBAM-DenseNet (MDCDenseNet) model was built by combining the multi-dilated module and convolutional block attention module (CBAM) attention mechanism. Five models of MDCDenseNet, DenseNet121, ResNet50, MobileNetV2, and NASNetMobile were compared and tested using three kinds of maize leave images from the PlantVillage dataset and field-collected at Northeast Agricultural University in China. Furthermore, auxiliary classifier generative adversarial network (ACGAN) and transfer learning were used to expand the dataset and pre-train for optimal identification results. When tested on field-collected datasets with a complex background, the MDCDenseNet model outperformed compared to these models with an accuracy of 98.84%. Therefore, it can provide a viable reference for the identification of maize leaf diseases collected from the farmland with a small sample size and complex background.     

全景绿视率自动识别和计算研究

绿视率是用于绿色空间感知的直观评价标准,传统研究的绿视率多基于平面影像进行计算,不能完全反映三维空间中人对绿量的主观感受。基于全景影像,提出全景绿视率的概念,通过全景相机获取球面全景照片,将等距圆柱投影转换为等积圆柱投影,利用基于语义分割的卷积神经网络模型,自动识别植被区域面积以实现全景绿视率自动化识别和计量。通过比较5项卷积神经网络模型对绿视率的识别效果,显示出Dilated ResNet-105神经网络模型具有最高的识别准确度。以武汉市武昌区紫阳公园为例,对各级园路和广场的全景绿视率进行计算和分析。将卷积神经网络的识别结果同人工判别结果进行对比研究,结果显示：使用Dilated ResNet-105卷积神经网络对绿植范围识别的平均交并比（mIoU）为62.53%,与人工识别的平均差异为9.17%。全景绿视率自动识别和计算可以为相关研究提供新的思路,实现客观准确、快速便捷的绿视率测量评估。     

Microscopic image recognition of diatoms based on deep learning

Diatoms are a crucial component in the study of aquatic ecosystems and ancient environmental records. However, traditional methods for identifying diatoms, such as morphological taxonomy and molecular detection, are costly, are time consuming, and have limitations. To address these issues, we developed an extensive collection of diatom images, consisting of 7983 images from 160 genera and 1042 species, which we expanded to 49,843 through preprocessing, segmentation, and data augmentation. Our study compared the performance of different algorithms, including backbones, batch sizes, dynamic data augmentation, and static data augmentation on experimental results. We determined that the ResNet152 network outperformed other networks, producing the most accurate results with top-1 and top-5 accuracies of 85.97% and 95.26%, respectively, in identifying 1042 diatom species. Additionally, we propose a method that combines model prediction and cosine similarity to enhance the model's performance in low-probability predictions, achieving an 86.07% accuracy rate in diatom identification. Our research contributes significantly to the recognition and classification of diatom images and has potential applications in water quality assessment, ecological monitoring, and detecting changes in aquatic biodiversity.     

Butterfly-plant network in urban landscape: Implication for conservation and urban greening

Butterflies (Insecta: Lepidoptera) contribute to the ecosystem services and thereby qualify as a group deserving conservation effort. Information on the butterfly-plant links is used as a foundation to sustain population and enhance conservation and management. Thus, in the present study, the structure of a butterfly-plant network in an urban landscape like Kolkata, India, was deciphered highlighting metrics like degree distribution, nestedness, and interaction strength and specialization index. A total of forty eight butterfly species associated with thirty different angiosperm plant species were observed during the study period of one year. While Lantana camara was observed to be the dominant plant species with 37 links to different butterflies, the Catopsilia pyranthe butterfly species was dominant in terms of the generalist pattern of links (21 links) with the plants. Differential ability of the shrubs and herbs in the sustenance of the butterflies was reflected in the network indices using the herbs and the shrubs, separately. In urban landscapes with restricted variety of flowering plants, an estimate of relative strength of interactions enables identification and further use of the concerned species in sustaining butterfly populations. In accordance with these propositions, the butterfly-plant network illustrated in the present instance may prove useful in selection of plant species required for the enhancement of population of desired butterfly species in urban areas like Kolkata, India.     

What can sown wildflower strips contribute to butterfly conservation?: an example from a Swiss lowland agricultural landscape

The objective of this study was to compare butterfly abundances and diversity between wildflower strips and extensively used meadows to identify which butterfly species can be supported by establishing wildflower strips. Butterflies were recorded along transects during one season in twenty-five sown wildflower strips and eleven extensively used meadows in a Swiss lowland agricultural landscape (600 ha). In total 1,669 butterflies of 25 species were observed (25 in the strips, 18 in meadows). This can be related to 38 species recorded in the region (lowland part of Kanton Fribourg) within the Swiss Biodiversity Monitoring Programme. In wildflower strips the number of butterflies per transect meter was significantly higher than in the meadows, but there was no significant difference in species richness. Butterfly communities, though, were quite different between the two habitat types. Habitat type, abundances of flowering plants and presence of forest within 50 m were identified as factors influencing butterfly species richness. Butterfly abundances were affected by habitat type and abundance of flowering plants. In wildflower strips, 65% of all flower visits by butterflies were observed on Origanum. It can be concluded that sown wildflower strips can support a substantial part of a regions species pool. This is mostly true for common species, but can apply to rare species when, for example, larval food plant requirements are met.     

An effective and fast solution for classification of wood species: A deep transfer learning approach

Classification and recognition of wood species have critical importance in wood trade, industry, and science. Therefore, accurate identification of wood species is a great necessity. Conventional classification and recognition of wood species require knowledge and experience on the anatomy of wood which is time-consuming, cost-ineffective, and destructive. Hence, convolutional neural networks (CNNs) -a deep learning tool- have replaced the conventional methods. In this study, classification of wood species via the WOOD-AUTH dataset and evaluating the performance of various deep learning architectures including ResNet-50, Inception V3, Xception, and VGG19 in classification with transfer learning was investigated in detail. The dataset contains macroscopic images of 12 wood species with three different types of wood sections: cross, radial and tangential. The experimental findings demonstrate that Xception produced a remarkable performance as compared to the other models in this study and the WOOD-AUTH dataset owners, yielding a classification accuracy of 95.88%.     

基于深度残差神经网络迁移学习的牙形刺图像识别

为了探索基于深度神经网络模型的牙形刺图像智能识别效果,研究选取奥陶纪8种牙形刺作为研究对象,通过体视显微镜采集牙形刺图像1188幅,收集整理公开发表文献的牙形刺图像778幅,将图像数据集划分为训练集和测试集。通过对训练集图像进行旋转、翻转、滤波增强处理,解决了训练样本不足的问题。基于ResNet-18、ResNet-34、ResNet-50、ResNet-101、ResNet-152五种残差神经网络模型,采用迁移学习方法,对网络模型进行训练以获取模型参数,五种模型测试Top-1准确率分别为85.37%、85.85%、83.90%、81.95%、80.00%, Top-2准确率分别为94.63%、94.63%、94.15%、93.17%、93.66%,模型对牙形刺图像具有较好的识别效果。通过对比研究发现,ResNet-34识别准确率最高,说明对于特征简单的牙形刺属种,增加网络深度并不一定能提升准确率,而确定合适深度的模型则不仅可以提高识别准确率,还可以节约计算资源。通过ResNet-34模型的迁移学习训练和重新训练效果对比可以看出,迁移学习不仅可以获得较高的准确率,而且可以较快获取模型参...     

Evaluating the DNA barcodes for identification of butterflies from Western Himalayas,Uttarakhand, India

DNA barcodes in species tagging have become a popular tool for taking inventories of species from different groups worldwide. The present study aimed to generate DNA barcodes of butterfly species from the Western Himalayas in Uttarakhand, India. The Indian Western Himalayan region (IWHR) has been explored to a limited extent about butterfly species’ diversity. However, the IWHR is prone to environmental change, and slight variations in climatic conditions can influence species diversity and change butterflies' range. The mitochondrial cytochrome c oxidase I (COI) gene was first used to generate the DNA barcode for butterflies from this region on a broad scale. 28 morphologically identified species, consisting of 102 sequences, were finally grouped into 26 species, with only two species showing ambiguity in species identification. These species had < 3% sequence variations from their neighboring relatives, suggesting cryptic species diversity. Generated sequences were also compared with the GenBank data of conspecific geographical locations, which showed intraspecies variation ranging from 1.3% to 7.3%. It was also noted that butterfly species have both intra and interspecies sequence divergence. In the phylogenetic-based species identification, a total of 28 species belonging to 4 families of butterflies were successfully discriminated, and two species at the genus level, which had high intra-specific divergence (0.025), were considered. However, the high intra-species sequence divergence observed may represent the presence of hidden species.     

<Emphasis Type="Italic">Wolbachia</Emphasis> endosymbiont infection in two Indian butterflies and female-biased sex ratio in the Red Pierrot, <Emphasis Type="Italic">Talicada nyseus</Emphasis>

The maternally inherited obligate bacteria Wolbachia is known to infect various lepidopteran insects. However, so far only a few butterfly species harbouring this bacterium have been thoroughly studied. The current study aims to identify the infection status of these bacteria in some of the commonly found butterfly species in India. A total of nine butterfly species belonging to four different families were screened using PCR with Wolbachia-specific wsp and ftsZ primers. The presence of the Wolbachia super group ‘B’ in the butterflies Red Pierrot, Talicada nyseus (Guerin) (Lepidoptera: Lycaenidae) and Blue Mormon, Papilio polymnestor Cramer (Papilionidae), is documented for the first time in India. The study also gives an account on the lifetime fecundity and female-biased sex ratio in T. nyseus, suggesting a putative role for Wolbachia in the observed female-biased sex ratio distortion.     

Quantitative description and discrimination of butterfly wing patterns using moment invariant analysis

Studies examining and using pattern variation in insects for identification and characterization of individuals and populations have been limited by the methods available for quantifying wing patterns objectively. In this paper, differences in wing pattern are demonstrated statistically using moment invariant data sets generated automatically from digitized images of the speckled wood butterfly, Pararge aegeria (Linnaeus). Studies with other biological subjects have already shown moment invariants to work well with outline shapes and silhouettes. A pilot study with replicated monochrome photographs of a single butterfly showed the method could detect pattern differences between wing surfaces, even in the presence of simulated wing fading and damage. In a further study of the wings of 228 specimens, multivariate analyses of variance using the moment data reliably detected differences between groups of butterflies according to sex, geographical origin and culture history. Potential applications and future improvements of the moment methodology are considered.     

Automatic strawberry leaf scorch severity estimation via faster R-CNN and few-shot learning

Using deep learning to estimate strawberry leaf scorch severity often achieves unsatisfactory results when a strawberry leaf image contains complex background information or multi-class diseased leaves and the number of annotated strawberry leaf images is limited. To solve these issues, in this paper, we propose a two-stage method including object detection and few-shot learning to estimate strawberry leaf scorch severity. In the first stage, Faster R-CNN is used to mark the location of strawberry leaf patches, where each single strawberry leaf patch is clipped from original strawberry leaf images to compose a new strawberry leaf patch dataset. In the second stage, the Siamese network trained on the new strawberry leaf patch dataset is used to identify the strawberry leaf patches and then estimate the severity of the original strawberry leaf scorch images according to the multi-instance learning concept. Experimental results from the first stage show that Faster R-CNN achieves better mAP in strawberry leaf patch detection than other object detection networks, at 94.56%. Results from the second stage reveal that the Siamese network achieves an accuracy of 96.67% in the identification of strawberry disease leaf patches, which is higher than the Prototype network. Comprehensive experimental results indicate that compared with other state-of-the-art models, our proposed two-stage method comprising the Faster R-CNN (VGG16) and Siamese networks achieves the highest estimation accuracy of 96.67%. Moreover, our trained two-stage model achieves an estimation accuracy of 88.83% on a new dataset containing 60 strawberry leaf images taken in the field, which indicates its excellent generalization ability.     

High-speed railways are not barriers to Pyronia tithonus butterfly movements

Linear infrastructures such as railways and roads can be barriers to the movements of individuals and, hence, may have strong impacts on populations. We tested the barrier effect of a high-speed railway for Pyronia tithonus, a butterfly species showing homing behaviour when displaced. We captured, marked and displaced 152?individuals in two different locations. One-third of the butterflies were released at a capture plot, one-third on the other side of the railway (in a similar habitat) and one-third on the same side but 100?m away from the capture plot. We obtained recapture rates of 40 and 23?% per location. Many (31?%) butterflies crossed the railway, showing homing behaviour. Thus, contrary to wide, busy roads, high-speed railways do not seem to be barriers for these butterflies. We suggest that in an intensive agrarian landscape, railway verges can play a substitution habitat role for grassland butterflies.     

Identifying Fagaceae and Lauraceae species using leaf images and convolutional neural networks

Lauraceae and Fagaceae are two large woody plant families that are predominant in the low- and middle-altitude regions in Taiwan. The highly interspecific similarity between some species of the family brings limitations on the management and utilization. This work proposed an approach for identifying 15 Lauraceae species and 20 Fagaceae species using leaf images and convolutional neural networks (CNNs). Leaf specimens of 35 species were collected from the northern, central, and southern parts of Taiwan. Images of the leaves were acquired using flat-bed scanners. Three CNN architectures—DenseNet-121, MobileNet V2, and Xception—were trained. Xception achieved the highest mean test accuracy of 99.39%, and MobileNet V2 required the shortest mean test time of 17.1 ms per image using a GPU. The saliency maps revealed that the characteristics learned by models matched the leaf features used by botanists. A pruning algorithm, gate decorator, was applied to the trained models for reducing the number of parameters and number of floating-point operations of the MobileNet V2 by 55.4% and 69.1%, respectively, while the model accuracy was maintained at 92.03%. Thus, MobileNet V2 has the potential to be used for identifying the Lauraceae and Fagaceae species on mobile devices.     

Kitam Bird Sanctuary,the only low elevation protected area of Sikkim: A conservation hotspot for butterflies in the Eastern Himalaya

The lowland areas of the Himalayan region are subjected to immense anthropogenic pressure because of least representation in the protected area network. Kitam Bird Sanctuary is the only representative protected area that occurs below 1000 m in Sikkim state of India (a part of globally significant biodiversity hotspot of Himalayas) and serves as the refuge for various species of flora and fauna. Here we studied butterfly diversity and community composition in Kitam Bird Sanctuary (a small protected area of 6 km² geographical area) following point count method spread across predetermined transects. Altogether 1674 butterflies belonging to 111 species and six families were recorded after completion of 240 point counts. Among these, 18 species are federally protected under the Wildlife (Protection) Act (1972) of India. Most of the butterflies were forest specialist in terms of habitat preference, whereas based on host plant specificity, the butterfly community was mostly dominated by generalist feeder (Oligophagous II and Polyphagous). Butterfly community parameters showed a strong correlation with habitat variables. While Kitam Bird Sanctuary is primarily designated for conservation of lowland birds, the high diversity of butterflies both in terms of taxonomic richness and trait composition suggests that the sanctuary harbors an ideal habitat for butterflies of the tropical region and invites conservation attention.     

BLeafNet: A Bonferroni mean operator based fusion of CNN models for plant identification using leaf image classification

Plants, the only natural source of oxygen, are the most important resources for every species in the world. A proper identification of plants is important for different fields. The observation of leaf characteristics is a popular method as leaves are easily available for examination. Researchers are increasingly applying image processing techniques for the identification of plants based on leaf images. In this paper, we have proposed a leaf image classification model, called BLeafNet, for plant identification, where the concept of deep learning is combined with Bonferroni fusion learning. Initially, we have designed five classification models, using ResNet-50 architecture, where five different inputs are separately used in the models. The inputs are the five variants of the leaf grayscale images, RGB, and three individual channels of RGB - red, green, and blue. For fusion of the five ResNet-50 outputs, we have used the Bonferroni mean operator as it expresses better connectivity among the confidence scores, and it also obtains better results than the individual models. We have also proposed a two-tier training method for properly training the end-to-end model. To evaluate the proposed model, we have used the Malayakew dataset, collected at the Royal Botanic Gardens in New England, which is a very challenging dataset as many leaves from different species have a very similar appearance. Besides, the proposed method is evaluated using the Leafsnap and the Flavia datasets. The obtained results on both the datasets confirm the superiority of the model as it outperforms the results achieved by many state-of-the-art models.     

Butterfly species diversity and their floral preferences in the Rupa Wetland of Nepal

Floral attributes often influence the foraging choices of nectar‐feeding butterflies, given the close association between plants and these butterfly pollinators. The diversity of butterflies is known to a large extent in Nepal, but little information is available on the feeding habits of butterflies. This study was conducted along the periphery of Rupa Wetland from January to December 2019 to assess butterfly species diversity and to identify the factors influencing their foraging choices. In total, we recorded 1535 individuals of 138 species representing all six families. For our examination of butterfly–nectar plant interactions, we recorded a total of 298 individuals belonging to 31 species of butterfly visiting a total of 28 nectar plant species. Overall, total butterfly visitation was found to be significantly influenced by plant category (herbaceous preferred over woody), floral color (yellow white and purple preferred over pink), and corolla type (tubular preferred over nontubular). Moreover, there was a significant positive correlation between the proboscis length of butterflies and the corolla tube length of flowers. Examining each butterfly family separately revealed that, for four of the families (Lycaenidae, Nymphalidae, Papilionidae, and Pieridae), none of the tested factors (flower color, plant category, and corolla type) were shown to significantly influence butterfly abundance at flowers. However, Hesperidae abundance was found to be significantly influenced by both flower color (with more butterflies observed at yellow flowers than purple) and flower type (with more butterflies observed at tubular flowers than nontubular flowers). Our results reveal that Rupa Lake is a suitable habitat for butterflies, providing valuable floral resources. Hence, further detailed studies encompassing all seasons, a greater variety of plants, and other influential factors in different ecological regions are fundamental for creating favorable environments to sustain important butterfly pollinators and help create balanced wetland ecosystems.     

Determinants of the biogeographical distribution of butterflies in boreal regions

Aim  We explored the relative contributions of climatic and land-cover factors in explaining the distribution patterns of butterflies in a boreal region.
Location  Finland, northern Europe.
Methods  Data from a national butterfly atlas survey carried out during 1991–2003, with a 10-km grain grid system, were used in these analyses. We used generalized additive models (GAM) and hierarchical partitioning (HP) to explore the main environmental correlates (climate and land-cover) of the realized niches of 98 butterfly species. The accuracy of the distribution models (GAMs) was validated by resubstitution and cross-validation approaches, using the area under the curve (AUC) derived from the receiver operating characteristic (ROC) plots.
Results  Predictive accuracies of the 98 individual environment–butterfly models varied from low to very high (cross-validated AUC values 0.48–0.99), with a mean of 0.79. The results of both the GAM and HP analyses were broadly concordant. Most of the variation in butterfly distributions is associated with growing degree-days, mean temperature of the coldest month and cover of built-up area in all six phylogenetic groups (butterfly families). There were no statistically significant differences in predictive accuracy among the different butterfly families.
Main conclusions  About three-quarters of the distributions of butterfly species in Finland appear to be governed principally by climatic, predominantly temperature-related, factors. This indicates that many butterfly species may respond rapidly to the projected climate change in boreal regions. By determining the ecological niches of multiple species, we can project their range shifts in response to changes in climate and land-cover, and identify species that are particularly sensitive to forecasted global changes.     

Automatic recognition and measurement of butterfly eyespot patterns

A favorite wing pattern element in butterflies that has been the focus of intense study in evolutionary and developmental biology, as well as in behavioral ecology, is the eyespot. Because the pace of research on these bull's eye patterns is accelerating we sought to develop a tool to automatically detect and measure butterfly eyespot patterns in digital images of the wings. We used a machine learning algorithm with features based on circularity and symmetry to detect eyespots on the images. The algorithm is first trained with examples from a database of images with two different labels (eyespot and non-eyespot), and subsequently is able to provide classification for a new image. After an eyespot is detected the radius measurements of its color rings are performed by a 1D Hough Transform which corresponds to histogramming. We trained software to recognize eyespot patterns of the nymphalid butterfly Bicyclus anynana but eyespots of other butterfly species were also successfully detected by the software.     

Butterfly diversity in Mediterranean islands and in Pentadaktylos <Emphasis Type="Italic">Pinus</Emphasis><Emphasis Type="BoldItalic"> </Emphasis><Emphasis Type="Italic">brutia</Emphasis> forests of Cyprus

We analysed the influence of contemporary geography on butterfly diversity for islands in the Mediterranean Basin. We found that island size and distance from the mainland has a significant effect on the number of species. We also used butterflies as an indicator group to identify the importance of forest habitats for biodiversity conservation in the island of Cyprus. To understand the relative importance of local vegetation characteristics of butterflies in the Pentadaktylos mountains transect counts were used to assess the abundance and butterfly diversity in two different forest types. A total of 1,602 butterflies and 23 species were recorded during this research. We observed highly significant effects of forest type on abundance and species richness of butterflies. For example, number of butterflies was significantly higher in old forest than young pine forest. Also, the abundance of endemic butterflies was highest in old forest habitats. Therefore, the survival of the majority of endemic butterflies in Cyprus may depend on conservation of old forests and their understorey plants.     

基于树木年轮径切特征的卷积神经网络树种识别

卷积神经网络可以通过树木年轮样本构造特征图像实现物种识别的自动化。本研究通过建立树木年轮样本构造特征图像集,选用LeNet、AlexNet、GoogLeNet和VGGNet 4个卷积神经网络模型,实现基于树木年轮横切面的计算机自动化树种精准识别,进而确定各模型的树种识别准确率,明晰不同树种在自动识别中的混淆情况,探测不同模型识别结果的差异。结果表明: 本研究训练的用于树种识别的卷积神经网络模型具有较好的可信度;4个模型中GoogLeNet模型树种识别准确率最高,为96.7%,LeNet模型识别准确率最低(66.4%);不同模型对于所选树种的识别结果具有一致性,表现为对蒙古栎识别准确率最高(AlexNet模型识别率达到100%),对臭冷杉的识别准确率最低。本研究中也存在类似结构树种的识别混淆情况。模型在科和属水平的识别准确率高于种水平;阔叶树种因其显著的结构差异容易区分,阔叶树树种的识别准确率高于针叶树。总体上,通过卷积神经网络,探测了树木年轮特征的深层信息,达到树种的精准识别,提供了一种快速便捷的自动树种初筛鉴定方法。