基于HSV与形状特征融合的花椒图像识别

光照条件变化会对花椒目标识别率产生影响,关系到机器视觉技术能否有效应用于生产现场的花椒采摘。通过对HSV特性图像识别技术的分析,提出HSV和形状特征融合的花椒识别算法。该算法采用同态滤波方法对光照进行补偿,解决因为光照不均匀而导致的花椒识别率低的问题,最后利用花椒圆度特征,排除树枝及树叶等的干扰,实现花椒的准确识别。利用同态滤波方法对光照进行补偿,对于光照不强或者发生遮挡的花椒图像有较大改善,通过试验得出其平均识别率达到94.0%,比单独采用HSV特性识别时,在顺光,背光和遮阴条件下,识别率分别提高4%,13%和21%,此外在遮阴条件下运行时间缩短14.6%。为遮阴条件下提高花椒识别率提供一种方法。     

基于随机森林的高寒湿地地区土地覆盖遥感分类方法

高寒湿地是青藏高原典型独特的生态系统,是全球气候变化的敏感地带和预警区。利用遥感技术快速、准确地分类提取高寒湿地的土地覆盖信息,对当地生态安全监测和保护具有重要意义。本文以若尔盖湿地国家级自然保护区为研究区,首先,以高分一号(GF-1)遥感影像为数据源,融合光谱特征、水体指数、地形特征、植被指数和纹理信息等26个变量进行随机森林(Random forest,RF)分类实验;然后,根据袋外数据(Out of bag,OOB)的特征变量重要性得分和精度评价结果,选出高寒湿地地区土地覆盖类型的最优分类方案和特征;最后,对特征变量进行降维,并基于相同的变量,采用极大似然法(Maximum likelihood classification,MLC)、支持向量机(Support vector machine,SVM)、人工神经网络(Artificial neural network,ANN)和RF等方法进行分类,比较不同方法的优适性。结果表明:结合GF-1影像光谱、水体、植被、纹理特征和地形信息,使用26个变量的RF模型的分类精度最高,总体精度(Overall accuracy,OA)为90.07%,Kappa系数为0.86;通过RF模型的变量重要性分析可以有效选出重要的特征信息,在降低特征变量维度的同时,还能保证较高的分类精度; 4种分类方法中,RF算法是高寒湿地地区较合适的分类方法,OA比MLC基准方法高17.63个百分点,比SVM和ANN等机器学习算法分别高6.98、6.56个百分点。     

基于机器学习和环境因子的温室网纹甜瓜纹理特征的预测

为了对网纹甜瓜纹理特征进行定量预测,本文结合多个环境因子(基质含水量、温度、湿度、光合有效辐射),使用支持向量机(SVM)与随机森林(RF)多元回归分析方法,对基于灰度共生矩阵提取的4个果实表面纹理特征——对比度(contrast)、熵(entropy)、相关性(correlation)、角二阶矩(ASM)进行了预测。结果表明,纹理特征受环境影响敏感度依次为基质水分、光合有效辐射、空气湿度或有效积温。对比RF模型和SVR模型,发现在不同环境因子输入条件下,RF模型均优于SVM模型,其中输入全部环境因子后RF模型的预测精度最高,对比度、熵、角二阶矩的模拟精度均达到了0.90,分别为R~2=0.945(RMSE=0.243)、R~2=0.940(RMSE=0.235)、R~2=0.934(RMSE=0.248)。上述结果表明,RF模型对于网纹甜瓜纹理特征具有较好的预测结果,本研究结果可为温室网纹甜瓜栽培过程中的质量监控和栽培管理提供全新的思路。     

翅形态特征在重阳木斑蛾成虫性别鉴定中的应用

为了明确重阳木斑蛾的翅形态特征在其性别鉴定中的作用,探索雌雄难于区分的成虫的数值鉴定方法。采用图像处理与分析技术,通过爱普生扫描仪获取重阳木斑蛾前翅图,利用BugShape v1.0软件提取雌雄成虫翅的轮廓特征,利用tpsDig2软件获取翅脉交叉点位置特征,并利用独立样本t检验和标志点空间普氏叠加和扭曲分析,以明确此类特征在该成虫雌雄鉴定中的作用。结果表明,重阳木斑蛾成虫左右前翅轮廓特征在同一性别中差异不显著。仅利用左前翅数据进行分析时,翅面积、短轴长度、等效圆半径、偏心率、球状性和圆形度在两个性别中存在极显著差异,周长和长轴长度的差异达到显著水平,而紧凑度和叶状性差异不显著。使用左右前翅作为数据集进行分析时,除紧凑度和叶状性差异仍不显著外,其余参数均达到极显著水平。雌雄前翅轮廓特征作为指标来判别性别时正确率为100%,但存在左右翅相互错判现象。翅膀内部翅脉交叉点的空间位置在个体间存在较大差异,主要表现在相对于翅的横向方向变异大。雌雄间的差异主要存在于外缘附近的翅脉交叉点。结论:通过前翅的轮廓特征可以区分重阳木斑蛾雌雄成虫的差异。翅脉交叉点空间位置在雌雄间和个体间存在一定的变异。     

Identification of an H1N1 subtype of swine influenza virus and serological analysis

To investigate the epizootic of swine influenza virus(SIV), 60 nasal swabs were collected from a clinical cases of pig farm in Tai'an City, Shandong Province of China in April 2017. SIV was isolated by inoculating into 10-day-old Special Pathogen Free embryonated eggs and the whole genome was sequenced. An H1N1 subtype SIV was isolated and designated as A/swine/Shandong/TA04/2017(H1N1). Phylogenetic analysis showed that apart from the polymerase A(PA) fragment belonging to the 2009 pandemic H1N1 branch, seven genome segments belonged to avian-like H1N1 influenza virus lineage. The cleavage site sequence of the hemagglutinin(HA) protein was PSIQSR↓G, which is a typical molecular biological characteristic. Five potential N-glycosylation sites(N14, N26, N277, N484 and N543) were found in the HA gene. To further investigate the epidemiology of SIV in this farm, the 995 serum samples were assessed with EAH1N1 2009 pandemic H1N1 and H3 N2 antigens. The results showed that the total positive rate was 65.43%. The positive rates of single virus infection detected by EAH1N1, 2009 pdmH1N1 and H3 N2 for serum HI(Hemagglutination inhibition) were 48.35, 30.85 and 7.47%, respectively. The results showed that SIV in Shandong Province has been reassorted in some segments and the SIV-positive rate was high on the SIV outbreak farm. These data provide evidence of an epizootic of SIV.     

基于注意力机制及多尺度特征融合的番茄叶片缺素图像分类方法

针对番茄早期缺素性状不明显及各生长期特征差异较大所导致的特征区域尺寸不一致、难提取、难辩别等问题，提出了一种基于注意力机制及多尺度特征融合卷积神经网络的番茄叶片缺素图像分类方法（Multi-Scale Feature Fusion Convolutional Neural Networks Based On Atte ntion Mechanism，MSFF-AM-CNNs）。首先根据番茄叶片缺素特点提出了多尺度特征融合结构（Multi-Scale Feature Fusion Module，MSFF Module）；其次在DenseNet基础上，结合浅层网络主要提取纹理、细节特征，深层网络主要提取轮廓、形状特征的特点分别提出具有针对性的特征提取方法，通过不同形式引入注意力机制及多尺度特征融合结构，使全局多尺度信息融合多个特征通道、选择性地强调信息特征并达到对特征精准定位的功能；同时引入Focal Loss函数以减少易分类样本的权重。试验结果表明，MSFF-AM-CNNs的平均召回率、平均F1得分、平均准确率较原模型DenseNet-121均大幅提升，其中缺氮和缺钾叶片的准确率分别提高了8.06和6.14个百分点，召回率分别提高了6.31和5.00个百分点，F1得分分别提高了7.25和5.55个百分点，平均识别准确率可达95.92%，具有较高的识别准确率及广泛的适用性，能够满足番茄叶片缺素图像的高精度分类需求，可为植物叶片缺素识别提供参考。     

Comparison of estimators and feature selection procedures in forest inventory based on airborne laser scanning and digital aerial imagery

Digital maps of forest resources are a crucial factor in successful forestry applications. Since manual measurement of this data on large areas is infeasible, maps must be constructed using a sample field data set and a prediction model constructed from remote sensing materials, of which airborne laser scanning (ALS) data and aerial images are currently widely used in management planning inventories. ALS data is suitable for the prediction of variables related to the size and volume of trees, whereas optical imagery helps in improving distinction between tree species. We studied the prediction of forest attributes using field data from National Forest Inventory complemented with ad hoc field plots in combination with ALS and aerial imagery data in Aland province, Finland. We applied feature selection with genetic algorithm and greedy forward selection and compared multiple linear and nonlinear estimators. Maximally around 40 features from a total of 154 were required to achieve the best prediction performances. Tree height was predicted with normalized root mean squared error value of 0.1 and tree volume with a value around 0.25. Predicting the volumes of spruce and broadleaved trees was the most challenging due to small proportions of these tree species in the study area.     

How do hedgerows influence soil organic carbon stock in livestock‐grazed pasture?

Hedgerows have the potential to influence ecosystem function in livestock‐grazed pasture. Despite this, they are often ignored when quantifying farmland ecosystem service delivery. In this study, we assess the contribution of hedgerows to the ecosystem function of carbon (C) storage, with a particular emphasis on soil organic carbon (SOC). We measured SOC stock (kg C m^?2), on an equivalent soil mass basis, at 0–0.15 m depth in pasture adjacent to 38 hedgerows (biotic) and 16 stone walls or fences (abiotic controls) across ten farms in the county of Conwy, Wales, UK. Pasture SOC stock (~7 kg C m^?2) was similar adjacent to biotic and abiotic field boundaries, positively associated with soil moisture and negatively with soil bulk density (BD). For biotic boundaries, two further variables were significantly associated with SOC stock, distance from hedgerow (decrease in SOC stock) and slope orientation (upslope SOC stock greater than downslope). For pasture adjacent to hedgerows, a model combining the aforementioned variables (BD, soil moisture, distance from hedgerow, slope orientation) explained 78% of variation in SOC stock. This study demonstrates that whilst hedgerows do have subtle positive effects on SOC stock in adjacent pasture, SOC storage adjacent to field boundaries is influenced more by soil moisture content and BD than field boundary type.     

Detect unique molecular structure associated with physiochemical properties in CDC varieties of oat grain with unique nutrient traits [Feed Type vs. Milling Type] in comparison with barley grain using advanced molecular spectroscopy as a non-destructive biological tool

The objective of this study was to determinate grain unique protein inherent molecular structure that are related physiochemical and nutrient profiles in CDC developed oat varieties [CDC Nasser (Feed Type) and CDC Seabiscuit (Milling Type)] grown in cool climate condition in western Canada in comparison with conventional barley variety of CDC Meredith as a control using advanced molecular spectroscopy. Multivariate analyses, including an agglomerative hierarchical cluster analysis (CLA) and principal component analysis (PCA), were performed to identify protein molecular structural differences among the grains. The results revealed that CDC Seabiscuit contained greater (P < 0.05) protein structural Amide I and II than CDC Nasser and CDC Meredith, while the greater (P < 0.005) structural Amide I to II area and height ratios was detected in CDC Meredith. New oat grains had greater (P < 0.05) β-sheet height than barley grains, however, there was no difference in α-helix to β-sheet ratio values among the varieties. In conclusion, CDC Nasser and CDC Meredith had no difference in protein molecular structural features, while CDC Seabiscuit contains different protein structural characteristics as compared to CDC Meredith grain. The molecular structure features are highly associated with physiochemical and nutrient profiles in grains, which indicate that it also affect nutrient utilization and availability.     

基于机器视觉的马铃薯晚疫病快速识别

晚疫病是马铃薯的一种严重病害,可造成减产甚至绝收。因此马铃薯晚疫病的识别与控制对提高其产量有非常重要的意义。该文基于机器视觉技术对马铃薯叶部晚疫病进行检测,根据马铃薯叶片上晚疫病斑的颜色、纹理和形状特征参数的不同,提取叶片表面的特征参数,并建立数学模型对病害程度做出评价。在RGB、HSV颜色空间中,根据马铃薯叶片在患病早期叶片颜色发生变化且与健康叶片不同,利用颜色特征,建立马铃薯晚疫病的无病和患病模型,该模型对马铃薯患病早期的识别率为67.5%。利用灰度共生矩阵,采用纹理统计参数进行病害等级评价,用熵值和能量值描述晚疫病的严重程度,纹理特征对患病程度的识别率比较稳定,对患病中期与后期的识别率分别为72.5%与80%。利用形状特征的相对特征,根据病斑面积比进行晚疫病诊断,该方法对马铃薯叶片晚疫病患病后期的诊断取得较好效果,识别率为90%,但由于叶片患病早期的病斑面积小且分散,识别难度大,识别率仅为50%。针对颜色、纹理及形状特征在识别马铃薯叶片晚疫病时的优势与局限性,提出颜色纹理形状特征结合的识别方法,对患病中期与后期的识别率分别为90%和92.5%。通常马铃薯晚疫病的理化值检测法耗时数天,但利用机器视觉识别马铃薯晚疫病患病情况非常快速,根据颜色特征进行病害识别的时间约为4 s,纹理特征识别的时间为7 s,形状特征特征识别的时间为3 s,综合颜色纹理形状特征的识别由于计算量较大,识别时间为9 s。该研究可为基于机器视觉的马铃薯晚疫病的快速检测提供理论依据。