基于多层感知神经网络的水稻叶瘟病识别方法

为实现水稻叶瘟病的快速诊断,综合利用图像处理技术和神经网络来进行叶瘟病斑的识别研究。该文设计了3个多层感知分类器来进行病斑识别准确率的对比验证,分别采用叶片正常区域和病斑区域的纹理特征、颜色特征以及纹理和颜色的组合特征作为不同分类器的输入单元;输出层采用1个单元用于输出病斑区域和正常区域的识别结果。首先,该文将采集到的RGB图像转换成灰度图像,利用灰度共生矩阵分别提取叶片正常区域与病斑区域的能量、对比度、熵、逆差距作为纹理特征;紧接着,将RGB彩色空间转换至HIS和Lab空间,分别提取病斑区域和正常区域的L、a、b值作为颜色特征。最后,采用不同的BP神经网络分类器进行病斑区域识别。该文共采用120副图像作为待测对象,试验结果表明,采用颜色和纹理的组合特征进行识别,准确率要比单独使用纹理特征和颜色特征高10%～15%。本文的研究结果为进一步实现水稻病害自动诊断打下了基础。     

Constructing naive Bayesian classifiers for veterinary medicine: A case study in the clinical diagnosis of classical swine fever

For diseases of which the clinical diagnosis is uncertain, naive Bayesian classifiers can be of assistance to the veterinary practitioner. These simple probabilistic models have proven to be very powerful for solving classification problems in a variety of domains, but are not yet widely applied within the veterinary domain. In this paper, naive Bayesian classifiers and methods for their construction are reviewed. We demonstrate how to construct full and selective classifiers from a data set and how to build such classifiers from information in the literature. As a case study, naive Bayesian classifiers to discriminate between classical swine fever (CSF)-infected and non-infected pig herds were constructed from data collected during the 1997/1998 CSF epidemic in the Netherlands. The resulting classifiers were studied in terms of their accuracy and compared with the optimally efficient diagnostic rule that was reported earlier by Elbers et al. (2002). The classifiers were found to have accuracies within the range of 67-70% and performed comparable to or even better than the diagnostic rule on the available data. In contrast with the diagnostic rule, the classifiers had the advantage of taking both the presence and the absence of particular clinical signs into account, which resulted in more discriminative power. These results indicate that naive Bayesian classifiers are promising tools for solving diagnostic problems in the veterinary field.     

基于遥感的塞罕坝机械林场树种分类研究

随着遥感应用的不断推广,各应用领域对遥感图象的监督分类结果的精度要求越来越高。在监督分类中有两个非常关键的问题:一是训练样本选取;二是分类器选择。本论文应用监督分类中的不同分类器对塞罕坝机械林场进行树种分类,对其结果进行分析比较,选择合适的分类器提高分类精度。     

面向对象的丘陵区水田遥感识别方法

中国南方丘陵区地形破碎,地物分布复杂,丘陵区水田的光谱特征相对于平原区较混杂,传统的基于像元的遥感数据获取受异质性因素的影响,无法利用单一时(季)像及特定的图像自动识别规则提取精度较高的水田分布信息。针对这一问题,该文基于多时像HJ-1A/1B卫星图像,结合地面调查,以湖南省湘潭市为研究区,在易康(e Cognition)软件平台上分别以光谱特征为主要参考的多层最邻近分类法和以在特征知识库支持下的决策树分类法对丘陵区水田进行图像识别。结果表明:分层最邻近分类法比单一最邻近分类提取的精度高,但在相同分割尺度下,利用特征知识库支持下的决策树分类提取水田的精度达到最高,为90.25%,总Kappa系数为0.79,说明特征知识库支持下的决策树分类方法比最邻近分类法更加适合丘陵区水田的遥感识别。     

Quality assessment of durum wheat storage centres in Sicily: Evaluation of vitreous, starchy and shrunken kernels using an image analysis system

The proportion of vitreous kernels in a sample is an internationally recognized specification for determining the value of durum wheat (Triticum durum Desf.). Vitreous kernels are mostly related to quality, which affects the pasta performance during cooking. Vitreousness and the amount of shrunken kernels are visually assessed during the grading process. This assessment is subjective and tedious.A machine vision system was developed to determine the percentage of vitreous, starchy, piebald and shrunken kernels in approximately 100 grain samples, using a trans-illuminated image of one layer of non-singulated kernels (in bulk) acquired by a digital camera. Classification models were developed with stepwise Linear Discriminant Analysis, as well as an on-line Bayesian classifier integrated with an image analysis system. The overall correct classification in Starchy classifier was high 98.58% in the Training set, made up of 6679 grains, following the Linear Discriminant Analysis classification, of 30 Italian cultivars harvested in 2005 in three localities. An independent Test set was constituted by samples collected in 30 Sicilian Storage Centres in the 2007 harvest season. The overall classification was 96.03%. For the Shrunken classifier 95.27% of the Training set and 99.58% of the Test set were correctly classified. The image analysis system was more reliable than the human inspectors who validated the system, both for the same samples measured many times and at different times.     

花生脱壳机脱出物的漂浮系数试验

为合理设计花生气力分选和二次脱壳气力输送装置,以辽宁主栽花生品种为研究对象,进行了花生脱壳机脱出物主要成分以及杂质等空气动力特性试验,得到了破损花生荚果、未脱净而需要二次脱壳的花生荚果、花生米、花生壳、石头的漂浮系数分别为0.168～0.246、0.102～0.146、0.080～0.186、3.287～6.037、0.031～0.045m-1。结果表明,花生荚果、破损花生荚果、花生米、花生壳和石头之间的漂浮速度的差异较大,有利于气力分选,可作为花生脱壳机分选装置和气力输送装置设计的重要参考依据。     

基于微波遥感极化目标分解的土地覆盖/土地利用分类

为有效利用微波遥感影像进行土地覆盖/土地利用分类,该研究以内蒙古河套灌区解放闸灌域为研究区域,采用春耕后试验区Radarsat-2全极化数据,利用极化目标分解方法提取得到了散射熵、平均散射角、反熵、平均特征值、单次反射特征值相对差异度、二次反射特征值相对差异度。结合实地数据,分析了各参数对于耕地、裸地、含植被水体、建筑等类别的可分离性。根据分析结果选取平均散射角、平均特征值、单次反射特征值相对差异度为分类特征变量,通过最小距离法计算了决策边界,最后结合树分类器对试验区影像进行了分类。整体分类精度93.89%,分类Kappa系数为0.914。结果表明,利用平均散射角可有效区分表面散射与二次散射及体散射;平均特征值可有效区分含植被水体与建筑物;单次反射特征值相对差异度参数可有效区分耕地与裸地。利用极化目标分解方法结合决策树分类器可精确地进行土地覆盖/土地利用分类。     

基于多分类器组合的个人信用评估模型

针对个人信用评估单一模型存在的不足，提出一种基于多分类器组合的个人信用评估模型。该模型综合了多元判别分析、logistic回归、神经网络、支持向量机等七种个人信用评估单一模型的预测结果，利用加权投票方法对其进行组合并输出最后预测结果。在某商业银行信用卡数据集上的测试结果表明，组合模型能有效地提高预测精度及稳健性，对信贷机构控制消费信贷风险具有很好的适用性。     

基于多分类器融合的玉米叶部病害识别

针对单分类器识别的局限性和玉米叶部病害的复杂性,该文提出了一种基于自适应加权的多分类器融合的玉米叶部病害识别方法。首先,对采集的玉米叶部病害图像的病害区域分别提取颜色、颜色共生矩阵和颜色完全局部二值模式3种特征,并相应地构建3个基于支持向量机的单分类器;然后,利用K近邻和聚类分析的方法计算各单分类器的自适应动态权值;最后,通过线性加权的方式进行融合判决,得到最终的分类结果。利用该方法对7种常见的玉米叶部病害图片进行了试验,平均识别率达94.71%。结果表明,其性能优于目前常见的单一特征或特征组合构建的同类分类器及多分类器融合方法。研究结果为其他农作物病害诊断提供了借鉴和参考。     

基于光谱技术和多分类器融合的异物蛋检测

为了提高鸡蛋中的血斑和肉斑的检测准确率,给消费者提供高品质的鸡蛋,该文利用微型光纤光谱仪采集鸡蛋的透射光谱,在单分类器的基础上,通过多分类器的融合对异物蛋进行检测。首先根据差异性度量选取朴素贝叶斯,Ada Boost和SVM分类器作为单分类器,然后通过特征级融合选取了5个基分类器。最后,5个基分类器以加权投票机制进行决策级融合。多分类器融合对正常蛋和异物蛋的检测准确率分别为92.86%和91.07%。试验结果表明,利用多分类器融合所建立的模型优于单一分类器的模型,提高了对异物蛋的检测准确率。