支持向量机在斑潜蝇虫害叶片光谱分析中的应用

测量了斑潜蝇虫害叶片的可见光反射光谱。运用图像处理技术计算出虫害叶片的破损率;采用光谱分析方法选出对叶片破损程度敏感的波长点。运用支持向量机(SVM)和多光谱法,建立基于敏感波长的受害叶片反射光谱分类模型并进行分类试验。试验结果表明,多光谱法分类精度为90%,SVM为93.8%(多项式核函数)和96.9%(RBF核函数),高于多光谱法。     

The Study of lmage Compression on JPEG ISO

The JPEEG algorithm for Stillimage ISO Encoding is introduced in this paper.Three encoding types for JPEG algorithm are discussed.Anew Fast DCT algorithm is proposed inthe paper. A software for JPEG algorithm on baseline code is developmented; The experimenceshows that the results are eatisfactory.     

Stepwise Decomposition Algorithm Research of the Intersecting Body Volume of Irregular Polyhedrons

We integrate the problems of deduction of intersecting grooves and bases in the constructional engineering into the problem calculating the intersecting body volume of irregular polyhedrons, and put forward the stepwise decomposition algorithm of the intersecting body volume of irregular polyhedrons based on computer graphics. The method takes advantage of the basic algorithms in the computer graphics and improves one of the algorithms which can be used to judge whether the dot is in the plane or not. By analyzing the method, we have developed the calculation formulae. Winth the segmentation function, the stepwise decomposition algorithm can be used to compute the intersecting body volume not only of the convex polyhedrons, but also the concave polyhedrons. By programming, the stepwise decomposition algorithm has been carried out, and its accuracy has been testified by a typical example.     

作物形态诊断人工神经网络专家系统的研究

本文使用计算机图像识别系统对作物形态进行自动提取，并设计了一种用于形态诊断的人工神经网络专家系统，它较好地解决了传统专家系统所面临的知识获取和表达、系统扩充困难等问题。达到计算机对作物形态性状信息自动提取和智能诊断之目的，为实现作物因苗管理、分类指导、提高科学管理水平提供技术支持。     

基于形状因子和分割点定位的粘连害虫图像分割方法

单个害虫的分割是进行害虫特征提取和识别的前提。针对害虫识别过程中出现的粘连等问题,提出了一种基于形状因子和分割点定位的害虫图像分割方法。该方法首先利用形状因子对图像中的每个区域进行粘连判定,然后对判定为粘连的区域进行逐层轮廓剥离和局部分割点的确定,接着根据局部分割点在原区域中搜索边界轮廓的两个分离点,最后连接局部分割点与分离点线段进行害虫分割。通过实验室人工随机散落桃蛀螟Conogethes punctiferalis(Guenée)和田间粘虫板诱捕梨小食心虫Grapholitha molesta(Busck)2种场景采集图像,验证算法的有效性,并与分水岭分割算法进行对比,采用分割率、分割错误率和分割有效性3项指标进行评价,结果表明:针对实验室环境下采集的2组桃蛀螟害虫图像,该文方法平均错误率为7%,约为分水岭分割方法的1/2,平均分割有效率为92.65%,比分水岭算法提高了5.7个百分点;在2组田间梨小食心虫图像分割中,该文方法平均错误率为2.24%,平均分割有效率为97.8%,分别比分水岭方法降低了4.29个百分点和提高了3.95个百分点,说明该文方法在分割准确性和有效性方面都可以获得更好的分割性能,应用于害虫多目标分割与自动识别系统中,可以有效地提高识别精度。     

Pampas deer conservation with respect to habitat loss and protected area considerations in San Luis, Argentina

Ozotoceros bezoarticus celer is the most endangered subspecies of pampas deer. Although common in the Argentine Pampas 100 years ago, it persists in only two small populations. The largest population has survived due to the rarity of roads, internal farm subdivisions, and the low cattle density. However, habitat condition for this population has changed dramatically in the last 16 years. Five Landsat images (1985, 1992, 1997, 1999, 2001), covering 4608 km², were used to quantify pampas deer habitat loss due to the replacement of natural grassland by exotic pastures and crops. Image classification showed that natural grassland cover was reduced from 84.5 to 37.8% between 1985 and 2001. The annual transformation rate increased significantly from 1.4 to 10.9%. Average paddock size was significantly reduced from 1470 to 873 ha, and the number of paddocks increased from 129 to 227. The land within this area proposed for a national park has not escaped these habitat changes. In the last 6 years the amount of replaced area within the proposed park has increased from 9.1 to 51.1% due to actions by ranchers to avoid inclusion within park boundaries. Three patches of natural grassland still remain within the pampas deer distribution, one of which is the proposed national park. The implementation of a national park is a decisive challenge for the survival of pampas deer and its habitat in Argentina.     

用彩色图像识别田间洋葱和杂草的研究

减少化学产品(化肥、农药等)的用量，是精确农业的主要目标之一。人们力图通过图像分析，测量田间一些地块的杂草和作物的比例并确定杂草的位置，但是在彩色图像里，作物叶子和杂草叶子的交叉给计算机区分它们带来了困难。文中针对两种不同情况，提出了两种不同的识别方法。一种方法是针对没有作物叶子和杂草叶子交叉的情况，另一种方法是针对有某些交叉的情况提出的。前一种方法计算时间短，后一种方法需要大量计算。还提出在图像分割过程中，用不同特征法和两步法进行区域合并的模式，其不同特征包括：区域像素各颜色成分的平均值，区域边界的比值和区域的形状。     

农产品质量的计算机辅助检验与分级（第Ⅰ报）——烟叶外观品质特征的定量检验

应用计算机图象处理技术和色度学理论，研究了烟叶外观品质特征（颜色、叶形、叶面等）的提取、描述与定量检测方法，为开发烟叶外观品质的检验仪器与分级设备提供了理论依据     

机器人采摘苹果果实的K-means和GA-RBF-LMS神经网络识别

为进一步提升苹果果实的识别精度和速度,从而提高苹果采摘机器人的采摘效率。提出一种基于K-means聚类分割和基于遗传算法(genetic algorithm,GA)、最小均方差算法(least mean square,LMS)优化的径向基(radial basis function,RBF)神经网络相结合的苹果识别方法。首先将采集到的苹果图像在Lab颜色空间下利用K-means聚类算法对其进行分割,分别提取分割图像的RGB、HSI颜色特征分量和圆方差、致密度、周长平方面积比、Hu不变矩形状特征分量。将提取的16个特征作为神经网络的输入,对RBF神经网络进行训练,以得到苹果果实的识别模型。针对RBF神经网络学习率低、过拟合等不足,引入遗传算法对RBF隐层神经元个数和连接权值进行优化,采取二者混合编码同时进化的优化方式,最后再利用LMS对连接权值进一步学习,建立新的神经网络优化模型(GA-RBF-LMS),以提高神经网络的运行效率和识别精度。为了获得更精确的网络模型,在训练过程中,苹果果实连同树枝、树叶一块训练;得到的模型在识别过程中,可一定程度上避免枝叶遮挡对果实识别的影响。为了更好地验证新方法,分别与传统的BP(back propagation)和RBF神经网络、GA-RBF优化模型比较,结果表明,该文算法对于遮挡、重叠果实的识别率达95.38%、96.17%,总体识别率达96.95%;从训练时间看,该文算法虽耗时较长,用150个样本进行训练平均耗时4.412 s,但训练成功率可达100%,且节省了人工尝试构造网络结构造成的时间浪费;从识别时间看,该文算法识别179个苹果的时间为1.75 s。可见GA-RBF-LMS网络模型在运行效率和识别精度较优。研究结果为苹果采摘机器人快速、精准识别果实提供参考。     

图像处理技术用于农机-土壤系统的研究

提出一种用于分析土壤工作部件扰动土壤孔隙变化的图像处理方法。主要步骤为：处理样品的制备；图像获取和孔隙识别；土壤孔隙形态学参数的计算。并给出一个用于研究鼠道犁扰动土壤的实例。结果表明图像处理技术用于研究农机－土壤系统是有效的。