基于YCbCr颜色空间和Fisher判别分析的棉花图像分割研究

棉花的分割是采棉机器人研究的关键技术。本文分别在HSV、HIS和YCbCr颜色空间下,首先根据棉花的颜色信息与背景颜色信息的差距,对样本图像中的各个对象(棉絮、棉枝、土壤等)分类; 其次根据分类结果分别提取各类在各颜色空间下的样本像素值; 再根据类间离散度最大和类内离散度最小的准则计算出Fisher判别向量和各类的质心; 最后按照像素值离各质心最近的准则进行图像分割。结果表明, 在YCbCr颜色空间下产生的分割噪声最小,选取此颜色空间,采用贴标签的方法自适应去噪。实验仿真表明,本方法可有效避免阳光直射和阴影的干扰,对各种情况都能准确分割,分割准确率达90.44%。     

基于形态学的黑背景下收获前棉花图像自动分割技术研究

以棉田实地采集的黑背景下收获前含铃壳棉花图像为样本(7级×9张/级),首先,对原始灰度图二值化以提取中心模板,进一步进行形态学高帽剪切预处理并提取其外围二值模板,两者叠加完成背景分割;试验结果表明:98.4%的含铃壳棉花二值模板图像被准确提取。其次,用形态学低帽剪切预处理后的二值图执行开启运算,以判别棉瓣间的连通性:⑴对于多连通的低品级棉花,直接提取未开启二值图中不超过4个区域的棉瓣二值模板,⑵对于单连通的高品级棉花,重新对原始灰度图二值化并提取最大区域的棉瓣二值模板;实验证明:基于棉瓣面积明显大于铃壳面积的棉瓣抽取算法是健壮的,74.6%的棉瓣二值模板图像被准确提取。     

基于HSV彩色模型的自然场景下棉花图像分割策略研究

 棉花成熟度和空间位置的识别是采棉机器人研究的关键技术,解决此问题必须对棉花图像进行分割。选取HSV模型中与亮度无关的S通道作为棉花图像的特征,排除了图像明暗变化对分割效果的影响。文中分两种情况对图像信息未缺失的棉花提出了成熟度判别的策略：正面时利用成熟棉花棉瓣的分散性和单朵棉花面积较大等特征进行判别;侧面时通过成熟棉花棉瓣相对棉荚面积比较大的特征进行识别。根据提取分割后棉花图像的几何信息,可确定棉花的重心位置。试验结果表明：该算法能很好地将成熟棉花从背景中分离出来,并较好地保存了棉花的轮廓信息。     

基于最小阈值法的棉花幼苗识别研究

 面对纯机械式分苗机构经常出现的漏分和多分的问题,本文提出使用基于最小阈值法的图像识别技术对棉花幼苗进行识别和分离。使用了HSV颜色空间内与光照条件无关的H值将RGB真彩色图像转化为灰度图像,以提高图像处理的实时性。基于最小值阈值法,提出一种能够有效地将处于复杂背景下的棉花幼苗茎部图像提取的方法,使用棉花幼苗茎部图像的面积和长宽比等几何信息能有效识别棉花幼苗,并得到棉花幼苗茎部的重心坐标。本文的研究证明了用图像处理中的最小阈值法识别棉花幼苗是可行的。     

基于数码图像的棉花叶片氮含量估测研究

为了探究利用数码图像处理技术诊断棉花氮素营养的最佳图像特征参数和最佳叶位,以‘新陆早53号’为材料,使用智能手机获取棉花不同叶位叶片图像,利用数字图像处理技术提取叶片图像颜色和纹理特征参数,分析叶片氮含量和特征参数的相关性,构建基于不同图像特征参数的氮含量估测模型。结果表明：颜色特征参数r、G/(B+R)、G/B和纹理特征参数COR、ASM与叶片氮含量相关性较好,相关性系数(r)均大于0.55。基于颜色-纹理综合特征参数构建的氮含量估测模型均优于基于颜色或纹理特征参数所建模型,其中倒4叶氮含量估测模型最优,模型决定系数(R²)为0.875、均方根误差(RMSE)为1.324、相对误差(RE)为8.09%。因此,利用数字图像处理进行棉花氮素营养诊断时,应选择的最佳图像特征参数为r、G/(B+R)、G/B、COR、ASM,应选择的最佳叶位为棉花倒4叶。     

黑背景下收获前棉花图像色特征生成及其品级聚类分析

为了客观地评价田间收获前棉花品级,依据我国子棉收购文字标准,运用机器视觉技术,对棉花的尺寸和色泽特征进行了研究,其色泽特征包括6个常用颜色空间下棉花和带壳棉花的黄色区域、黄色深度、白度、色差等8个色泽特征.结果表明,RGB、NTSC、亨特、HSI颜色空间下的特征参数区分度较好;棉花尺寸与色泽特征的相关性表明棉花白度为无效特征.在各颜色空间下,基于有效图像特征对7个品级的棉花样本进行了K均值聚类分析.分析结果说明,棉花品级的聚类结果独立于颜色空间;铃壳色泽对品级的贡献显著;由于其聚类品级与图像特征的相关性普遍较高、较均衡,且算法运行时间较短,HSI颜色空间可能是棉花品级聚类的最优颜色空间.与人工分级相比,机器视觉的区分度更好,因此机器视觉技术可用于提高收获前棉花分级精度.     

基于图像特征的收获前籽棉品级聚类融合分析

为客观评价收获前籽棉品级,依据籽棉品级国家标准,基于机器视觉在3个颜色空间中选取棉棉瓣大小、色泽特征用K-均值、竞争学习网络方法对7个品级的样本进行聚类融合分析。结果表明,亮度修正后特征之间极显著相关,Hunter颜色空间较好。肉眼对第1、2、7品级的识别率为73%~100%,3~6品级为26%~46%,总计47.7%;聚类融合对各品级的识别率为65%~100%,总计78.6%。聚类融合方法基于人类的先验知识,在更宽的视觉范围内更均衡所有特征,可克服个体聚类器的过度训练,能够客观地识别收获前籽棉品级,提高其采摘、收购质量。     

基于Visual c++矩形标尺智能识别

通过利用图像处理和图像识别领域的边界分割Robert局部边缘检测算子方法及梯度算子区域迭代的阈值图像分割算法相结合来对图像特征进行二次提取,不同于单独采用边界分割与区域分割算法,该算法通过对边缘检测结果的差分结果的迭代分析,有效的增强了识别算法对识别条件的适应性及识别精度,并采用Visual C 编程进行计算机图像辅助设计作为实现手段.     

基于颜色特征和AdaBoost算法的麦穗识别的研究

单位面积麦穗数是小麦产量预测中的一个重要参数。在应用图像处理技术识别麦穗的个数研究中,提出了一种小麦颜色信息和AdaBoost算法相结合的麦穗检测方法。先用麦穗颜色分割法排除非小麦区域的干扰,然后用AdaBoost算法训练的分类器对麦穗区域进行定位识别。该方法在保证检测率的同时,大大减少了目标区域的误检率,提高了小麦检测准确率。试验证明所用方法能有效去除叶片和秆茎的干扰,对于角度倾抖有一定鲁棒性。在随机抽取的100个样本中检测率为88.7%,有很强的识别能力。     

山东圣丰种业田间地头开展徐豆14间苗技术指导

山东圣丰种业科技有限公司成立于2004年9月,控股山东圣丰豆业有限公司和济宁嘉诚种业有限公司,是以棉花、大豆、小麦种子为主导产业的省级高新技术企业。2005—2007年,大豆和棉花种子总销售额、总销售量和市场占有率在全国名列前茅,其中大豆种子居全国首位。     

Cotton Image Segmentation Algorithm Based on Fusion Method of Markov Random Field and Quantum Particle Swarm Cluster

[Objective] The aim of this study was to improve the cotton image segmentation accuracy in a picking robot image processing system. [Method] An image segmentation algorithm based on a fusion method of Markov random field and quantum particle swarm optimization clustering was proposed. The process of the proposed algorithm is as follows: first, transform the RGB (red, green, blue) images into grayscale; second, use it to segment these images; finally, the threshold of the connected area is set on the basis of the segmented image to obtain the target area. Then, the cotton front image and the cotton side image are selected from the images collected from different angles. The segmentation experiment was carried out by using this algorithm, and compared with the Otsu algorithm, the fuzzy C-means algorithm, the quantum particle swarm image segmentation algorithm and the Markov random field image segmentation algorithm. [Result] The results showed that the segmentation accuracy and peak signal to noise ratio of the proposed algorithm were 98.94% and 77.48 dB. When compared with the Otsu algorithm, fuzzy C-means algorithm, quantum particle swarm optimization algorithm and Markov random field algorithm, the average segmentation accuracy and peak signal to noise ratio of the proposed algorithm increased by 2.47%–4.56%, and 9.81–13.11 dB, respectively. [Conclusion] The proposed algorithm had higher segmentation accuracy and higher peak signal to noise ratio than the other algorithms tested.     

Segmentation of Color Image Based on Feature Divergence and Fuzzy Theory

In order to improve general adaptive capability of algorithm,the new color image segmentation algorithm based on feature divergence and fuzzy theory(FDCIS) is proposed.The algorithm introduces feature divergence and fuzzy dissimilarity function into calculation in order to measure the dissimilarity of feature vector,clusters data by means of feature divergence,and accomplishes the merge of image region.The experimental results demonstrate that the color image segmentation result of the proposed approach reduce calculation on large sample of color image,simply and effectively solve over-segmentation of color image,avoid the dependence of the algorithm on initial condition,and hold favorable consistency in terms of human perception.     

不同灌水量对新疆高产棉花冠层结构映像特征及产量影响的研究

讨论了北疆棉区推广棉种新陆早12号与新陆早13号在不同灌水量条件下冠层结构特征的差异，得出新陆早12号对限量滴灌反应不敏感，而新陆早13号对限量滴灌反应较敏感。因此在限量滴灌条件下选择新陆早12号、适量滴灌条件下选择新陆早13号栽培种植较为适宜。由此得出了不同品种对灌水量的反应表现及理想群体冠层结构的合理指标，为增加棉花产量和提高水分利用效率提供了理论依据。     

棉花黄萎病菌在土壤—植株微生态系中的分布

在土壤垂直分布中,棉花黄萎病菌主要分布于0~40 cm耕作层中,占总菌量的90%多,而40 cm以下土层中仅占总菌量的10%；在土壤水平分布中，棉花黄萎病菌随病害的加重，由聚积分布型转变为均匀分布型。在病株中,病菌分布也不均,叶脉>主茎>果枝。     

Adaptive Image Segmentation Method Based on Ncut

In order to solve the problem that requires some factors by manual in the traditional Ncut algorithm, limit the generality of the algorithm, an adaptive image segmentation method is proposed by improving the traditional Ncut algorithm. First, instead of the two control parameters on the calculation of weight matrix that influence the segmentation results in the traditional Ncut algorithm by groups of potential theory; then in order to reduce the sensitive to the number of the cluster and the center of the cluster in the K-means algorithm, calculate on the eigenvector of the Ncut algorithm by the minimum spanning tree, to get the final number of cluster and the center, and then uses the K-means clustering algorithm to get the final segmentation result. The experimental results show that the proposed method not only improves the versatility of the algorithm, and the segmentation is good.     

Appoach To Object Semanteme Based Image Segmentation and Classification

There exist potential problems in current region-based image retrievals. This paper proposes a novel approach to object semanteme based image segmentation and classification. First, a hierarchical region growing image segmentation is established using HRGSeg algorithm, which can effectively get rid of weak object semantemes and play down the side effect of over-segmentation. Based on it, low-level features like color, edge and texture are extracted mapped into high-level object semantics hierarchically by using SVM. A fairly good experiment result is achieved and shows the feasibility of our approach.