结合Sentinel-2影像和特征优选模型提取大豆种植区

准确获取大豆的空间分布对于产量估计、灾害预警和农业政策调整具有重要意义,目前针对种植结构复杂地区所开展的大豆遥感识别研究鲜有报道。该研究以安徽省北部平原的典型大豆产区——龙山、青疃镇为研究区,基于Sentinel-2数据提出一种分层逐级提取策略的大豆识别方法。该方法首先构建决策树筛选规则,剔除研究区内非农田地物,获得田间植被的总体分布;然后生成19个候选特征因子,包括分辨率小于等于20 m的10个波段反射率以及9个植被指数。在典型地物类型样本的支持下,将ReliefF特征权重评估算法与随机森林(RandomForest,RF),BP神经网络(Back-Propagation Neural Network,BPNN)和支持向量机(Support Vector Machine, SVM)相结合,分别构建ReliefF-RF、ReliefF-BPNN、ReliefF-SVM三种组合模型筛选出对于大豆识别最有效的特征,并基于布设在研究区内6个样方(大小为1 km×1 km)的无人机影像提取得到的大豆分布来评估3种模型在大豆制图中的表现。结果表明,ReliefF-RF模型表现最佳,基于该模型筛选出7个优选特征因子,大豆制图的总体精度介于85.92%～91.91%,Kappa系数在0.72～0.81之间,各个样方的提取效果均优于其他两种模型。此外,基于优选特征达到的提取精度明显高于原始波段反射率,虽然略低于全部19个特征的结果,但是数据量降低了63.16%。该研究可以为农田景观破碎、种植结构复杂地区的大豆种植区提取相关研究提供有价值的参考和借鉴。     

基于双目视觉的香蕉园巡检机器人导航路径提取方法

为实现移动机器人香蕉园巡检自动导航,研究提出了一种基于双目视觉的香蕉园巡检路径提取方法。首先由机器人搭载的双目相机获取机器人前方点云,进行预处理后对点云感兴趣区域进行二维投影并将投影结果网格化,得到网格地图;然后采用改进的K-means算法将道路两侧香蕉树分离,其中初始聚类中心通过对网格地图进行垂直、水平投影以及一、二阶高斯拟合确定;最后基于最小包围矩形提取导航路径,将道路两侧网格以最小矩形框包围,提取两包围框中间线作为期望导航路径。测试结果表明,改进的K-means算法聚类成功率为93%,较传统方法提高了12%;导航路径提取平均横向偏差为14.27 cm,平均航向偏差为4.83°,研究方法可为香蕉园巡检机器人自动导航提供支持。     

基于中值点Hough变换玉米行检测的导航线提取方法

为解决机器视觉对早期玉米苗带在多环境变量下导航线提取耗时长、准确率低的问题,该研究提出了一种基于中值点Hough变换作物行检测的导航线提取算法。首先,改进了传统的2G-R-B算法,再结合中值滤波、最大类间方差法和形态学操作实现土壤背景与玉米苗带的分割。其次,通过均值法提取玉米苗带特征点,然后采用中值点Hough变换拟合垄间两侧玉米苗列线,最后将检测出的双侧玉米苗列线为导航基准线,利用夹角正切公式提取导航线。试验结果表明：改进的灰度化算法能够正确分割玉米苗带与土壤,处理一幅640×480像素彩色图像平均耗时小于160 ms,基于中值点Hough变换检测玉米苗列再提取导航线的最大误差为0.53°,相比于传统Hough变换时间上平均快62.9 ms,比最小二乘法平均精确度提高了7.12°,在农田早期玉米苗带多环境变量影响因素下导航线提取准确率均达92%以上,具有较强的可靠性和准确性。     

采用K均值聚类和环形结构的狭叶锦鸡儿木质部提取算法

针对木质部交互统计误差大、效率低、重现性差、劳动强度高和传统图像处理算法精度不理想等问题,该文以狭叶锦鸡儿木质部切片图像为研究对象,根据木质部特点提出基于K均值聚类算法和环形结构提取算法相结合,实现木质部准确提取的方法。首先通过动态巴特沃斯同态滤波法对30幅供试图像进行光照不均校正,然后采用K均值聚类法对光照补偿后图像初分割,最后采用环形结构提取算法实现木质部提取计数。试验结果表明:采用K均值聚类算法对光照补偿后的木质部图像初分割分割误差R(section error, R)、过分割误差OR(over-segmentation error, OR)和欠分割误差UR(under-segmentation error, UR)均值分别为5.15%、1.48%和6.46%,优于未光照补偿和3R-G-B算法;该文提出的环形结构提取算法对初分割后木质部图像检测的平均相对误差为2.26%,比分水岭法低11.69个百分点,比凹点匹配法低4.93个百分点。从速度上看,该算法平均耗时3.17 s,比分水岭法快1.40 s,比凹点匹配法快4.88 s。该算法检测的均方根误差RMSE(root mean squared error, RMSE)为0.52%,约相当于分水岭法的1/3,约相当于凹点匹配法的1/2,该算法优于其他2种分割算法;在图像结构复杂、光照不均匀、内部分布不均等缺陷条件下,该文算法也能很好地实现木质部的分割和提取。该方法不仅能对狭叶锦鸡儿木质部自动分割和提取,也可为其他植物木质部分割提取提供参考。     

Copper,Zinc, and Lead Fractions in Soils Long-Term Irrigated with Municipal Wastewater

Long-term irrigation with municipal wastewater may lead, even in spite of intense farming, to an accumulation of organic matter, nutrient elements, and trace metals in soils. Excessive increases of heavy metals may pose a potential risk to the food chain and provoke restrictions for the further cultivation of sensitive crops. Copper (Cu), zinc (Zn), and lead (Pb) forms in soils under long-term irrigation (for 100–120 years) with treated wastewater of Wroclaw were investigated by using selective seven-step sequential extraction (procedure of Zeien-Bruemmer) for partitioning the metals into operationally defined fractions, likely to be released in solution under various environmental conditions. The largest fraction of Cu, Pb, and particularly Zn in nonirrigated (control) soils was strongly bound in a residual form, while the percentage of exchangeable and the most labile fractions were negligible. Total concentration of metals in irrigated soils was elevated, and significant redistribution of metals among phases was observed. Percentages of residual fraction of Cu and Pb were no more than 25% (Zn < 40%), while significantly increased contribution of fractions occluded on iron (Fe) oxides and organically bound Cu. Exchangeable and readily mobile forms of Zn are predominant zinc fractions in soils irrigated with wastewater.     

基于图像特征选择识别田间籽棉品级

依据中国籽棉品级文字标准,在HSI颜色空间提取了反映籽棉颜色和杂质含量的14个纹理特征以及反映棉瓣大小、结构的16个形状特征,该特征集存在维数灾难,需要进行降维。面向籽棉品级识别的特征选择问题属于非多项式(NP)难题,该文基于交叉验证、混合Filter-Wrapper和启发式搜索提出了一种求解算法。首先,以最优特征组合和浮动搜索为启发式搜索策略,基于10-折交叉验证在每一个训练集上用Filter启发式搜索最优l维特征子集（l=1, 2, 3,…, 30）,评价函数为类可分性准则;其次,在10个训练集上用Wrapper从最优l维特征子集中选择最优特征子集的容量（l=1,2,3,…,30）,评价函数为Bayes分类器的误分率,10个验证集的平均误分率极小处产生最优特征子集的容量;最后,在最优特征子集容量处验证预测集的平均误分率。结果表明,所选择的10个最优特征子集在预测集上的平均识别率为88.39%,混合Filter-Wrapper和浮动搜索的特征选择算法效率高、效果好。     

缓控释肥料缓释期的快速测定研究

采用25℃水浸提法和高温快速浸提法,研究了3种不同类型的缓控释肥料缓释期快速测定的最佳温度,测定三者在25℃下的缓释期,并对其在缓释期最佳测定温度和25℃下达到相同氮素累积释放率所需时间进行相关分析.结果表明:微溶性二价金属磷酸铵钾盐为包裹层的多层包裹粒状水溶性肥料(1#)肥料缓释期最佳预测温度为40℃,热塑性树脂包膜...     

黄土丘陵沟壑区地表过程响应单元划分及其地形特征提取

地表过程解析与调控是保障人与自然和谐相处的重要途径,高效、合理地划分地表过程响应单元则是开展地表过程研究的重要前提.针对黄土丘陵沟壑区土壤侵蚀的地形地貌分异特征,提出在小流域内以沟谷线、山脊线、分水线和沟缘线相互交叠,构成具有相对均一坡向、坡度与坡位的地表过程响应单元,在ArcGIS9.2中,以Hc-DEM为基础数据,通过确定沟谷线、山脊线、分水线和沟缘线的提取方法,以及响应单元平均坡度、整体坡向、面积与周长等地形特征的测算技术,实现地表过程响应单元的快速划分及其地形特征的高效提取.研究结果对促进黄土丘陵沟壑区以土壤侵蚀为代表的地表过程研究具有重要的基础意义.     

棉花异性纤维中麻绳与羽毛的分类特征

摘要：为准确识别棉花异性纤维中较难识别的羽毛和麻绳异性纤维,采用机器视觉技术,通过图像处理方法采集异性纤维目标,对羽毛和麻绳异性纤维的色彩和纹理特征进行有效的特征提取,形成异性纤维目标的特征向量。再通过一种自底向上的凝聚型层次聚类算法对提取的羽毛和麻绳的色彩与纹理特征进行层次聚类分析,选择最优特征向量。将8个特征向量进行降维分析并比较各维数下的层次聚类效果,试验结果表明,选取红色(R_ave)、绿色(G_ave)、蓝色(B_ave)、能量、熵、惯性矩等6个特征进行层次聚类效果最好,羽毛识别率达到94%,麻绳识别率达到95%, 说明选择的特征向量对这2种异性纤维具有理想的区分性。该研究可为棉花异性纤维的正确识别提供参考。     

协助自然修复微量元素污染酸性土壤研究：8年田间试验

There are many remediation techniques for organic contaminated soils,but relatively few of them are applicable to trace elementcontaminated soils.A field experiment was carried out to investigate assisted natural remediation（ANR） of an acid soil contaminated by As,Cd,Cu,Zn and Pb using one inorganic amendment,sugar beet lime（SL）,and two organic amendments,biosolid compost（BC）and leonardeite（LE）.The experiment was arranged in a completely randomized block design with four treatments in three replicates:1) a non-amended control（NA）;2) SL amended at 30 Mg ha^-1 year^-1;3) BC amended at 30 Mg ha^-1 year^-1 and 4) LE amended at 20 Mg ha^-1 year^-1 plus SL amended at 10 Mg ha^-1 year^-1（LESL）.The amended plots received two doses of each amendment（DO2）:one in October 2002 and another in October 2003.The plots were then divided in half into two subpolts and one subplot received another two doses of the same amendments（DO4） in October 2005 and October 2006.In 2011,the pH values of the amended soils were greater than that of the NA soil,with the SL-amended soil showing the highest pH.Total organic carbon（TOC） was also greater in the amended soil,and greater with DO4 than with DO2.Amendments reduced the concentrations of 0.01 mol L^-1 CaCl₂-extractable Cd,Cu and Zn,especially in the SL-amended soil.Plant cover of colonizing vegetation was enhanced by amendments,but was independent of amendment doses.Changes in soil properties from 2003 to 2011 showed that the first amendment application of DO2 caused a high differentiation between all the amendment treatments and the NA treatment.After the second application of DO2,soil pH and TOC continued increasing slowly.Further two applications of amendments（DO4） caused differences only in the organic treatments.Plant cover increased over time in all the treatments including NA.It could be concluded that the slow and steady natural remediation of this soil could be enhanced by amendment application（ANR）.