复杂背景下黄瓜病害叶片的分割方法研究

利用图像处理和模式识别技术进行复杂背景下黄瓜叶部病害的自动识别,需要先把目标叶片从复杂背景中分割出来,才能进行后续的特征提取和病害识别。为实现复杂背景下黄瓜叶片的分割,首先利用K-均值聚类算法去除图片中的非绿色部分,再采用基于laplacian of gaussia(LOG)算子的方法对待分割的叶片进行区域检测,然后进行基于形状上下文(shape context)的模板匹配和分割。为了提高匹配速度,先检测叶片的生长点和叶尖,以确定叶片的位置、尺寸和方向;然后使用基于超像素(superpixel)的最优匹配搜索方法来减少搜索的复杂度。对20幅黄瓜叶部病害图像进行分割测试,并与人工分割法进行对比,结果表明,本文所采用的分割算法能较好地从复杂背景下提取出黄瓜叶部病害图像,分割准确率达94.7%,为后期黄瓜病斑的特征提取等工作奠定了良好的基础。     

黄河流域水资源、能源与粮食生产的时空匹配及阻尼效应

水、能源与粮食是人类生活与可持续发展的重要保障,探究水资源、能源对粮食生产作用对于优化资源配置与保障粮食安全有着重要的意义。该研究运用洛伦兹曲线与基尼系数评估黄河流域上中下游水资源、能源与粮食生产之间的匹配度,并基于柯布-道格拉斯函数引入粮食增长阻尼模型测算出黄河流域水资源与能源对于粮食生产的制约程度。结果表明：1）黄河流域水资源、能源对粮食的基尼指数呈现出先减小后增大情形,黄河下游区域的匹配度要比上中游区域更为合理,2019年,下游区域水资源、能源对粮食都呈现出高度匹配状态情形。2）黄河流域水资源对于粮食生产的阻尼系数波动较大,阻尼系数变化范围为0.005～0.032,且水资源对粮食增长阻尼作用呈现出6 a一周期情形。能源对于粮食生产的阻尼作用呈现出稳步上升情形,且在2015后上增较快。在水资源与能源约束条件下,2019年黄河流域粮食产量增长要比上一年分别降低0.76%与5.28%。3）黄河流域水资源阻尼系数呈现出西部小东部大情形,能源阻尼效应呈中高约束状态集中在黄河上游东部区域与下游区域,低约束状态集中于黄河中游区域。另外,黄河流域水资源与能源阻尼系数具有典型的空间集聚特征,水资源阻尼效应高-高集聚区主要分布在黄河下游区域,能源阻尼效应低-低集聚区分布在中游北部区域。研究结论可为黄河流域粮食的稳定增长与资源有效配置提供科学依据。     

基于改进YOLOv4模型的全景图像苹果识别

苹果果园由于密植栽培模式,果树之间相互遮挡,导致苹果果实识别效果差,并且普通的图像采集方式存在图像中果实重复采集的问题,使得果实计数不准确。针对此类问题,该研究采用全景拍摄的方式采集苹果果树图像,并提出了一种基于改进YOLOv4和基于阈值的边界框匹配合并算法的全景图像苹果识别方法。首先在YOLOv4主干特征提取网络的Resblock模块中加入scSE注意力机制,将PANet模块中的部分卷积替换为深度可分离卷积,且增加深度可分离卷积的输出通道数,以增强特征提取能力,降低模型参数量与计算量。将全景图像分割为子图像,采用改进的YOLOv4模型进行识别,通过对比Faster R-CNN、CenterNet、YOLOv4系列算法和YOLOv5系列算法等不同网络模型对全景图像的苹果识别效果,改进后的YOLOv4网络模型精确率达到96.19%,召回率达到了95.47%,平均精度AP值达到97.27%,比原YOLOv4模型分别提高了1.07、2.59、2.02个百分点。采用基于阈值的边界框匹配合并算法,将识别后子图像的边界框进行匹配与合并,实现全景图像的识别,合并后的结果其精确率达到96.17%,召回率达到95.63%,F1分数达到0.96,平均精度AP值达到95.06%,高于直接对全景图像苹果进行识别的各评价指标。该方法对自然条件下全景图像的苹果识别具有较好的识别效果。     

瘦肉型三元杂交生长育肥猪日粮蛋白质主要必需氨基酸配比试验

试验选用体重30 kg和60 kg的杜长大三元杂交生长育肥猪各48头,各阶段随机分为4组(生长猪试1、2、3、4组,育肥猪试5、6、7、8组),每组12头,分别喂给蛋白模型1、2、3、4、5、6、7、8日粮。结果表明,生长猪以试验Ⅰ组平均日增重最高,分别比试验Ⅰ、Ⅲ、Ⅳ组高4.08%、7.33%、14.47%;试验Ⅱ组的料重比最低,分别比试验Ⅰ、Ⅲ、Ⅳ组低0.40%、11.70%、8.46%,能量转化效率最高,分别比试验Ⅰ、Ⅲ、Ⅳ组高2.5%、10.06%、6.62%,PER最高,分别比试验Ⅱ、Ⅲ、Ⅳ组高10.25%、9.80%、10.25%;育肥猪以试验Ⅶ组日增重最高,分别比试验Ⅴ、Ⅵ、Ⅶ组高6.69%、9.23%、8.32%;料重比以试验Ⅴ组最低,分别比试验Ⅵ、Ⅶ、Ⅷ组低8.55%、3.89%、4.71%;DE转化效率以试验Ⅴ组最高,比试验Ⅵ、Ⅶ、Ⅷ组分别高7.68%、3.15%、3.71%,PER最高,分别比试验Ⅵ、Ⅶ、Ⅷ组高8.95%、4.02%、4.02%。综合考虑,生长猪日粮蛋白质模型以日粮2、育肥猪以日粮5较为理想。     

Enigmatic hotspot of crayfish diversity at risk: Invasive potential of non‐indigenous crayfish if introduced to New Guinea

1. The large island of New Guinea has a rich indigenous astacofauna represented by numerous parastacids from the genus Cherax. The western half of the island is part of Indonesian territory.
2. Indonesia is known to be the main exporter of ornamental crayfish globally, and certain New Guinean species are exploited as ornamentals within the international pet trade. Moreover, one non‐indigenous species has been previously recorded being cultured in Java, Indonesia. This species, the North American Procambarus clarkii, is a vector of crayfish plague, the disease that is lethal to most parastacids. This population has already tested positive for the disease.
3. As the transport of non‐indigenous crayfish within the Indonesian territory is not restricted, their introduction to New Guinea can be expected. The Indonesian market was therefore surveyed for ornamental crayfish and their environmental suitability evaluated, as represented by temperature during the drought and rainy seasons in New Guinea.
4. Four North American and one Australian species were found advertised for sale. One of them, P. clarkii, was assessed as the most damaging species, followed by other North American species. A total ban on the culture and transport of the highest risk crayfish species in Indonesia and Papua New Guinea is recommended.

     

基于典型叶片模板自动匹配的虫损叶面积测量

为了定量化评估农作物的虫害程度,提出了一种基于典型叶片模板自动匹配的叶片虫损面积测量新方法。先将叶片图像二值化并提取其外轮廓;再对提取的轮廓进行多边形近似,以多边形的顶点为端点将叶片外轮廓划分成若干子轮廓;然后采用形状上下文对完整叶片与虫损叶片之间的子轮廓进行自动配准,找出其间的相互映射关系;最后根据映射关系对虫损叶片进行重建,计算出虫损面积。对10类不同叶片的测量分析表明：该方法平均每叶片耗时0.962 s,最大相对误差为8.22%,平均相对误差为4.78%。其中,形状复杂度高的叶片平均相对误差为7.48%,复杂度中等的叶片为5.99%,复杂度低的叶片为1.84%。结果表明,该方法能准确而快速地测量虫损叶面积。     

Vitality rating of picea abies by defoliation class and other vigour indicators

The study was made to assess the relationship between visually estimated defoliation class and a number of other variables depicting the vitality of Norway spruce (n=50). Variables characterizing crown size and condition were determined on the standing trees. Electrical impedance (an indication of the physiological status of a tree) was measured in the inner bark tissue of the living trees. Shoot growth and needle variables were determined from the detached sample branches. Negative correlations were found between defoliation class and the growth parameters (5‐year height, radial and volume growth of the trunk). Positive correlations were observed between the needle loss class and the abundance of branch damage, secondary shoots and impedance values. 18 variables (defoliation excluded) describing tree size and vitality were summarized in a factor analysis incorporating 5 factors. These were interpreted as 1) vitality, 2) foliage discoloration, 3) tree size, 4) needle number and 5) needle size factors. The tree‐specific score values for the first factor were interpreted as “vitality indices”;. The rank correlation between these score values and defoliation degree was 0.835. This result suggests that the defoliation class and vitality parameters contributing to the first factor give a consistent estimate of the tree condition.     

应用遥感技术增加连清间隔期时效信息产出的探讨

探讨了利用遥感技术在5年一次的连清间隔期插入一次或数次调查的方法,其关键技术是遥感判读点建立在连清地面样地调查基础上,这样产出的信息数据可靠、时效性强。此方法费用低、灵活实用,可满足各级政府和林业主管部门对森林资源信息的不同需求。     

适地适树研究进展

阐述了适地适树的概念、途径,及近年来我国在适地适树的决策方法、判断标准方面的研究进展,并提出了相关建议。     

Examining brain oedema through brain electrical impedance topographic mapping

Change of Brain Electrical Impedance has great relations with brain oedema and brain block, a new method of Examining Brain Oedema through Brain Electrical Impedance Topographic Mapping(BEITM)is reviewed. Network topology structure that used in BEITM are also presented. According to topology, we discuss the drawing algorithm that based on triangle element. Some patients have been measured to make clinical data, according to this data Brain Electrical Impedance topographic come into being by computer at last, functional imaging has been implemented. Clinical case analysis compared with change state of an illness and repeatability of steady patient's data has proved the method is feasible.