基于Faster R-CNN的野外环境中蝗虫快速识别

蝗虫是常见的害虫之一,对农作物和生态系统具有很大的危害,采用常规的方法对蝗虫进行监测存在一定局限性,为了有效应用海量野外影像数据实现对蝗虫实时监测,本文建立了一种基于深度学习网络的蝗虫自动识别模型。利用手机模拟摄像头获取的内蒙古锡林浩特附近草原的280张蝗虫的RGB图像,采用深度学习算法中的Faster R-CNN（Faster Region-based Convolutional Neural Network）网络结构建立了蝗虫识别模型。经验证该模型的精确度为0.756,可以较准确地将蝗虫从野外复杂环境中识别出来,与以往同类研究相比,在识别结果和实用性方面均有较大的进步。该模型是建立蝗虫实时监测系统的基础,可以为蝗虫的防治提供辅助信息,同时该网络结构还可以应用于其他害虫的识别,具有较强的推广性,拓宽了深度学习算法的应用领域。     

Improved geological mapping using landsat-5 TM data in Weixi area, Yunnan province China

Different image processing algorithms have been evaluated in the context of geological mapping using Landsat TM data. False color composites, the principal component imagery, and IHS decorrelation stretching method for Landsat-5 TM data have been found useful for delineating the regional geological features, mainly to provide the maximum geological information of the studied area . The study testifies that using which image processing yields best results for geological mapping in arid and semiarid regions by preserving morphological and spectral information. Generally, the studied area can be divided into three main geological units: Basaltic intrusive rocks, Metamorphic with varying intensities and Sedimentary rocks.     

Biogeochemical evaluation of Nannorrhops ritchiana : A Mg-flora from Khuzdar, Balochistan, Pakistan

Nannorrhops ritchiana (Mazari Palm) is a distinctive flora growing in the Saharo-Sindian region. It is well distributed on the ultramafic soil, derived from the Bela Ophiolite in the Khuzdar District, Balochistan, Pakistan. Quantitative estimation of Ca, Mg, Fe and Ni in soil and plant ash has been carried out. The constituents of plant ash have been discussed in relation to soil chemistry, pH, climate, mobility, average abundance in plant ash and exclusion mechanism of the flora. Relationship among Ca, Mg, Fe and Ni has been established using scattergrams to evaluate the biogeochemistry of the plant. High contents of Mg and high coefficient of biological absorption allow it to be classed as Mg-flora. Both Ca and Fe appeared to be antagonistic to Mg. The metal assemblage ofN. ritchiana nicely reflected the nature of the bed rock as being serpentinized ultramafic, and its corresponding soils. Good exclusion mechanism ofN. ritchiana did not allow it to absorb high Ni from the soil. Relatively high concentrations of Ni inN. ritchiana from the Baran Lak area can be used to localize Ni-mineralization in the study area. Statistical analyses, such as minimum, maximum, mean, mode, median, standard deviations, and coefficient of correlation, were also made to improve raw geochemical data and interpretations.     

Cluster analysis of the space-time distribution of sunspot groups during solar cycle no. 20

Cluster analysis (a Bayesian iteration procedure) was used to study the space-time distribution of sunspot groups in the time interval from 1965 to 1977. (Data were taken from the Greenwich and Debrecen Heliographic Results.) The distribution proved to be significantly non-random for the 8–10 groups cluster^–1 (gr cl^–1) level of clustering. Convincing evidence also favours non-random behaviour for other levels of clustering from the lowest (3–4 gr cl^–1) up to the highest ( 150 gr cl^–1) level. The rotation rate of the non-random pattern is generally slightly lower than the Carrington rate.The 8–10 gr cl^–1 level, crudely corresponding to the sunspot nests investigated earlier, was studied in more detail. The cycle- and latitude-averaged rotational rate of the nests is slightly ( 1%) but significantly lower than the Carrington rate. Their differential rotation is strongly reduced: the cycle-averaged rotational rate varies only by 2–3% within the sunspot belt. A slight but significant bimodality is seen in the differential rotation curve: the intermediate latitudes ( 10°–20°) show a somewhat slower rotation than both the equatorial and the higher latitude regions. This might be explained by a time-dependence of the rotation rate coupled with the butterfly diagram.