针对果园生产管理的需求,基于微信小程序开发智慧果园大数据移动管理服务平台,以便对果园实现精准管理和服务.以果园管理为场景需求,综合运用移动GIS、数据库和AI目标检测等技术,实现生产服务、咨询服务、果园管理和数字采集等功能,设计开发智慧果园大数据移动管理服务平台.通过构建智慧果园大数据管理和数据采集服务,实现园区信息的数字化管理和可视化,为数字果园、智慧果园管理和综合服务提供技术支撑.该平台可以作为智慧农业的信息服务管理基础支撑平台,为数字果园的数据采集、数据展示、数据分析与应用提供有效的手段.

为实现数字果园的机器视觉系统快速准确识别果树关键物候期,采集四川地区苹果、杧果、石榴、柑橘4种果树4个物候期的图像15000幅,按6∶2∶2的比例随机划分训练、验证和测试数据集,训练VGG16、ResNet50、MobileNetV2及Swin Transformer 4个深度学习图像分类模型,评测不同模型的精度和性能.结果表明,各模型识别物候期精度分别为98.9%、99.3%、99.7%、99.8%,其中杧果成熟期的识别误差较大,精度分别为96.7%、98.2%、99.0%、99.5%;模型识别测试集图像的计算量(GFLOPs)分别为15.50、4.12、0.32、15.14,识别单张图像耗时分别为3.00ms、2.33ms、3.00ms、4.67ms.该结果可为果园嵌入式设备、服务器端的机器视觉系统选择模型提供参考.

Citrus is widely planted in southern China. Due to cloudy and rainy weather, complex planting types, and other factors, it is difficult to use spectral information to directly identify citrus orchard information. Based on the unique phenological characteristics of citrus, this study put forward the hypothesis that “the vegetation information of citrus orchards may be weakened during the growth and expansion of citrus fruit”. According to this feature, a method of citrus orchard information identification is proposed, and the threshold of the key time window is determined. Taking Wuming District, Nanning City, and Guangxi Zhuang Autonomous Region as the research area, an empirical study on remote sensing identification of citrus orchard information is carried out. First, multi-temporal Sentinel-2 remote sensing images of the study area in 2018 were obtained, and a normalized difference vegetation index was constructed. NDVI, Green Normalized Difference Vegetation Index (GNDVI), Difference Vegetation Index (DVI), Sentinel-derived red-edge spectral indices (RESI), and other vegetation spectral indices Secondly, according to the ground sample point information, the difference in remote sensing vegetation information of different vegetation types in different periods was compared, and then the optimal features of citrus orchard identification were determined. The results showed that there was no significant difference in spectral characteristics between citrus orchards and other major crop types in the study area (such as sugarcane, banana, corn, rice, etc.), but the multi-temporal remote sensing vegetation index of the study area showed that the NDVI of citrus orchards in October was 0.47 lower than that of November, which was significantly lower than that of other crop types. In October, the GNDVI of the citrus orchard also showed a low value of 0.43, but the difference was not obvious compared with other months. However, the dispersion degree of citrus orchard DVI was low, and the separation was not strong. According to the crop phenological calendar, the period of rapid expansion of citrus fruits was from September to October, which verified the scientific hypothesis proposed in this study that the vegetation information of citrus orchards would be weakened during this period. The dispersion degree of different vegetation indexes in the citrus fruit expansion stage was obviously different, and the dispersion degree of NDVI was the highest, and the difference was the strongest. According to the phenological characteristics of the citrus orchard NDVI in October, to further build the normalized index, by using the threshold value method to identify the spatial distribution of the citrus orchard, the identification method had an overall accuracy of 82.75%, better than other identification results of vegetation index. The results of the study for citrus orchard information and remote sensing identification research provide better support for theory and practice.

[目的]传统数字化果园管理中存在数据存储分散、数据处理效率低、服务能力弱和整个生命周期的农业生产数据缺乏等问题,开发集数据存储、管理、挖掘分析和决策支持等功能于一体的果园大数据智能管理与服务平台迫在眉睫.[方法]文章通过结合"空-天-地"一体化、数字农业、智慧农业等理念,集成深度学习、概率图模型和知识图谱等人工智能方法,开发了果园大数据智能管理与服务平台.[结果]平台实现了果树数量自动统计、地理环境感知、农机设备自动化精准感知、果园生产周期数据管理和智能化更新功能,同时平台还提供水、肥、药智能决策服务和智能农机生产作业管控服务.[结论]该平台在成都市龙泉驿现代农业园区进行应用,实现了桃园生产大数据的智能管理和提供精准生产指导服务.该平台的实现为促进果园智能管理和精准生产指导服务提供技术参考,对实现果园智慧化管理具有重要意义.

[目的]为智慧果园管理和综合服务提供技术支撑.[方法]传统的数据库无法满足智慧果园多模态数据融合和全场景查询分析等问题,文章通过分析智慧果园数据来源和应用逻辑,以都市龙泉驿现代农业园区为示范应用,从多级数据局部存储策略、冷热数据存储和空间数据存储等方面探索了存储策略,提供了一站式数据库解决方案.[结果]该解决方案满足果园数据高效存储、管理的需求,结果显示YMatrix充分发挥了其高吞吐、低延迟、高并发、准实时数据加载等优点.[结论]该研究为数字农业和果园数字化基础设施建设和农业数字孪生建设提供技术参考,为实现现代化智慧农业提供经验.

[目的]平武县作为涪江上游的水源涵养地与重要生态功能区,开展生态系统服务价值评估,对生态环境质量监测与自然资源资产化管理具有重要现实意义.[方法]本研究以单位面积价值当量因子法为基础,基于三期土地利用/土地覆被遥感监测数据,对平武县生态系统6个一级类型,14个二级类型,按11种生态系统服务功能细分类型,评估2005-2015年平武生态系统服务价值,并构建时空动态调节因子,对2015年内生态系统服务价值变化进行时空动态分析.[结果]①2005-2015年间,平武县土地利用格局比较稳定,生态系统服务价值略微增加,2015年达到333.30亿元,其中森林生态系统价值服务占比最大,接近65％.在服务功能类型中,气候调节功能价值总量最大,达88.36亿元.②2015年内,平武生态系统服务功能价值8月最高、2月最低,变化范围为150～422亿元.[结论]2005-2015年,平武县生态系统服务价值总体稳定、略微增加,年内逐月变化显著.生态系统服务价值时空动态分析方法能有效服务于平武生态系统监测与自然资源资产化管理.

[目的]提取四川省粮食生产核心区川西平原油菜种植区域空间分布信息,分析种植区域的空间特征和面积变化,以期为四川省农业政策制定、产业结构进一步调整和优化提供参考.[方法]文章将GF-1、Sentinel-2和Landsat-8全色波段影像和多光谱影像进行融合,分别得到空间分辨率为15 m、10m和15 m的融合图像;利用最大似然分类法对融合图像的RGB假彩色图像进行地物分类,提取油菜种植区域信息.以Google Earth发布的历史影像建立验证样区,通过目视解译提取油菜种植区域,验证遥感影像的提取精度.[结果]获得了该地区2016年和2018年共2期油菜种植区域空间信息.川西平原油菜主要分布在北部、东北部、西部和南部浅丘地带,零散分布于中部平原地区.2018年油菜种植面积较2016年有所增长,主要增加区域在川西平原崇州市、大邑县、邛崃市等市县.[结论]川西平原近年来油菜种植面积总体保持增长势态.四川省油菜产业的发展使得油菜种植区域差异化逐渐明显,需要进行针对性分析,因地制宜地研究和确定产业的战略目标与规划、区域农业产业化经营应遵循的法律或规则.

[目的]为实现果园自然场景下智能农业机器人对桃花的准确、快速、有效检测.[方法]文章采用相机获取桃花图片数据,通过LabelImg软件进行人工标记建立桃花目标识别的检测样本数据集,训练Darknet深度学习框架下的YOLO v4模型对桃花进行识别.[结果]模型精度评估表明,YOLO v4模型的平均准确率MAP值(86％)比Faster R-CNN的MAP值(51％)高出35％.[结论]YOLO v4与经典的算法相比,对各种自然环境下的桃花检测具有较好的实时性和鲁棒性,可为精准识别桃花提供重要参考价值,桃花精准识别为疏花疏果作业奠定了基础.

[目的]针对自然资源资产管理绩效评估中自然资源数据收集手段落后、要素不全、数据断档、更新频率不一致、缺乏空间信息等现状,构建了自动化、业务化自然资源遥感云计算动态监测服务平台.[方法]文章以四川省理县为例,利用Landsat,MODIS和Senti-nel 等多源遥感数据,通过计算植被指数、水体指数和干旱指数等指标,综合运用机器学习方法识别与提取自然资源地物类别.[结果]根据该文提出的高效计算方案,构建自然资源动态监测云平台,并基于多源数据信息的互补特性实现了复杂地物的高精度识别和提取,提升自然资源自动化、业务化动态变化监测能力.[结论]该平台可为生态系统价值评估、县域自然资源资产管理以及生态环境质量监测工作提供思路和参考.

Maize stalk mulching is a conservation tillage method that has been currently promoted in northeastern China Plain. Remote sensing estimation of regional crop residue cover (CRC) can quickly obtain the information of straw mulching in a large area, which plays an important role in monitoring and popularizing the work of straw mulching. In this study, the normalized difference til-lage index (NDTI), simple tillage index (STI), normalized difference residue index (NDRI), and normalized difference index 7 (NDI7) were extracted from Sentinel-2A image and used to establish a linear regression model for CRC and spectral indices in Lishu County of Jilin Province. The results showed that soils had strong spatial heterogeneity in the study area, which would lead to a significant impact on the spectral index regression model. Using soil texture classification (zoning) to establish regression model could improve the inversion accuracy. Soil spatial heterogeneity would increase the estimation error of the model. The four spectral indices had a strong correlation with CRC, among which the NDTI and STI models performed better. The zonal linear regression model based on NDTI and STI verified that R2 was 0.84 and RMSE was 13.3%, which was better than the non-zonal model (R2 was 0.75 and RMSE was 16.5%) and thus effectively improved the inversion accuracy.

为运用遥感手段掌握云南省甘蔗种植区域空间分布特征,评估甘蔗种植优势度,以云南甘蔗种植区为研究对象,采用Landsat8-OLI为数据源,结合地面调查结果,通过极大似然法提取甘蔗种植区域空间分布及面积,利用比较优势指数法、使用甘蔗遥感面积评估各区甘蔗种植优势度.2014-2016年遥感影像监测显示,云南省甘蔗种植面积约为24.12万hm2,主要集中分布在该省西南以及南部地区的主要河流沿线地带,临沧、德宏、保山、普洱、文山等区域的甘蔗面积约占全省的88％.比较优势指数法评估结果显示,临沧、德宏、普洱、保山、西双版纳的甘蔗规模优势指数分别为4.76、5.25、2.07、1.53、3.26,为云南甘蔗种植的优势区域;文山、玉溪、红河、大理等地区规模优势指数均小于1,种植优势不明显.研究结果可为云南甘蔗种植空间布局与优化调整提供参考.

[目的]基于深度学习的分类方法是使用高分辨率遥感影像快速提取作物种植空间信息的新方法.[方法]以云南省陇川县甘蔗种植园为研究区,收集空间分辨率为0.5 m的Google Earth开放影像进行数据预处理,建立样本数据集,构建U-Net神经网络模型,训练模型参数;使用U-Net模型提取甘蔗种植空间信息,通过地面样方数据验证甘蔗提取精度.[结果](1)基于深度学习方法的甘蔗分类总体精度和Kappa系数分别为92.76％和0.8480,面积总精度为94.41％;平坝区、丘陵区分类精度存在差异,总精度和Kappa系数分别为97.10％、0.9221和88.42％、0.7673;(2)受部分地物RGB影像特征与甘蔗相似的影响,分类结果存在错分现象.[结论]基于U-Net神经网络模型的方法可用于高分辨率影像的甘蔗提取,更准确的分类精度还有待进一步研究和验证.

[目的]开展农地确权工作旨在放活土地经营权,促进农地流转,实现农业规模经营,然而现实中农地流转出现与预期效果不一致的现象.文章重点为四川省农业规模经营的发展提出可行路径.[方法]文章以农户有限理性为基本假设,运用前景理论分别对四川省成都平原地区、浅丘地区和山区的农户,在农地确权前后对农地流转的心理认知进行分析,并针对不同地区农户农地流转心理状态及当地资源禀赋提出适宜的农业规模经营实现路径.[结果]山区农户生存能力较弱,对土地依赖程度强,土地用于满足其生存需求,通过农地确权降低了流转意愿;浅丘地区农户生存能力较强,对土地依赖程度较弱,土地用于满足其安全需求,在农地确权之后,部分农户倾向于短期流转,部分农户愿意流入土地;成都平原地区农户生存能力强,对土地依赖程度弱,土地用于满足其社交需求和尊重需求,流转土地的态度在农地确权之后趋于开放.[结论]对于四川省山区,提出"充分发挥生态优势,第三产业带动发展"与"细化生产分工,实现农业服务规模经营"路径建议;对于四川省浅丘地区,提出"发展土地股份合作社,按面积入股分红退股"与"产品标准化生产,产业品牌化发展"路径建议;对于成都平原地区,提出"互联网+新型农业经营主体+智慧农业"与"打造田园综合体,开发新型产业发展亮点"路径建议.

Remote sensing technique is known as an inexpensive and effective tool for retrieving crop variables in a large area.The existing methodologies can be identified into two categories:the methodologies based on statistical predictive models and the methodologies based on canopy reflectance (CR) models inversion.The latter is relatively universal.Thus,it has great potential in wisdom agriculture for crop monitoring in regional scale.However,CR model inversions suffer from the so-called "ill-posed problem".Therefore,the multi-stage,sample-direction dependent,target-decisions (MSDT) inversion technique and the object-based inversion technique were previously proposed.They are similar in technical routes:the progress of an inversion is partitioned into several stages.In each stage,only a part of variables were estimated.The results of preliminary stages are used as prior knowledge of later stages of inversion.In this way,the uncertainties in parameter optimization are reduced,the ill-posed problem is therefore limited.Concretely speaking,the MSDT method firstly estimates the sensitivity and uncertainties of variables before each stage of inversion.The most sensitive and uncertain variables were firstly retrieved using a subset of remote sensing data which is sensitive to the retrieved variables.The scheme of parameterization is then updated based on the preliminary results.Another subset of sensitive variables was subsequently retrieved using another subset of sensitive data.The object-based inversion defines an "object" as a plot or a gliding window,in which the crop has similar attributes.Such attributes are referred to as "object signatures".A remotely sensed image is firstly segmented into objects.Within each object,object signatures are firstly retrieved,and used as prior knowledge in subsequent pixel-wise retrieval of spatial heterogeneous or interested variables.In this way,spatial constrains,i.e.,the spatial distribution of variables,are extracted and imposed on the inversion.It can be seen the MSDT and object-based inversion essentially follow the same procedure.The major difference between them is that MSDT method makes the scheme of inversion according to the sensitivity and uncertainty of variables,while object-based inversion is based on the spatial distribution of variables.In this review,MSDT and object-based inversions were summarized into an integrated conceptual framework of "multi-stage inversion".Based on this framework,the following technical problems and the potential solutions can be summarized as follows.1) The schemes of MSDT and object-based inversions are practically in conflict.In future studies,multi-step inversion strategies need further comparison,verification and improvement to ensure their rationality and effectiveness.The thoughts of MSDT and object-based inversions should be integrated,to develop more sophisticated inversion schemes under the conceptual framework of multi-step inversion.2) Multi-step inversions might be significantly affected by the accuracy of preliminary parameterization of CR model.In future studies,the integrated application of multi-sources data could be helpful for CR model parameterization,and for detecting errors in each stage of inversion.For instance,same variables can be retrieved from satellite,aerial and ground remote sensing data,or obtained directly from in-situ measurements and existing remote sensing products.With approaches of scale transformation,the variables retrieved from multi-source data can be compared,in order to obtain prior-knowledge,or detect error in inversions.3) Multi-step inversions might be distorted by error propagation.In future studies,on the one hand,gross errors and systematic errors should be detected and corrected in each stage of inversion according to the statistical distributions of retrieved variables,or by using multiple data sources.On the other hand,the schemes of multi-step parameter optimization should be customized for each variable according to its sensitivity and spatial heterogeneity.Not to fix sensitive or spatially heterogeneous variables if the accuracy and reliability of prior knowledge or the preliminary inversions could not be guaranteed.

It is vitally important to know leaf area index ( LAI) and its dynamics for crop growth monitoring and yield prediction. In recent years, the method of LAI estimation by inverting canopy reflectance ( CR) models has-been widely applied because of its universality and independence from ground truths of LAI. However, even if pa-rameterizations of CR models are tried to be accurate by substituting in-situ measurements or best estimations of variables, there are always bias between the simulated and remotely sensed spectra. In this paper such error is re-ferred to as "error of spectral simulation"or "simulation error". Based on the in -situ measurements of winter wheat variables, ACRM ( a Two-layer Canopy Reflectance Model) model is used to obtain the optimum simulated spectra. The optimum simulated spectra are compared with the in-situ measured winter wheat canopy spectra to reveal the distribution of simulation errors in different bands and in different sample plots. Four schemes of band se-lection are proposed and tested for retrieving LAI, in order to explore how the simulation errors affect LAI retrieval, and to discuss the principles of band selection for avoiding such affectations. Regression coefficient ( R2 ) between estimated and measured LAIs, root mean square error ( RMSE) and mean relative error ( MRE) of LAI estimation are used to evaluate the accuracy and stability of LAI retrieval. The experiment shows that, first, errors of spectral simulation differ significantly in different bands. Second, the LAI estimation without considering simulation errors yields much larger MRE (14. 31%) than that considered simulation errors (MRE 7. 84% ~ 9. 55%). However, the regression coefficients are similar (0. 8512 ~ 0. 8662). That means simulation errors cause significant system-atic errors in estimated LAI. Nevertheless, it hardly affects the stability of LAI retrieval. Third,generally speaking, hyperspectral bands with minimum simulation errors should be selected to estimate LAI, in order to achieve opti-mum accuracy. The result of dimension reduction using stepwise regression merely provides general indications on band number and band placement for LAI retrieval. Based on such indications, the bands with minimum local sim-ulation errors in each wavelength range should be selected for LAI estimation. However, simulation errors in each band differ slightly with different sample sites. Therefore, for each band, the average simulation error of a randomly selected subset of sample sites is computed and used as the basis of band selection. The results of this study are helpful for improving schemes of band selection, and for utilizing hyperspectral data more effectively for LAI estima-tion.

The deterioration that lies in fruit’s epidermis or adjacent pulp eroded by insect pests or Mechanical damage will lead to fruit rot,which affects fruit quality and economic value.A novel method of damage fruit detection is proposed in this paper.A Portable Detector for Damaged Fruit based on MCU is designed by using the diffuse reflectance principle of visible light and near-infrared.The detection method based on standard deviation is presented by analyzing the test voltage of red-Fuji apples.After testing,recognition accuracy rate of the detector is over 90%.