Dehai Zhu

Estimating regional winter wheat yield by assimilation of time series of HJ-1 CCD NDVI into WOFOST-ACRM model with Ensemble Kalman Filter

Abstract Regional crop yield prediction is a significant component of national food security assessment and food policy making. The crop growth model based on field scale is limited when it is extrapolated to regional scale to estimate crop yield due to the uncertainty of the input parameters. The data assimilation method which combines crop growth model and remotely sensed data has been proven to be the most effective method in regional yield estimation. The methods based on cost function are powerless with crop dynamic growth simulation and state variable dynamic update. However, sequence assimilation method has more advantages to overcome these problems, this paper presents a method of assimilation of time series HJ-1 A/B Normalized Difference Vegetation Index (NDVI) into the coupled model (e.g. WOrld FOod STudies (WOFOST) crop growth model and A two layer Canopy Reflectance Model (ACRM) radiative transfer mode) for winter wheat yield estimates using Ensemble Kalman Filter (EnKF) at the regional scale. The WOFOST model was selected as the crop growth model and calibrated and validated by the field measured data in order to accurately simulate the state variables and the growing process of winter wheat. The theoretically optimal time series LAI profile was obtained with the EnKF algorithm to reduce the errors which existed in both time series HJ-1 CCD NDVI and WOFOST–ACRM model. Finally, the winter wheat yield at the county level was estimated based on the optimized WOFOST model running on the wheat planting pixel. The experiment illustrates that in the potential mode, the EnKF algorithm has significantly improved the regional winter wheat yield estimates ( R 2 = 0.51 , RMSE=775xa0kg/ha) over the WOFOST simulation without assimilation ( R 2 = 0.25 , RMSE=2168xa0kg/ha) at county level compared to the official statistical yield data. Meanwhile, in the water-limited mode the results showed a high correlation ( R 2 = 0.53 , RMSE=3005xa0kg/ha) with statistical data. In general, our results indicate that EnKF is a reliable optimization method for assimilating remotely sensed data into the crop growth model for predicting regional winter wheat yield.

Object-oriented feature selection of high spatial resolution images using an improved Relief algorithm

Abstract In object-oriented classification approaches, small sample size and high dimensionality of features are the two main characteristics. In order to effectively use the rich features of the image objects, in this study, the Relief algorithms were improved in terms of aspects of randomly drawing samples, the influence of sample quantity variance, and iteration times to evaluate the features. Through experiments in WorldView-2 images, the two classification results were contrasted between 36 dimension feature sets selected from Initials and initial 111 dimension feature sets. The results showed that the dimension disaster was avoided efficiently, while the precision and speed of classification were improved as well. Respectively, the overall accuracy and kappa coefficient using the optimized feature sets with the improved Relief algorithm were increased by 6% and 11% compared with using all feature sets. Therefore, the feature space of object-oriented classification is effectively optimized with the proposed methods.

Mapping of Daily Mean Air Temperature in Agricultural Regions Using Daytime and Nighttime Land Surface Temperatures Derived from TERRA and AQUA MODIS Data

Air temperature is one of the most important factors in crop growth monitoring and simulation. In the present study, we estimated and mapped daily mean air temperature using daytime and nighttime land surface temperatures (LSTs) derived from TERRA and AQUA MODIS data. Linear regression models were calibrated using LSTs from 2003 to 2011 and validated using LST data from 2012 to 2013, combined with meteorological station data. The results show that these models can provide a robust estimation of measured daily mean air temperature and that models that only accounted for meteorological data from rural regions performed best. Daily mean air temperature maps were generated from each of four MODIS LST products and merged using different strategies that combined the four MODIS products in different orders when data from one product was unavailable for a pixel. The annual average spatial coverage increased from 20.28% to 55.46% in 2012 and 28.31% to 44.92% in 2013.The root-mean-square and mean absolute errors (RMSE and MAE) for the optimal image merging strategy were 2.41 and 1.84, respectively. Compared with the least-effective strategy, the RMSE and MAE decreased by 17.2% and 17.8%, respectively. The interpolation algorithm uses the available pixels from images with consecutive dates in a sliding-window mode. The most appropriate window size was selected based on the absolute spatial bias in the study area. With an optimal window size of 33 × 33 pixels, this approach increased data coverage by up to 76.99% in 2012 and 89.67% in 2013.

Texture extraction for object-oriented classification of high spatial resolution remotely sensed images using a semivariogram

A Semivariogram, as defined in geostatistics, is a powerful tool for texture extraction of remotely sensed images. However, the traditional texture features extracted by a semivariogram are generally for pixel-based classification. Moreover, most studies have been based on the original computation mode of semivariogram and discrete semivariance values. This article describes a set of semivariogram texture features (STFs) based on the mean square root pair difference (SRPD) to improve the accuracy of object-oriented classification (OOC) in QuickBird images. The adaptive parameters for the calculation of a semivariogram were first derived from semivariance analysis, including directions, moving window size, and lag distance. Then, 22 STFs were extracted from the discrete and mean/standard deviation semivariance, and 15 features were selected from the extracted STFs based on feature optimization. Then five grey-level co-occurrence matrix (GLCM) texture features (mean, homogeneity, contrast, angular second moment, and entropy) were calculated based on segmented image objects using the panchromatic band. A comparison of classification results demonstrates that the STFs described in this article are useful supplement information for the spectral OOC, and the spectral + STFs classification method can be used to obtain a higher classification accuracy than can the combination of spectral and GLCM features.

A new hierarchical moving curve-fitting algorithm for filtering lidar data for automatic DTM generation

Recent advances in laser scanning hardware have allowed rapid generation of high-resolution digital terrain models (DTMs) for large areas. However, the automatic discrimination of ground and non-ground light detection and ranging (lidar) points in areas covered by densely packed buildings or dense vegetation is difficult. In this paper, we introduce a new hierarchical moving curve-fitting filter algorithm that is designed to automatically and rapidly filter lidar data to permit automatic DTM generation. This algorithm is based on fitting a second-degree polynomial surface using flexible tiles of moving blocks and an adaptive threshold. The initial tile size is determined by the size of the largest building in the study area. Based on an adaptive threshold, non-ground points and ground points are classified and labelled step by step. In addition, we used a multi-scale weighted interpolation method to estimate the bare-earth elevation for non-ground points and obtain a recovered terrain model. Our experiments in four study areas showed that the new filtering method can separate ground and non-ground points in both urban areas and those covered by dense vegetation. The filter error ranged from 4.08% to 9.40% for Type I errors, from 2.48% to 7.63% for Type II errors, and from 5.01% to 7.40% for total errors. These errors are lower than those of triangulated irregular network filter algorithms.

Real-time control of 3D virtual human motion using a depth-sensing camera for agricultural machinery training

Abstract To recreate human movements in a virtual environment in real time, we propose a new method for real-time tracking of 3D virtual full-body motion using a depth-sensing camera. The method uses natural interaction and a non-contact mode. The 3D virtual environment was constructed using a 3D graphics engine and human joint data were calculated using images acquired from a Prime Sense depth-sensing camera. Then skeletal data for the human model in a skinned mesh animation were separated by improving the mesh modules using a 3D graphics engine. Finally, motion data from the depth sensor were combined with joint data for the human model to yield full-body control of a virtual human (VH). Experimental results show that the proposed method can drive VH full-body movements in real time based on motion-sensing data. The method was applied in virtual driving training for agricultural machinery. Trainees can become familiar with the basic operations required for driving agricultural machinery using full-body motion instead of a mouse and keyboard. The training system is inexpensive and has high safety and a strong sense of immersion.

Design and Implementation of Locust Data Collecting System Based on Android

This Paper is mainly focused on the Locust Data Collecting System based on Android platform, and designs the structure of data acquisition and data transmit system. The system includes data acquisition unit, data transmission unit and process unit in the server. In addition, the paper discusses the implementation of the Data Collection in mobile and the realization of the programming. The experiment indicates that the method is feasible and effective.

LSSA_CAU: An interactive 3d point clouds analysis software for body measurement of livestock with similar forms of cows or pigs

Abstract As increasing number of studies for shape measurement purposes in livestock farming by using consumer depth cameras, many software have been developed in order to measure livestock conformation. However, many of these softwares were designed only for specific livestock or body part of specific livestock with very limited body measurements. To be more flexible and general compared to the current software provided in the literature, an interactive software LSSA_CAU is developed to estimate body measurements of livestock based on 3d point clouds data. Livestock with similar forms of cows or pigs and standing with her head forward is assumed for designing algorithm used in LSSA_CAU. This software provides a set of tools for loading, rendering, segmenting, pose normalizing, measuring point clouds data of whole body surface of livestock in a semiautomatic manner. In order to validate the software, both synthetic and real world point clouds data of livestock were processed by using the LSSA_CAU. Our experiments show that the proposed software generalizes well across livestock species and supports customized body measurements. An updated LSSA_CAU version can be downloaded freely from https://github.com/LiveStockShapeAnalysis to livestock industry and research.

A portable and automatic Xtion-based measurement system for pig body size

Abstract Body measurement plays an important role in animal breeding and production. In this paper, we develop a novel portable and automatic measurement system for pig body size. Firstly, we utilize two depth cameras to capture the point clouds of the scene with a pig from two viewpoints and implement the registration of the obtained point clouds. Secondly, we resort to Random Sample Consensus (RANSAC) to remove the background point cloud and extract the foreground pig point cloud with a Euclidean clustering. Finally, body measurement is conducted via pose normalization and morphological constraints on pig cloud. We evaluate the proposed system on 20 sets of the scenes with a pig in a commercial pig farm. Experimental results show that the pig object extraction algorithm achieves good performance. The average relative errors for body width, hip width, and body height are 10.30%, 5.87% and 7.01% respectively, which demonstrates the efficacy of the proposed system.

A WebGIS-based decision support system for locust prevention and control in China

Abstract Locust swarms are destructive agricultural and biological disasters in China. The green prevention and control (GPC, such as ecological regulation and physical control) of locusts is a comprehensive and complex process, especially in information technology. In this study, a web-based decision support system (DSS) integrated with geographic information system (GIS) is developed to prevent and control locusts efficiently, accurately, and rapidly. The locust prevention and control DSS (LPCDSS) is developed to assist farmers and local government agencies in Chinese provinces with high incidence of locust by providing spatial decision-making information. LPCDSS offers online access to county, city, provincial, and national level data queries and is capable of storing, spatial analyzing, and displaying geographically referenced information of locust data. The system can also provide the real-time tracking of global positioning system (GPS) location, as well as goods scheduling of locust plagues prevention. Six types of web service, real-time data synchronization model, and locust population estimation model are developed and implemented to improve the decision-making usability and feasibility of LPCDSS by adopting a three-layer system architecture. The system is developed by using several programming languages, libraries, and software components. As a result, this system has been running successfully for several years and has improved efficiency of the locust prevention and control in China with high efficiency and great accuracy. The approaches and methodologies presented in this paper can serve as a reference for those who are interested in developing integrated pest control system applications.