[ieee 2012 first international conference on agro-geoinformatics - shanghai, china...

6
Design of Remote Sensing System for Rice Information Extraction GUO Rui-fang, HUANG Jing-feng Institute of Remote Sensing and Information Application, Zhejiang University key Laboratory of Polluted Environment Remediation and Ecological Health, Ministry of Education, College of Natural Resources and Environmental Science, Zhejiang University Key Laboratory of Agricultural Remote Sensing and Information System, Zhejiang Province Hangzhou China e-mail: [email protected]; [email protected] WANG Xiu-zhen Institute of Remote Sensing and Earth Sciences, Hangzhou Normal University Hangzhou China e-mail: [email protected] AbstractRice is the world's most important crop for ensuring food security. Remote sensing has advantages in collecting data objectively and accurately on rice planting area, growth and development, and yield at large scales, but processing is slow due to vast amount of data, and there is need for an information extraction system specially designed for rice. This paper mainly discussed the key techniques for the rice data extraction system based on IDL including its graphical user interface, Xmanager Incident Response, object graphics system, image roaming and image scaling, as well as the function and structure of this system for rice area extraction, growth monitoring, and yield forecasting. Rice cultivation data extraction thus becomes automatic and efficient. Key words IDL; System; Rice; Rice Information Extraction I. INTRODUCTION Rice is a major staple food for 60% of China’s population. Rice production is crucial to national food security. Extracting remote sensing information on rice growing area and harvest plays an important role in monitoring rice production and policy-making related to rice farming and rice price prediction. Application of remote sensing in agriculture is a part of earth observations from space. Remote sensing data have advantages such as macro-scale, comprehensiveness, objectivity, timeliness, and eco-efficiency, and play important roles in extracting data on rice planting area, reproductive stage identification, growth monitoring, and yield estimation [1], but slow processing speed due to the vast amount of raw data and complex formats limits the use of remote sensing data. Development of new remote sensing applications can meet new challenges in data processing, and general purpose application software may be tailored for specific needs [2]. Multi-source remote sensing data fusion can provide a wealth of information on remote sensing applications for rice farming. As a fourth generation language, IDL is capable of extracting information from raw image data and analyzing shapes [3]. Thus, IDL is a language of choice for devising the rice information extraction systems. II. DESIGN OF RICE REMOTE SENSING INFORMATION SYSTEM A. System dataflow diagram The dataflow diagram shown in Fig. 1 meets the need of rice growth monitoring and yield estimation.

Upload: wang

Post on 01-Mar-2017

215 views

Category:

Documents


3 download

TRANSCRIPT

Page 1: [IEEE 2012 First International Conference on Agro-Geoinformatics - Shanghai, China (2012.08.2-2012.08.4)] 2012 First International Conference on Agro- Geoinformatics (Agro-Geoinformatics)

Design of Remote Sensing System for

Rice Information Extraction

GUO Rui-fang, HUANG Jing-feng

Institute of Remote Sensing and Information Application, Zhejiang University key Laboratory of Polluted Environment Remediation and Ecological Health, Ministry of Education, College of Natural Resources and Environmental Science, Zhejiang UniversityKey Laboratory of Agricultural Remote Sensing and Information System, Zhejiang Province

Hangzhou Chinae-mail: [email protected]; [email protected]

WANG Xiu-zhen

Institute of Remote Sensing and Earth Sciences, Hangzhou Normal UniversityHangzhou Chinae-mail: [email protected]

Abstract—Rice is the world's most important crop

for ensuring food security. Remote sensing has

advantages in collecting data objectively and

accurately on rice planting area, growth and

development, and yield at large scales, but

processing is slow due to vast amount of data, and

there is need for an information extraction system

specially designed for rice. This paper mainly

discussed the key techniques for the rice

data extraction system based on IDL including its

graphical user interface, Xmanager Incident

Response, object graphics system,

image roaming and image scaling, as well as the

function and structure of this system for rice area

extraction, growth monitoring, and yield

forecasting. Rice cultivation data extraction thus

becomes automatic and efficient.

Key words IDL; System; Rice;

Rice Information Extraction

I. INTRODUCTION

Rice is a major staple food for 60% of

China’s population. Rice production is crucial to

national food security. Extracting remote sensing

information on rice growing area and harvest

plays an important role in monitoring rice

production and policy-making related to rice

farming and rice price prediction.

Application of remote sensing in agriculture

is a part of earth observations from space.

Remote sensing data have advantages such as

macro-scale, comprehensiveness, objectivity,

timeliness, and eco-efficiency, and play

important roles in extracting data on rice planting

area, reproductive stage identification, growth

monitoring, and yield estimation [1], but slow

processing speed due to the vast amount of raw

data and complex formats limits the use of

remote sensing data.

Development of new remote sensing

applications can meet new challenges in data

processing, and general purpose application

software may be tailored for specific needs

[2]. Multi-source remote sensing data fusion can

provide a wealth of information on remote

sensing applications for rice farming. As a fourth

generation language, IDL is capable of extracting

information from raw image data and analyzing

shapes [3]. Thus, IDL is a language of choice for

devising the rice information extraction systems.

II. DESIGN OF RICE REMOTE SENSING

INFORMATION SYSTEM

A. System dataflow diagram

The dataflow diagram shown in Fig. 1 meets

the need of rice growth monitoring and yield

estimation.

Page 2: [IEEE 2012 First International Conference on Agro-Geoinformatics - Shanghai, China (2012.08.2-2012.08.4)] 2012 First International Conference on Agro- Geoinformatics (Agro-Geoinformatics)

(1) Input of remote sensing data. This system

utilizes vector data(vector image of study area),

raster data as sources for extracting information

on rice planting area, reproductive stages

identification, growth monitoring, and yield

estimation. The system supports entry of a

number of raster data format such as tiff, img and

binary data, and only vector data with shp type

(2) Image pre-processing. Image

pre-processing includes rectifying,

subsetting, calculation, and masking. The

calculation includes EVI Enhanced vegetation

index NDVI Normalized difference

vegetation index NDSI Normalized

difference snow index), and LSWI Land Surface

Water Index), which are used in the functional

templates.

(3) Time series vegetation indices. Extracting

time series vegetation indices, based on which

rice planting area and reproductive stage

identification are calculated, is the key in rice

growth monitoring and yield estimation. Rice

planting area extraction is based on relations

among various vegetation indices in 1-year time

series, and the reproductive stages are identified

based on NDVI or EVI value variation in the

time series.

(4) Rice yield models. Rice yield is forecasted

by establishing models on relations between

vegetation indices and yield, and the yield in a

study area is predicted using the models.

(5) Data exporting and mapping. Each model

has its own calculation and result, most of which

are presented in raster format, and some are

displayed in text and table forms. Mapping rice

planting areas, tillering and heading status with

IDL are the final output.

B. Model design

To build a system that meets specific needs

in rice growth monitoring and yield forecasting,

with outstanding expandability, flexibility,

performance, and efficiency, the system consists

of data processing template, main template,

output template, and rice growth monitoring

template. The system can be further divided into

data pre-processing template, rice area extraction

template, rice reproductive stage identification

template, rice growth monitoring template, rice

yield estimation template, and data output

template. The structure of system templates are

shown in Fig. 2.

1) Data preprocessing modelThis includes image mosaicing,

subsetting, calculation, and masking. Image

mosaicing program uses multiple remote sensing

images to create an image that covers larger

areas. Subsetting program removes areas outside

the border of study area. Noises need to be

removed because they disguise signals and

reduce quality of images [4]. This template also

calculates EVI, NDVI, NDSI, and LSWI.

2) Rice area extraction model Extraction of rice planting area is the basic of

rice growth monitoring and yield estimation.

Rice fields are irrigated before seedling, and

irrigation continues until mature stage (about one

week before harvesting). During this period the

soil is saturated with water, which is a

characteristic of rice cultivation. Soil water

saturation is the basic of rice area extraction.

Each of the two rice area extraction methods has

its own advantages [5, 6]. The method 1 [5] is

suitable for paddy rice fields in Southern China,

and the method 2 [1] is good in all types of rice

fields.

3) Rice growth stage identification model Plant growth stage identification is

important in global ecological monitoring and

climate change research. This program

identifies rice reproductive stage by analyzing

time series vegetation indices. Rice growth

stages includes seedling, tillering, heading, and

maturing. MOD09A1 data is used to calculate

time series vegetation indices, and change of the

vegetation indices due to change of characteristic

Page 3: [IEEE 2012 First International Conference on Agro-Geoinformatics - Shanghai, China (2012.08.2-2012.08.4)] 2012 First International Conference on Agro- Geoinformatics (Agro-Geoinformatics)

spectrum of rice growth stages indicates the stages of rice growth [7].

Figure 1. Data flowchart of remote sensing system for rice information extraction

Figure 2. The function and structure of remote sensing system for rice information extraction

4) Rice growth monitoring model

Rice growth monitoring is done by

comparing averaged annual rice growth statistics

in previous years, this method provides reference

in rice growth monitoring, which offers timely

information for rice field management and early

yield forecasting [8]. Remote sensing image can

reveal characteristics of vegetation growth at

macroscales. Rice growth status affects yield, and

growth monitoring is an important element in

crop yield forecasting.

5) Rice yield forecast model

Rice harvest is important to China and world

food security as well as environmental

protection. Harvest forecast in a study area is

Data preprocessing model Rice area extraction model

Rice yield forecast model Output model

Rice reproductive stage

identification model

Rice growth monitoring

model

Remote sensing system for

rice information extraction

Out and Mapping

Yield

data

Growth

monitoring

Growth stage

identification

Yield forecast

Image data

Mosaic

Subset

Reduce noise

Calculate

Vector data

Time of flooding

and transplanting Extraction of rice

pixel (method2)

Extraction of rice pixel

(method1)

NDVI, NDSI, LSWI, DEM and NIR NDVI, LSWI and DEM

Page 4: [IEEE 2012 First International Conference on Agro-Geoinformatics - Shanghai, China (2012.08.2-2012.08.4)] 2012 First International Conference on Agro- Geoinformatics (Agro-Geoinformatics)

achieved using models, which are established

with relations between vegetation indices and

known rice yield, and source data are

combination of statistics and field sampling data

[9].

6) Output model Data output template is to export rice area and

growth monitoring information in the form of

Thematic maps according to usage requirements.

The output format can be any format supported

by Window such as JPG and TIFF.

C. User interface design

User interface is designed with user’s needs

and operational patterns and interaction in

mind. As well as, to creat an

initialization command layer, existing formal or

non-formal guidelines should be followed. The

user's interaction should be made as simple and

efficient as possible. The user interface should

tolerate user errors and can undo actions. User is

informed with a progress bar if a task takes long

time to complete. User interface design can

utilize GUI tools to simplify designing processes.

Therefore, the software adopts typical windows

multiple document interface, and the interface

includes menu bar, toolbar, activity panel, Layer

manager window and image display window.

1) Menu designThe Menu consists of a main navigation tab,

subtabs, and their connective dropdown menus,

through which template tasks are executed. Place

the cursor over a tab and left click can trigger a

dropdown menu. Some submenus have

sub-sub-tabs and sub-sub menus. Fig. 3 shows a

submenu of pre-processing templat.

Figure 3. Menu of remote sensing system for rice information extraction

2) Interaction interfaceHuman-computer interaction is how the

system completes tasks with user input and

manipulation. Fig. 4 shows user interface of rice

harvest forecast template:

Fig 4 Interface of batch calculation of NDVI

3) Dialog boxDialog box is the software’s way to

communicate with user. User sets reference

values in the dialog box to execute specified

commands. There are “OK” and “Cancel”

buttons in most dialog boxes. Clicking the “OK”

button will implement the command, and

clicking the “Cancel” button will close the dialog

box without taking action. An error dialog box

will pop up if the system encounters an

error. Fig. 5 shows a dialogue box displaying

error message.

4) Display windowThe system’s display window is used to

display result images. For instance, after

extracting rice area, there will be a display

window on rice area information screen to

Page 5: [IEEE 2012 First International Conference on Agro-Geoinformatics - Shanghai, China (2012.08.2-2012.08.4)] 2012 First International Conference on Agro- Geoinformatics (Agro-Geoinformatics)

display a map showing distribution of rice planting areas as demonstrated in Figure 6:

Figure 5. Error dialog of remote sensing system for rice information extraction

Figure 6. Model interface of remote sensing system for rice information extraction

III. HOW TO USE THE RICE INFORMATION

SYSTEM

This section demonstrates how to use the rice

remote sensing extraction system with an

example: extracting single season rice planting

area and tillering stage growth monitoring data in

Hubei Province in 2010.

Enter the reference values in the cells as

displayed in Fig. 7 The system will automatically

draw the map showing distribution of the rice

planting areas in Hubei Province. The green

areas are not rice planting areas, and it shows that

the single season rice planting areas are mostly

distributed in the central Hubei.

fields in Hubei Province in 2010

Rice tillering stage growth monitoring map compares current data with previous years’ data, the growth status is labeled as excellent, good, fair, poor, and very poor.

Rice tillering stage growth monitoring map

compares current data with previous years’ data,

the growth status is labeled as excellent, good,

fair, poor, and very poor. Enter the reference

values in cells prompted as shown in Fig. 8, the

system can create the rice tillering stage growth

monitoring map automatically. The green color

represents excellent growth status , red is good,

blue is fair, pink is poor, and yellow is very poor

growth status. It can be seen that the overall rice

growth status is about average, and growth status

in some regions’is better than average.

Figure 7. Spatial distribution of single paddy rice

Page 6: [IEEE 2012 First International Conference on Agro-Geoinformatics - Shanghai, China (2012.08.2-2012.08.4)] 2012 First International Conference on Agro- Geoinformatics (Agro-Geoinformatics)

Figure 8. The growth status of single paddy rice of the tillering period in Hubei Province in 2010

IV. CLOSING REMARKS

This paper describes the design and

construction of rice remote sensing information

extraction system. The rice information

extraction system is developed using MODIS

data as source data, IDL techniques. Based on

demand of popularization development and

scalability, this paper used the redevelopment

technology of IDL language and ENVI , as well

as designed the function and structure of this

system for rice area extraction, growth status

monitoring, and yield forecasting. Rice

cultivation data extraction thus becomes

automatic and efficient.

ACKNOWLEDGEMENTS

This research was supported by Ph.D.

Programs Foundation of Ministry of Educational

of China (200100101110035) and National Key

Technologies R&D Program (2011BAD32B01)

REFERENCE

[1] Sun H S. Extracting Planting Area and Growth

Information of Paddy Rice using Multi-temporal

MODIS Data in China, Hangzhou, Zhejiang

University, 2008, 1-157.

[2] Tang Q, Niu Z. Approach for remote sensing system

development based on IDL and ENVI redevelopment.

Computer Applicafions, 2008, 28: 270-276

[3] Yan D W. The introduction and improvement of IDL

visualization [M]. Machine Press, 2003.

[4] Zhao Y S. The principle and method of remote sensing

application and analysis[M].Beijing, Science

Press,2003.

[5] Xiao X M, Boles S, Liu J Y., et al. Mapping paddy rice

agriculture in southern China using multitem MODIS

images. Remote Sensing of Environment, 2005, 95:

480-492

[6] Xiao X M, Boles S, Frolking S, et al. Mapping paddy

rice agriculture in South and Southeast Asia using

multi-temporal MODIS images, Remote Sensing of

Environment , 2006, 100: 95 – 113.

[7] Sun H S, Huang J F, Peng D L. Detecting major

growth stages of paddy rice using MODIS data.

Journal of Remote Sensing, 2009, 1007-4619:

1122-1137.

[8] Wu S X, Mao R Z, Li H J, et al. Review of crop

condition monitoring using remote sensing in china.

Chinese Agriculture Science Bulletin, 2005, 21(3):

319-322.

[9] Deng R, Huang J F, Wang F M, et al. Research on

yield estimation of rice based on remote sensing using

moderate-Resolution Imaging Spectroradiometer

(MODIS) Data : A Case Study of Jiangsu Province,

China. Chinese Journal Rice Science, 2010 , 24 (1) :

87 92