CN105740759A - Middle-season rice information decision tree classification method based on multi-temporal data feature extraction - Google Patents
Middle-season rice information decision tree classification method based on multi-temporal data feature extraction Download PDFInfo
- Publication number
- CN105740759A CN105740759A CN201610030646.3A CN201610030646A CN105740759A CN 105740759 A CN105740759 A CN 105740759A CN 201610030646 A CN201610030646 A CN 201610030646A CN 105740759 A CN105740759 A CN 105740759A
- Authority
- CN
- China
- Prior art keywords
- image
- semilate rice
- rice
- semilate
- decision tree
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/188—Vegetation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
According to a middle-season rice information decision tree classification method based on multi-temporal data feature extraction of the invention, selected GF-1 image data has the advantages of high spatial resolution and high temporal resolution. On the basis, a variety of characteristic parameters for rice distribution extraction based on a single-temporal image are used, the advantages of timing analysis based on a multi-temporal image are utilized, multiple parameters and multiple temporal phases are combined organically, and the distribution of middle-season rice is extracted by means of knowledge decision tree classification. Through use of a variety of characteristic parameters, non-target surface features can be eliminated better. Multi-temporal analysis is conducive to the elimination of wrongly-classified surface features caused by 'different surface features, same spectrum' and the extraction of target surface features. Decision tree classification has the characteristics of being flexible, visual, efficient, and the like. Therefore, by integrating all the advantages, the precision of middle-season rice extraction is further improved. The method is of positive significance both to the food security system of a country and to the commercial application of remote sensing in agriculture.
Description
Technical field
The present invention relates to Mapping remote sensing technology field, be specifically related to a kind of semilate rice information decision tree classification approach based on feature extraction in multi-temporal data.
Background technology
Oryza sativa L. is one of world's Three major grain crops, and sown area accounts for the 15% of world's total area under cultivation.China is Rice Production state the biggest in the world, and Rice Cropping region is very wide, reaches Hainan Province in the south, and north is to Heilongjiang Province, and to the east of Taiwan Province, there is the plantation of Oryza sativa L. in Xi Da Xinjiang Uygur autonomous region.Different geographical Different climate, the kind of Rice Cropping is also not quite similar, and can be divided mainly into three major types according to implantation time: early rice, semilate rice, late rice.Different geographical, early rice, semilate rice, late rice trophophase can difference to some extent.For Hubei Province, throughout the year plantation double; two season early late rice, one season semilate rice, one season late rice etc., the chances are for early rice transplanting time in the first tenday period of a month in the last ten-days period in April to May, and harvest time is in the middle ten days and the last ten days in July.Semilate rice transplanting time last ten-days period in may, harvest time was in JIUYUE last ten-days period to early and middle ten days in October to early and middle ten days in June.Late rice transplanting time is in the last ten-days period in July, and harvest time is in the first tenday period of a month in the last ten-days period in October to November.Monitoring of Paddy Rice Plant Area accounts for the 30% of China's cereal crops gross area, and its yield accounts for the half nearly of total output of grain.Accurately, timely, objective, obtain Monitoring of Paddy Rice Plant Area rapidly, there is many-sided significance perhaps, one, the cultivated area of Oryza sativa L. is related to national economy, the correct cultivated area grasping Oryza sativa L. in time, to establishment socio-economic development planning, it is ensured that national food security, macro adjustments and controls crop mix has very important effect；Its two, Monitoring of Paddy Rice Plant Area promptly and accurately can be country and the decision-makings of local governments at all levels, the macro adjustments and controls measure taked, it is provided that objectively scientific basis；Its three, timely, objective, accurately rice area can also bring huge commercial value.
The method of traditional acquisition large area Monitoring of Paddy Rice Plant Area mainly has sampling survey method and agricultural statistics report method.Sampling survey method refers to according to statistical theory, from whole units of object of study, extract a part of unit carry out investigating and analyzing, and go to infer overall quantative attribute by the quantative attribute of this fractional unit, namely utilize sample characteristics to calculate the investigation method of general characteristic.The precision of the method and the sample number of investigation have and directly contact, if the sample size of investigation is few, then the overall accuracy of statistics can exist bigger deviation, if the sample of investigation is many, then inquiry agency time spent, energy, cost again can be excessive.Agricultural statistics report method is according to administrative unit, from village level unit summarizing and reporting gradually to small towns, counties and cities, province and country.Although the comparison of the method statistics is comprehensive, but it is ageing poor, generally require the cost substantial amounts of time, and give the correct time on step by step there is also certain fail to report, the phenomenon of misrepresentation, certain defect is there is in the accuracy causing data with reliability, on the other hand, by reporting the data of statistics as data refer, but can not reflect Oryza sativa L. plantation distribution situation spatially.Therefore, these traditional methods obtaining crop area increasingly can not meet the needs of reality in ageing, accuracy and Financial cost etc..
At present, along with the development of satellite sensor technology and remote sensing, remote sensing technology application agriculturally also becomes more and more extensive.Remote sensing has macroscopic view, the feature such as objective, real-time, accurate, it is possible to timely, objective and extract large-scale proportion of crop planting information with relatively low Financial cost.Being continuously increased and sensor improving constantly on spatial resolution, spectral resolution, temporal resolution and gray level resolution particular with sensor type, the information utilizing remote sensing image acquisition crops is also more accurate.How to utilize different image data sources to obtain accurate Crop Information and also become the research topic of current remote sensing fields.Oryza sativa L., as main crops, becomes the study hotspot of current many experts and scholars especially.
HUANGQING etc., based on MODIS time series data, the NDVI value rule of each node of analyzing rice growth period, by the NDVI value of each node is set suitable threshold value, extract the distribution of semilate rice, finally with statistical data over the years for standard, have done accuracy assessment.He Yue etc. are by choosing Oryza sativa L. waiting period and the MODIS image of each growth critical period, the separability of analyzing rice field and other atural objects, build decision-tree model, extract late rice cultivated area, and be analyzed with statistics, from administrative district yardstick, the late rice cultivated area extracted is carried out precision analysis and evaluation, achieve good effect.But, the resolution of MODIS image is relatively low, is only 250m, many discrete rice terraces, corresponding connect on MODIS image a Pixel size all without, so, the extraction effect of the rice terrace of Discrete Distribution is unsatisfactory, and particularly hills, southern area is a lot, and the distribution of Oryza sativa L. is not concentrated.It addition, MODIS data are extracted in the process of Oryza sativa L. distribution, simply having used the spectral signature parameters such as NDVI, LSWI, be not combined use with non-spectral characteristic parameter, cause that characteristic parameter is single, the effect of extraction neither be fine.
Relative MODIS data, TM and HJ-1A, spot data have higher image spatial resolution.Zhou Yi etc., with TM image for data source, introduce the non-spectral feature such as DEM, textural characteristics, go out the distribution of semilate rice in conjunction with the Spectra feature extraction of rice ear sprouting period, finally have chosen a number of sample and demonstrate extraction accuracy, achieve good effect.Zheng Changchun etc., on the basis of SPOT5 image, have extracted trial zone semilate rice at heading stage with other atural objects difference of DN value and NDVI value on each wave band by comparative analysis semilate rice, and have adopted visual interpretation and quantitative statistics to evaluate extraction result.But, Mono temporal high resolution image data extract Oryza sativa L., there is a lot of foreign body phenomenon with spectrum in image, can cause the situation of a lot of wrong point, and the precision therefore Oryza sativa L. extracted has large effect.Pan Zhi waits by force TM image by choosing three different phases, and analyzing rice is the Changing Pattern of NDVI value within these three period, obtains Oryza sativa L. distribution substantially, and through verifying, precision has reached 84%, achieves good effect.But, the revisiting period of TM image data is longer, it is desirable to the TM sequential image difficulty obtaining Oryza sativa L. different times is relatively big, and particularly sexual intercourse weather is easy to the generation negative impact of the quality to image.
Summary of the invention
In view of this, it is necessary to provide one can better reject non-targeted atural object, thus Target scalar being carried out the semilate rice information decision tree classification approach based on feature extraction in multi-temporal data of accurately extraction.
A kind of semilate rice information decision tree classification approach based on feature extraction in multi-temporal data, the described semilate rice information decision tree classification approach based on feature extraction in multi-temporal data comprises the following steps:
S1, obtain number image data of high score of respective regions difference phase, and number image data of high score is done radiation calibration pretreatment and atmospheric correction pretreatment respectively；
S2, identify to be combined with the image of transplanting time by the image interpretation of atural objects various in image and be analyzed, the atural object in heading stage image is carried out tentatively distinguishing of typical feature；
The different atural object sample point spectral value on each wave band of S3, record, and draw different atural object sample point spectral value statistical table, find out different atural object difference between each wave band glazing spectrum by the numerical value in analytic statistics table, set up the characteristic parameter of a number image data of multidate high score；
S4, indices to image data carry out threshold value setting, extract the distributed intelligence of semilate rice, extract the distribution of semilate rice accurately in number image data of high score；
S5, adopt the mode of stochastic sampling to lay sample point, by the classification of on-site inspection sample point and Classification in Remote Sensing Image results contrast, adopt the appraisal procedure based on error matrix to evaluate the accuracy of Classification in Remote Sensing Image.
The GF116M image data that the present invention chooses both had had the advantage that high spatial resolution also has high time resolution, on this basis, the various features parameter that the present invention had used when both having employed the distribution of Mono temporal Extraction of Image Oryza sativa L., combine again multi_temporal images and carry out the advantage of Time-Series analysis, many reference amounts are made organically to be combined with multidate, by the mode of Knowledge Decision-making tree classification, extract the distribution of semilate rice.Join quantitative analysis by various features to extract, can better reject non-targeted atural object, simultaneously to multi temporal analysis, it is possible to contributing to rejecting foreign body and compose together the mistake caused and divide atural object can also extract Target scalar further, decision tree classification has flexible, directly perceived, efficiency high.So, no matter comprehensively these advantages, it is possible to improve further the precision that semilate rice extracts, be the grain safety system to country, or to remote sensing commercial applications agriculturally, has positive meaning.
Accompanying drawing explanation
Fig. 1 is the flow chart of steps of the semilate rice information decision tree classification approach based on feature extraction in multi-temporal data of the present invention；
Fig. 2 is the FB(flow block) of the semilate rice information decision tree classification approach based on feature extraction in multi-temporal data of the present invention；
Fig. 3 is the FB(flow block) of step S4 in Fig. 1；
Fig. 4 is the FB(flow block) of step S41 in Fig. 1；
Fig. 5 is another FB(flow block) of the semilate rice information decision tree classification approach based on feature extraction in multi-temporal data of the present invention.
Detailed description of the invention
In order to make the purpose of the present invention, technical scheme and advantage clearly understand, below in conjunction with drawings and Examples, the present invention is further elaborated, it should be understood that, specific embodiment described herein only in order to explain the present invention, is not intended to limit the present invention.
The embodiment of the present invention provides a kind of semilate rice information decision tree classification approach based on feature extraction in multi-temporal data, and as depicted in figs. 1 and 2, the described semilate rice information decision tree classification approach based on feature extraction in multi-temporal data comprises the following steps:
S1, obtain number image data of high score of respective regions difference phase, and number image data of high score is done radiation calibration pretreatment and atmospheric correction pretreatment respectively.
Concrete, number image data of high score that the present invention adopts is the GF116M image data downloading respective regions difference phase from China Resource Satellite Applied Center.Mainly include the image data of three phases: the transplanting time of semilate rice, time range are a number image data of high score in the first tenday period of a month in the last ten-days period in May to June；The heading stage of semilate rice, time range are number image data of high score of early and middle ten days in the last ten-days period in July to August；The milk stage of semilate rice, time range are a number image data of high score in the JIUYUE last ten-days period to the first tenday period of a month in October.It addition, in ENVIClassic, select the TopographicModeling in menu bar, it is possible to derive gradient file according to the shp vector border file of respective regions and DEM image data.
Meanwhile, utilize ENVI that the GF116M image of three phases does radiation calibration pretreatment and atmospheric correction pretreatment respectively.If projection coordinate's system of two width images is different, also need projection coordinate's system of converting image.Utilize shp vector border file that the image of two phases is done cutting.Image after cutting needs to do geometry essence registration in ENVI, global error have in a pixel coverage, to ensure that same ground point is completely superposed on the image that different time, different-waveband and different sensors obtain.
S2, identify to be combined with the image of transplanting time by the image interpretation of atural objects various in image and be analyzed, the atural object in heading stage image is carried out tentatively distinguishing of typical feature.
Wherein, described image interpretation mark includes the shape of various atural objects in image, size, color, texture, shade, figure, three-dimensional appearance etc.；Described preliminary differentiation is that the atural object in heading stage image is divided into semilate rice, late rice, corn and soybean, Cotton Gossypii, settlement place, road, water body, open ground, the ten kinds of typical features in evergreen ground.
Therefore, is combined with the image of transplanting time by image interpretation marks such as the shape of atural objects various in image, size, color, texture, shade, figure, three-dimensional appearance and is analyzed, it is possible to tentatively the atural object in heading stage image is divided into semilate rice, late rice, corn and soybean, Cotton Gossypii, settlement place, road, water body, open ground, evergreen ten kinds of typical features.
The different atural object sample point spectral value on each wave band of S3, record, and draw different atural object sample point spectral value statistical table, find out different atural object difference between each wave band glazing spectrum by the numerical value in analytic statistics table, set up the characteristic parameter of a number image data of multidate high score.
Concrete, on semilate rice image at heading stage, choose a number of typical feature sample point.Record different atural object sample point spectral value on each wave band (mainly visible red wave band, near infrared band and short infrared wave band), and draw different atural object sample point spectral value statistical table, statistical table comprises different atural object in the maximum of each wave band glazing spectrum, minima, average and standard deviation.Different atural object difference between each wave band glazing spectrum is found out by the numerical value in analytic statistics table.
Different ground object sample point, it is also possible to as the checkpoint of trial inspection after extracting.Different atural objects, due to the difference of its composition, structure, physicochemical property and growing environment, the eigenvalue such as its NDVI, EVI, LSWI and DEM also there are differences each other, and these eigenvalues that can be used to distinguish difference atural object are referred to as characteristic parameter.
S4, indices to image data carry out threshold value setting, extract the distributed intelligence of semilate rice, extract the distribution of semilate rice accurately in number image data of high score.
As it is shown on figure 3, described step S4 include following step by step:
S41, in semilate rice image at heading stage, forest land, mountain region and the rice terrace of the overwhelming majority, nonvegetated area, settlement place, road, open ground and corn and soybean are rejected；
Wherein, as shown in Figure 4, described step S41 include following step by step:
S411, according to the growth characteristics of semilate rice, it is usually and is planted in field in easy flights, the actual plantation situation therefore in semilate rice image at heading stage, according to semilate rice, gradient condition is reasonably set threshold value, it is possible to the forest land, mountain region of the overwhelming majority and rice terrace are distinguished and weeded out.
Concrete, if described gradient threshold value is T1, namely in described semilate rice image at heading stage, if the gradient on ground is less than T1, then this region is likely the field being suitable for doing plantation semilate rice, if the gradient on ground is more than T1, then this region belongs to the forest land, mountain region being not suitable for doing the field of plantation semilate rice, therefore, by gradient threshold value, forest land, mountain region is rejected.
S412, choose the GF116M image in semilate rice heading period, obtain NDVI, EVI value of image, by NDVI and EVI setting reasonable threshold value, tentatively vegetation region in image and nonvegetated area being made a distinction.
Owing to vegetation is very low at visible red wave band reflectance, significantly high near infrared band reflectance, so vegetation is relative to non-vegetation, its NDVI and EVI value is higher, but owing to precipitation and moisture affect, NDVI and the EVI value of vegetation can be less than normal than ever, even has overlapping the intersection with NDVI and the EVI value of non-vegetation.Therefore, it can, by NDVI and EVI is set reasonable threshold value, tentatively vegetation region in image and nonvegetated area be made a distinction.
If the reasonable threshold value of described NDVI is T2, the reasonable threshold value of described EVI is T3, namely in described semilate rice image at heading stage, if the NDVI value of partial image more than T2, EVI value more than T3, then this partial image is vegetation area, otherwise, then be nonvegetated area territory.Therefore, forest land, mountain region can be rejected by the reasonable threshold value of NDVI and EVI.
S413, in semilate rice image at heading stage, visible red wave band reflectance value is set threshold value, it is possible to semilate rice and settlement place, road, open ground are made a distinction and rejected.
Analyze the spectral reflectance values of various atural objects on semilate rice image at heading stage it appeared that, settlement place, road, open ground spectral reflectance values all higher on each wave band, and the spectral reflectivity of semilate rice is relatively low on visible red wave band, so the spectral reflectance values of settlement place, road, open ground and semilate rice differs greatly at visible red wave band.
If described visible red wave band reflectance threshold value is T4, namely in described semilate rice image at heading stage, if the visible red wave band reflectance of partial image is more than T4, then this region is settlement place, road, open ground, if the visible red wave band reflectance of partial image is less than T4, then this region is the planting site of semilate rice, corn and soybean, Cotton Gossypii, evergreen, late rice, therefore, by visible red wave band reflectance threshold value, settlement place, road, open ground can be made a distinction and be rejected.
S414, in semilate rice image at heading stage, the reflectance value of short infrared wave band is set threshold value, it is possible to the corn and soybean in semilate rice image at heading stage rejected.
On semilate rice image at heading stage, the spectral reflectance values of Semen sojae atricolor and Semen Maydis is higher on short infrared wave band, the spectral reflectance values of semilate rice is relatively low on short infrared wave band, therefore, by the reflectance value of short infrared wave band is set threshold value, it is possible to the corn and soybean in semilate rice image at heading stage is rejected.
If the reflectance threshold value of described short infrared wave band is T5, namely in described semilate rice image at heading stage, if the reflectance value of the short infrared wave band of partial image is less than T5, then this region is the planting site of semilate rice, Cotton Gossypii, evergreen, late rice, if the visible red wave band reflectance of partial image is more than T5, then this region is the planting site of corn and soybean, therefore, by visible red wave band reflectance threshold value, the planting site of corn and soybean can be made a distinction and be rejected.
S42, in the image of semilate rice Different Transplanting Periods, LSWI value is set threshold value, by middle rice field and cotton field with evergreen make a distinction, and reject cotton field with evergreenly.
In the image of semilate rice Different Transplanting Periods, just having transplanted due to semilate rice or be about to transplant, in now, the water content in rice field is more, and corresponding LSWI value is higher, and the LSWI value of Cotton Gossypii, evergreen ground is relatively low.
If described LSWI threshold value is T6, namely in the image of described semilate rice Different Transplanting Periods, if the reflectance value of the short infrared wave band of partial image is more than T6, then this region is the planting site of semilate rice, if the visible red wave band reflectance of partial image is less than T6, then this region is the planting site of Cotton Gossypii, evergreen, late rice, therefore, by visible red wave band reflectance threshold value, Cotton Gossypii, planting site evergreen, late rice can be made a distinction and be rejected.
For improving the precision that semilate rice extracts further, as it is shown in figure 5, the described semilate rice information decision tree classification approach based on feature extraction in multi-temporal data also includes step S4a,
Step S4a: by the transplanting time of semilate rice, heading stage and period milk stage three image NDVI set threshold value, and to three period image short-wave infrared spectral value set threshold value, will " foreign body is with spectrum " pixel and the late rice rejecting do not rejected completely.
Concrete, the heading stage of described semilate rice is the period that semilate rice growth is very vigorous, from the transplanting time of semilate rice to heading stage, the NDVI value of semilate rice can significantly increase, but, from the heading poplar bloassom phase of semilate rice, arriving the milking maturity harvest time of semilate rice again, the NDVI value of semilate rice can be gradually reduced again.Further, since the change of middle rice field water content, the spectral value of the short infrared wave band of water content sensitivity also can significantly be increased.Therefore, by the transplanting time of semilate rice, heading stage and period milk stage three image NDVI set threshold value, and to three period image short-wave infrared spectral value set threshold value, it is possible to improve the precision that semilate rice extracts further.Meanwhile, by this step, more " foreign body is with the spectrum " pixel divided wrong in Mono temporal and the late rice do not rejected completely effectively can be rejected.
S5, adopt the mode of stochastic sampling to lay sample point, by the classification of on-site inspection sample point and Classification in Remote Sensing Image results contrast, adopt the appraisal procedure based on error matrix to evaluate the accuracy of Classification in Remote Sensing Image.
Concrete, the actual cultivated area of semilate rice is obtained by the mode of artificial field exploring, using the high-resolution remote sensing image of survey region as base map, carries out on-site inspection, obtain semilate rice cultivated area and distributed data accurately in units of independent field.The Accuracy Verification of the semilate rice distribution that remote sensing image extracts adopts the appraisal procedure based on error matrix, the mode adopting simple random sampling lays sample point, by the classification of on-site inspection sample point and Classification in Remote Sensing Image results contrast, the appraisal procedure based on error matrix is adopted to evaluate the accuracy of Classification in Remote Sensing Image, different spatial resolutions remote sensing image, difference according to scope of experiment size, the number of its sample point there are differences.
Concrete, described appraisal procedure is: first set up the sample point nicety of grading on-site inspection table described in the error matrix form of semilate rice classification and table one with corresponding categorical data according to the sample data of on-site inspection, then calculate, according to error matrix form, total area precision, user's precision and the Producer precision that semilate rice plantation classification is extracted, just can reflect the nicety of grading of semilate rice based on this three classes nicety of grading from different sides.
Described total area precision, user's precision and Producer precision describe the nicety of grading of semilate rice from different sides, are simple and easy to do and have the evaluation index of statistical significance.Wherein, overall classification accuracy statement is to each random sample, the probability that the result classified is consistent with testing data type.User's precision refers to appoint from classification results take a random sample, the conditional probability that its type having is identical with ground actual type.Producer accuracy representing is relative to any one random sample in testing data, the conditional probability that on classification chart, the classification results in same place is consistent with it.
Wherein, the sample number computing formula of described stochastic sampling is:
Wherein, P is sample variation degree, and N is overall number, and Z is the confidence level of survey result, and e is sampling error scope.
Using the field operation on-site inspection acquired results actual value data as Experimental Area, the semilate rice plantation data obtained with utilizing Remote sensing image classification to extract contrast, calculate the precision carrying out semilate rice crop coverage measurement based on remote sensing image, including the total area precision of classification, user's precision and Producer precision, wherein
Total area accuracy computation formula is:
User's accuracy computation formula:
Producer accuracy computation formula:
Wherein, SSurveyGained semilate rice area, S is extracted for remote sensing imageRealFor on-site inspection gained semilate rice cultivated area.I1For sample point is correctly divided into counting of semilate rice, I2For sample point belonging to non-semilate rice but being classified as counting of semilate rice, I3For sample point belonging to semilate rice but being divided into counting of non-semilate rice.According to investigation situation, investigation sample point and distribution on the spot can be generated and extract charting and the error matrix form of point, such as table 1 below:
Table 1 sample point nicety of grading on-site inspection table
The precision of the precision through verifying on the spot with traditional extraction nicety of grading and the extracting method of other scholars is contrasted, as shown in table 2 below:
Table 2 present invention extracts the accuracy comparison of semilate rice with additive method
In sum, semilate rice information decision tree classification approach based on feature extraction in multi-temporal data of the present invention, the GF116M image data that it is chosen both had had the advantage that high spatial resolution also has high time resolution, on this basis, the various features parameter that the present invention had used when both having employed the distribution of Mono temporal Extraction of Image Oryza sativa L., combine again multi_temporal images and carry out the advantage of Time-Series analysis, many reference amounts are made organically to be combined with multidate, by the mode of Knowledge Decision-making tree classification, extract the distribution of semilate rice.Various features parameter, it is possible to better reject non-targeted atural object, multi temporal analysis, it is possible to contributing to rejecting foreign body and compose together the mistake caused and divide atural object can also extract Target scalar further, decision tree classification has flexible, directly perceived, efficiency high.So, no matter comprehensively these advantages, it is possible to improve further the precision that semilate rice extracts, be the grain safety system to country, or to remote sensing commercial applications agriculturally, has positive meaning.
The foregoing is only presently preferred embodiments of the present invention, not in order to limit the present invention, all within the spirit and principles in the present invention, any amendment of making, equivalent replacement, improvement etc., should be included within protection scope of the present invention.
Claims (7)
1. the semilate rice information decision tree classification approach based on feature extraction in multi-temporal data, it is characterised in that the described semilate rice information decision tree classification approach based on feature extraction in multi-temporal data comprises the following steps:
S1, obtain number image data of high score of respective regions difference phase, and number image data of high score is done radiation calibration pretreatment and atmospheric correction pretreatment respectively；
S2, identify to be combined with the image of transplanting time by the image interpretation of atural objects various in image and be analyzed, the atural object in heading stage image is carried out tentatively distinguishing of typical feature；
The different atural object sample point spectral value on each wave band of S3, record, and draw different atural object sample point spectral value statistical table, find out different atural object difference between each wave band glazing spectrum by the numerical value in analytic statistics table, set up the characteristic parameter of a number image data of multidate high score；
S4, indices to image data carry out threshold value setting, extract the distributed intelligence of semilate rice, extract the distribution of semilate rice accurately in number image data of high score；
S5, adopt the mode of stochastic sampling to lay sample point, by the classification of on-site inspection sample point and Classification in Remote Sensing Image results contrast, adopt the appraisal procedure based on error matrix to evaluate the accuracy of Classification in Remote Sensing Image.
2. the semilate rice information decision tree classification approach based on feature extraction in multi-temporal data according to claim 1, it is characterised in that number image data of high score of described different phases includes:
The transplanting time of semilate rice, time range are a number image data of high score in the first tenday period of a month in the last ten-days period in May to June；
The heading stage of semilate rice, time range are number image data of high score of early and middle ten days in the last ten-days period in July to August；
The milk stage of semilate rice, time range are a number image data of high score in the JIUYUE last ten-days period to the first tenday period of a month in October.
3. the semilate rice information decision tree classification approach based on feature extraction in multi-temporal data according to claim 1, it is characterised in that described step S4 include following step by step:
S41, in semilate rice image at heading stage, forest land, mountain region and the rice terrace of the overwhelming majority, nonvegetated area, settlement place, road, open ground and corn and soybean are rejected；
S42, in the image of semilate rice Different Transplanting Periods, LSWI value is set threshold value, by middle rice field and cotton field with evergreen make a distinction, and reject cotton field with evergreenly.
4. the semilate rice information decision tree classification approach based on feature extraction in multi-temporal data according to claim 1, it is characterised in that described step S41 include following step by step:
S411, in semilate rice image at heading stage, the actual plantation situation according to semilate rice, gradient condition is reasonably set threshold value, it is possible to forest land, mountain region and the rice terrace of the overwhelming majority are distinguished and weeded out；
S412, in semilate rice image at heading stage, obtain NDVI, EVI value of image, NDVI and EVI set reasonable threshold value, it is possible to preliminary vegetation region in image and nonvegetated area are made a distinction；
S413, in semilate rice image at heading stage, visible red wave band reflectance value is set threshold value, it is possible to semilate rice and settlement place, road, open ground are made a distinction and rejected；
S414, in semilate rice image at heading stage, the reflectance value of short infrared wave band is set threshold value, it is possible to the corn and soybean in semilate rice image at heading stage rejected.
5. the semilate rice information decision tree classification approach based on feature extraction in multi-temporal data according to claim 1, it is characterised in that the sample number computing formula of described stochastic sampling is:
Wherein, P is sample variation degree, and N is overall number, and Z is the confidence level of survey result, and e is sampling error scope.
6. the semilate rice information decision tree classification approach based on feature extraction in multi-temporal data according to claim 1, it is characterized in that, described carry out the accuracy computation of semilate rice crop coverage measurement based on remote sensing image and include three below parameter: the total area precision of classification, user's precision and Producer precision；
Described total area accuracy computation formula is:
User's accuracy computation formula:
Producer accuracy computation formula:
Wherein, SSurveyGained semilate rice area, S is extracted for remote sensing imageRealFor on-site inspection gained semilate rice cultivated area.I1For sample point is correctly divided into counting of semilate rice, I2For sample point belonging to non-semilate rice but being classified as counting of semilate rice, I3For sample point belonging to semilate rice but being divided into counting of non-semilate rice.
7. the semilate rice information decision tree classification approach based on feature extraction in multi-temporal data according to claim 1, it is characterised in that the described semilate rice information decision tree classification approach based on feature extraction in multi-temporal data also includes step S4a,
Step S4a: by the transplanting time of semilate rice, heading stage and period milk stage three image NDVI set threshold value, and to three period image short-wave infrared spectral value set threshold value, will " foreign body is with spectrum " pixel and the late rice rejecting do not rejected completely.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610030646.3A CN105740759B (en) | 2016-01-15 | 2016-01-15 | Semilate rice information decision tree classification approach based on feature extraction in multi-temporal data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610030646.3A CN105740759B (en) | 2016-01-15 | 2016-01-15 | Semilate rice information decision tree classification approach based on feature extraction in multi-temporal data |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105740759A true CN105740759A (en) | 2016-07-06 |
| CN105740759B CN105740759B (en) | 2019-01-15 |
Family
ID=56246295
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610030646.3A Active CN105740759B (en) | 2016-01-15 | 2016-01-15 | Semilate rice information decision tree classification approach based on feature extraction in multi-temporal data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105740759B (en) |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106599844A (en) * | 2016-12-14 | 2017-04-26 | 中国科学院南京地理与湖泊研究所 | Method for automatically extracting paddy rice growing region based on MODIS |
| CN107273820A (en) * | 2017-05-26 | 2017-10-20 | 中国科学院遥感与数字地球研究所 | A kind of Land Cover Classification method and system |
| CN107314990A (en) * | 2017-08-28 | 2017-11-03 | 北京师范大学 | A kind of spring maize remote sensing recognition method |
| CN107527014A (en) * | 2017-07-20 | 2017-12-29 | 武汉珈和科技有限公司 | Crops planting area RS statistics scheme of sample survey design method at county level |
| CN107832697A (en) * | 2017-11-01 | 2018-03-23 | 中国科学院地理科学与资源研究所 | The processing method and system of notoginseng planting information rapid extraction |
| CN108364284A (en) * | 2018-01-18 | 2018-08-03 | 中国科学院遥感与数字地球研究所 | A kind of opium poppy extracting method and system based on remote sensing |
| CN109345555A (en) * | 2018-10-15 | 2019-02-15 | 中科卫星应用德清研究院 | Rice, which is carried out, based on multidate multi- source Remote Sensing Data data knows method for distinguishing |
| CN110189043A (en) * | 2019-06-04 | 2019-08-30 | 中国地质大学（北京） | It is a kind of based on high score satellite remote sensing date using land resource analysis system |
| CN110516552A (en) * | 2019-07-29 | 2019-11-29 | 南京航空航天大学 | A kind of multipolarization radar image classification method and system based on timing curve |
| CN110647873A (en) * | 2019-11-14 | 2020-01-03 | 山东省农业可持续发展研究所 | Peanut planting area remote sensing identification method and system |
| CN111368716A (en) * | 2020-02-28 | 2020-07-03 | 重庆市规划和自然资源调查监测院 | Geological disaster catastrophe farmland extraction method based on multi-source time-space data |
| CN111507303A (en) * | 2020-04-28 | 2020-08-07 | 同济大学 | Wetland plant species detection method |
| CN111521563A (en) * | 2020-03-20 | 2020-08-11 | 航天信德智图（北京）科技有限公司 | Rice extraction method based on GF-1 and MODIS space-time fusion |
| CN113221788A (en) * | 2021-05-21 | 2021-08-06 | 广东省科学院广州地理研究所 | Method and device for extracting ridge culture characteristics of field blocks |
| CN113313671A (en) * | 2021-04-30 | 2021-08-27 | 武汉光谷信息技术股份有限公司 | Rice yield estimation method combining satellite images and MODIS data |
| CN115631419A (en) * | 2022-12-08 | 2023-01-20 | 航天宏图信息技术股份有限公司 | Method and device for extracting rice planting area and spatial distribution based on change detection |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101858971A (en) * | 2010-06-02 | 2010-10-13 | 浙江大学 | Rice yield remote sensing estimation method based on MODIS data |
| CN102668899A (en) * | 2012-03-28 | 2012-09-19 | 北京师范大学 | Crop planting mode recognition method |
| CN104915674A (en) * | 2014-10-24 | 2015-09-16 | 北京师范大学 | Landsat8 and MODIS fusion-construction high space-time resolution data identification autumn grain crop method |
| CN105005782A (en) * | 2014-04-24 | 2015-10-28 | 中国科学院遥感与数字地球研究所 | Fine method for global vegetation classification based on multi-temporal remote sensing data and spectroscopic data |
-
2016
- 2016-01-15 CN CN201610030646.3A patent/CN105740759B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101858971A (en) * | 2010-06-02 | 2010-10-13 | 浙江大学 | Rice yield remote sensing estimation method based on MODIS data |
| CN102668899A (en) * | 2012-03-28 | 2012-09-19 | 北京师范大学 | Crop planting mode recognition method |
| CN105005782A (en) * | 2014-04-24 | 2015-10-28 | 中国科学院遥感与数字地球研究所 | Fine method for global vegetation classification based on multi-temporal remote sensing data and spectroscopic data |
| CN104915674A (en) * | 2014-10-24 | 2015-09-16 | 北京师范大学 | Landsat8 and MODIS fusion-construction high space-time resolution data identification autumn grain crop method |
Cited By (22)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106599844A (en) * | 2016-12-14 | 2017-04-26 | 中国科学院南京地理与湖泊研究所 | Method for automatically extracting paddy rice growing region based on MODIS |
| CN107273820A (en) * | 2017-05-26 | 2017-10-20 | 中国科学院遥感与数字地球研究所 | A kind of Land Cover Classification method and system |
| CN107527014A (en) * | 2017-07-20 | 2017-12-29 | 武汉珈和科技有限公司 | Crops planting area RS statistics scheme of sample survey design method at county level |
| CN107314990A (en) * | 2017-08-28 | 2017-11-03 | 北京师范大学 | A kind of spring maize remote sensing recognition method |
| CN107314990B (en) * | 2017-08-28 | 2020-01-17 | 北京师范大学 | Spring corn remote sensing identification method |
| CN107832697B (en) * | 2017-11-01 | 2019-11-08 | 中国科学院地理科学与资源研究所 | The processing method and system of notoginseng planting information rapidly extracting |
| CN107832697A (en) * | 2017-11-01 | 2018-03-23 | 中国科学院地理科学与资源研究所 | The processing method and system of notoginseng planting information rapid extraction |
| CN108364284A (en) * | 2018-01-18 | 2018-08-03 | 中国科学院遥感与数字地球研究所 | A kind of opium poppy extracting method and system based on remote sensing |
| CN109345555B (en) * | 2018-10-15 | 2020-08-25 | 中科卫星应用德清研究院 | Method for identifying rice based on multi-temporal multi-source remote sensing data |
| CN109345555A (en) * | 2018-10-15 | 2019-02-15 | 中科卫星应用德清研究院 | Rice, which is carried out, based on multidate multi- source Remote Sensing Data data knows method for distinguishing |
| CN110189043A (en) * | 2019-06-04 | 2019-08-30 | 中国地质大学（北京） | It is a kind of based on high score satellite remote sensing date using land resource analysis system |
| CN110189043B (en) * | 2019-06-04 | 2021-08-03 | 中国地质大学（北京） | Usable land resource analysis system based on high-score satellite remote sensing data |
| CN110516552A (en) * | 2019-07-29 | 2019-11-29 | 南京航空航天大学 | A kind of multipolarization radar image classification method and system based on timing curve |
| CN110516552B (en) * | 2019-07-29 | 2021-02-05 | 南京航空航天大学 | Multi-polarization radar image classification method and system based on time sequence curve |
| CN110647873A (en) * | 2019-11-14 | 2020-01-03 | 山东省农业可持续发展研究所 | Peanut planting area remote sensing identification method and system |
| CN111368716A (en) * | 2020-02-28 | 2020-07-03 | 重庆市规划和自然资源调查监测院 | Geological disaster catastrophe farmland extraction method based on multi-source time-space data |
| CN111368716B (en) * | 2020-02-28 | 2023-04-28 | 重庆市规划和自然资源调查监测院 | Geological disaster damage cultivated land extraction method based on multi-source space-time data |
| CN111521563A (en) * | 2020-03-20 | 2020-08-11 | 航天信德智图（北京）科技有限公司 | Rice extraction method based on GF-1 and MODIS space-time fusion |
| CN111507303A (en) * | 2020-04-28 | 2020-08-07 | 同济大学 | Wetland plant species detection method |
| CN113313671A (en) * | 2021-04-30 | 2021-08-27 | 武汉光谷信息技术股份有限公司 | Rice yield estimation method combining satellite images and MODIS data |
| CN113221788A (en) * | 2021-05-21 | 2021-08-06 | 广东省科学院广州地理研究所 | Method and device for extracting ridge culture characteristics of field blocks |
| CN115631419A (en) * | 2022-12-08 | 2023-01-20 | 航天宏图信息技术股份有限公司 | Method and device for extracting rice planting area and spatial distribution based on change detection |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105740759B (en) | 2019-01-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105740759A (en) | Middle-season rice information decision tree classification method based on multi-temporal data feature extraction | |
| Azar et al. | Assessing in-season crop classification performance using satellite data: A test case in Northern Italy | |
| CN107527014A (en) | Crops planting area RS statistics scheme of sample survey design method at county level | |
| CN105372672B (en) | Southern winter kind crops planting area extracting method based on time series data | |
| CN109345555A (en) | Rice, which is carried out, based on multidate multi- source Remote Sensing Data data knows method for distinguishing | |
| CN105608293A (en) | Forest aboveground biomass inversion method and system fused with spectrum and texture features | |
| CN110909933B (en) | Agricultural drought rapid diagnosis and evaluation method coupling crop model and machine learning language | |
| CN102175209B (en) | Effective sampling method for crop cultivated area measurement under support of historical remote sensing product data | |
| CN108984803A (en) | A kind of method and system of crop yield spatialization | |
| CN105004320A (en) | Gaofen satellite data land surface vegetation coverage retrieval method and system | |
| Dehkordi et al. | Yield gap analysis using remote sensing and modelling approaches: Wheat in the northwest of Iran | |
| CN102194127A (en) | Multi-frequency synthetic aperture radar (SAR) data crop sensing classification method | |
| Kotsuki et al. | SACRA–a method for the estimation of global high-resolution crop calendars from a satellite-sensed NDVI | |
| Fan et al. | Agro-ecological suitability assessment of Chinese Medicinal Yam under future climate change | |
| CN104199027A (en) | Method for realizing large-area near real-time monitoring on phenological period of rice based on compactly polarimetric radar | |
| Dong et al. | Using RapidEye imagery to identify within-field variability of crop growth and yield in Ontario, Canada | |
| CN104834971A (en) | Farmland nutrient management partitioning method based on GIS and RS | |
| CN108205718A (en) | Production method and system are surveyed in a kind of cereal crops sampling | |
| Boschetti et al. | Estimation of rice production at regional scale with a Light Use Efficiency model and MODIS time series. | |
| Belal et al. | Precision farming technologies to increase soil and crop productivity | |
| Zhu et al. | UAV flight height impacts on wheat biomass estimation via machine and deep learning | |
| Li et al. | Monitoring spatial and temporal patterns of rubber plantation dynamics using time-series landsat images and *** earth engine | |
| Ren et al. | Based on historical weather data to predict summer field-scale maize yield: Assimilation of remote sensing data to WOFOST model by ensemble Kalman filter algorithm | |
| Wei et al. | Application of remote sensing technology in crop estimation | |
| Yang et al. | Feature extraction of cotton plant height based on DSM difference method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| PE01 | Entry into force of the registration of the contract for pledge of patent right |
Denomination of invention: Mid season rice information decision tree classification method based on feature extraction from multi temporal data Effective date of registration: 20211123 Granted publication date: 20190115 Pledgee: Wuhan area branch of Hubei pilot free trade zone of Bank of China Ltd. Pledgor: WUHAN JIAHE TECHNOLOGY CO.,LTD. Registration number: Y2021420000130 |
|
| PE01 | Entry into force of the registration of the contract for pledge of patent right |