CN108363949B - Cotton remote sensing monitoring method based on phenological analysis - Google Patents
Cotton remote sensing monitoring method based on phenological analysis Download PDFInfo
- Publication number
- CN108363949B CN108363949B CN201711500930.3A CN201711500930A CN108363949B CN 108363949 B CN108363949 B CN 108363949B CN 201711500930 A CN201711500930 A CN 201711500930A CN 108363949 B CN108363949 B CN 108363949B
- Authority
- CN
- China
- Prior art keywords
- cotton
- remote sensing
- sensing image
- information
- interfering
- 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.)
- Active
Links
Images
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
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
Abstract
The invention belongs to the field of agricultural remote sensing, and particularly relates to a cotton remote sensing monitoring method based on phenological analysis, which comprises the following steps: combining the spectral characteristic analysis of cotton in the region and the phenological analysis of interference crops to determine the key time phase of cotton identification; obtaining a key phenological period remote sensing image and preprocessing the image; performing multi-scale segmentation on the multi-temporal image, constructing classifiers in a layering manner, and preliminarily extracting cotton information; performing secondary segmentation on the primary cotton information result and the objects which are possibly subjected to mixed separation, constructing a classifier by combining spectral features and textural features of cotton, removing the mixed separation, and supplementing missed cotton information; and merging the extracted cotton information subclasses and outputting a drawing. According to the invention, the automatic extraction of the cotton planting information is realized by combining the phenological period analysis of cotton and interfering crops with the layered classification method based on the remote sensing image, and the spectral difference and the textural features of cotton and other interfering crops are fully considered, so that the extraction result of the cotton planting information is more accurate and rapid, and the method is easy to popularize and apply.
Description
Technical Field
The invention belongs to the technical field of remote sensing application, and particularly relates to a cotton remote sensing monitoring method based on phenological analysis.
Background
China is the largest world cotton-producing country, and cotton is the main raw material for making clothes and quilts in China. The quick and accurate extraction of the cotton planting information has important significance for mastering the national cotton planting scale, predicting the cotton yield, even for the economic development of the textile industry and the like.
At present, in the aspect of remote sensing monitoring of crops, remote sensing monitoring research aiming at cotton crop planting area is less, and great challenges exist in method precision and popularization, particularly in northern areas of China, the agricultural planting structure is complex, the land parcel is small, and great challenges are provided for accurate acquisition of cotton planting information. With the development of remote sensing technology, novel satellite sensors are continuously emerging, and multi-temporal and multi-resolution remote sensing image data sequences are formed in the same area, so that the application of multi-temporal remote sensing images is more common. Therefore, based on the multi-temporal remote sensing data sequence and in combination with the growth period characteristics of cotton, the automatic and accurate extraction of cotton planting information is a problem to be solved urgently.
Disclosure of Invention
The invention aims to solve the technical problem of providing a cotton remote sensing monitoring method based on phenological analysis, which can automatically and accurately extract cotton planting information.
In order to achieve the purpose, the invention provides a cotton remote sensing monitoring method based on phenological analysis, which comprises the following steps:
a cotton remote sensing monitoring method based on phenological analysis comprises the following steps:
firstly, combining the analysis of the main crop phenology and the cotton spectral characteristic analysis in the area to determine the key time phase of the cotton remote sensing monitoring;
1.1, determining the types and the phenological conditions of main interfering crops in the cotton planting period, and making a phenological table, wherein the main interfering crops comprise winter wheat, corn, peanut and soybean;
1.2, drawing a cotton spectral characteristic curve;
1.3, establishing a remote sensing image interpretation mark of the cotton growth stage by combining the analysis of the cotton spectral characteristic curves of different growth stages and based on the visual interpretation of the remote sensing image;
the specific process is as follows:
the cotton is in a sowing period and a seedling emergence period, and the remote sensing images show bare land information; when the cotton is in the bud stage, the remote sensing image is pink, and the ground characteristics of the cotton at the initial vegetation stage are presented; the cotton is in the boll stage and boll opening stage, the remote sensing image is rosy, and the vegetation ground characteristic of flourishing cotton is presented; when the cotton is in the harvesting period, the remote sensing image is light red, and vegetation ground characteristics of withered cotton are presented;
1.4, gradient division;
dividing a key cotton identification period into four gradients, wherein the first gradient is an ungrooved period, a sowing period and a seedling period, and the surface characteristic is a bare land characteristic; the second gradient is a bud period, a flower bell period and a boll opening period, and the ground surface characteristic is a vegetation characteristic; the third gradient is a mature harvest period, and the ground surface characteristic is a vegetation characteristic; the fourth gradient is that after the mature harvest period, the surface characteristic is a bare land characteristic;
1.5, selecting a key time phase for cotton identification in each gradient range by adopting a moving window method;
the specific process is as follows:
in each gradient, moving a window in a phenological table by taking ten days as a step length, and determining a time phase when the number of the interfering crop species opposite to the surface characteristics of the cotton surface is maximum as a key time phase for identifying the cotton in the gradient;
and 1.6, determining the key cotton remote sensing monitoring time phase in the key cotton identification time phase in each gradient by combining the availability of remote sensing data.
Selecting a remote sensing image of a period with vigorous vegetation growth in the first stage between the time phases of a boll stage and a boll opening stage of cotton;
acquiring a key time phase remote sensing image, and preprocessing the key time phase remote sensing image;
taking the remote sensing image of one time phase as a reference, and carrying out image registration on the remote sensing images of other time phases;
step four, carrying out multi-scale segmentation on the multi-temporal remote sensing image, constructing classifiers in a layering manner, and preliminarily extracting cotton;
the specific process is as follows:
4.1, performing multi-scale segmentation on the remote sensing images of a plurality of cotton identification key time phases, wherein the optimal segmentation scale is the average value of segmentation scale values of each time phase corresponding to the time phase when the spectrum segmentation index of the remote sensing image object of each time phase is the peak value;
4.2 constructing different image information layers LtT is the number of different image periods used;
4.2.1, selecting a remote sensing image of a vegetation vigorous growth period between the time phases of a boll stage and a boll opening stage of cotton; the 1 st layer of the classification layer is marked as L for the remote sensing image construction of the selected vegetation in the vigorous growth period1At the classification level L1In (1), a multi-temporal spectral feature classifier f (L) is established11) Extracting all vegetation land features in the remote sensing image, wherein the vegetation land features comprise garden lands, cotton and interfering crops; at the classification level L1In the method, a multi-temporal spectral feature classifier f (L) is constructed12) Removing all garden lands in the remote sensing image, wherein the vegetation land information left in the remote sensing image comprises cotton and interfering crops;
4.2.2, selecting the remote sensing image of the first gradient cotton identification key time phase, and constructing the 2 nd layer of the classification layer as L2At the classification level L2In (1), a multi-temporal spectral feature classifier f (L) is established2) Eliminating interfering crops which grow vigorously in spring;
4.2.3, selecting the remote sensing image of the second gradient cotton identification key time phase, and constructing the 3 rd layer of the classification layer as L3At the classification level L3In (1), a multi-temporal spectral feature classifier f (L) is established3) Interfering crops such as corn and soybean with bare land characteristics are removed, and the residual vegetation land feature information in the remote sensing image is cotton and peanut;
4.2.4, selecting the remote sensing image of the third gradient cotton identification key time phase, and constructing the 4 th layer L of the classification layer4At the classification level L4In (1), a multi-temporal spectral feature classifier f (L) is established4) Interfering crops peanuts with bare land characteristics are removed, and the residual vegetation land feature information in the remote sensing image is a preliminary information result of cotton.
In the formula, VndviNormalizing index value, V, for vegetationBrightnessIs the band average brightness value, VGLCMContrast(all dir)As texture contrast value, VBlue_MeanIs the average value of blue band, VMax.diffAnd a, b, c, d, e, f, g and h are threshold values.
Step five, performing secondary multi-scale segmentation on the primary information result of the cotton in the remote sensing image obtained in the step four and other interfering crop information which may be mixed into the cotton, constructing a classifier, and finely extracting the cotton information;
5.1, performing secondary multi-scale segmentation on the preliminary information result of the cotton in the remote sensing image obtained in the fourth step and other interfering crop information which may be mixed into the cotton; the interfering crops mixed with cotton are peanuts and soybeans; the remote sensing image based on the secondary multi-scale segmentation is the remote sensing image corresponding to the last gradient cotton identification key time phase;
and (3) selecting optimal scales of secondary multi-scale segmentation: taking a secondary multi-scale segmentation object as a target area, taking the mean variance of the near-infrared wave band of the minimum remote sensing image interfering crops as an evaluation parameter, taking 3 as the lower limit of the segmentation scale, taking 2 as the step length, taking the optimal segmentation scale of the primary segmentation in the fourth step as the upper limit, and selecting the segmentation scale corresponding to the maximum mean variance value as the optimal segmentation scale of the secondary multi-scale segmentation;
5.2, constructing a classifier;
constructing a classifier f (L) by combining a cotton spectral characteristic curve and textural characteristics, removing interfering crops mixed with cotton from the primary cotton information result after secondary multi-scale segmentation as shown in the following formula, extracting cotton subclass 1 information, and respectively extracting missed cotton subclass 2 information and cotton subclass 3 information from peanuts and soybeans of other interfering crops;
wherein, VndviNormalizing index value, V, for vegetationBrightnessIs the band average brightness value, VGLCM Contrast(all dir)As texture contrast value, VMean4The average value of the wave bands of the near infrared wave bands is shown, and k, l, m, n, o and p are threshold values;
step six, merging the extracted various cotton information, and outputting a drawing;
further, in the first step, the cotton spectral characteristic curve construction step is as follows:
1) selecting cotton sample S at random according to Neuman (2011) sampling rulejN, n is the number of selected cotton squares;
2) calculating cotton sample size S in the image band b of the remote sensing image at the cotton growth stage pjThe mean of the DN values or the mean of the spectra,
p ∈ [ unsewn, sown, seedling, bud, boll opening, mature harvest]；
3) For all cotton samples S of each cotton growth stage p and each image band bjMean value of DN values or mean value of spectra
Averaging again to obtain the values of all cotton samples in each cotton growth stage p and each image band bIs marked as
4) Drawing the abscissa as b and the ordinate as
P cotton spectral characteristic curves; plotting the abscissa as p and the ordinate as
B cotton spectral characteristic curves.
The invention has the following beneficial effects:
the invention realizes the remote sensing automatic extraction of the cotton planting information by combining the phenological period analysis of cotton and interfering crops with the layered classification method based on the remote sensing image, and makes the cotton planting information extraction result more accurate and rapid by fully considering the spectral characteristics of cotton in different growing periods and the spectral difference and texture difference with other interfering crops, and is easy to popularize and apply.
Drawings
FIG. 1 is a technical flow chart of the present invention;
FIG. 2 is a constructed hierarchical classification system diagram for remote sensing monitoring of cotton;
FIG. 3 is a phenological analysis table of main crops in Nangong;
FIG. 4 is a schematic diagram of cotton sample plot before and after random sampling in Nangong City;
FIG. 5 is a schematic diagram of spectral characteristics of cotton in Nangong City at different periods;
FIG. 6 is a variation curve of spectral characteristics of cotton in different wavebands in Nangong City;
FIG. 7 is a schematic representation of cotton growth period interpretation (pseudo color synthesis);
FIG. 8 is a schematic diagram of a hierarchical moving window screening method;
FIG. 9 is a remote sensing distribution diagram of cotton in Nangong.
Detailed Description
The cotton remote sensing monitoring method based on the phenological analysis is suitable for shooting multi-temporal remote sensing images by the same satellite, the specific implementation flow of the cotton remote sensing monitoring method is shown in a figure 1, and the cotton remote sensing monitoring method based on the phenological analysis comprises the following steps:
firstly, combining the analysis of the main crop phenology and the cotton spectral characteristic analysis in the area to determine the key time phase of the cotton remote sensing monitoring;
1.1, determining the types and the phenological conditions of main interfering crops in the cotton planting period and making a phenological table.
1.2, drawing a cotton characteristic curve and determining a key time phase of remote sensing monitoring of cotton.
And selecting remote sensing images according to the growth stages p and p belonging to the cotton in the phenological table [ the non-sowing period, the seedling period, the bud period, the boll opening period and the mature harvesting period ], and drawing a cotton spectral characteristic curve. The spectral characteristics of the cotton in the growing period are respectively drawn into two cotton spectral characteristic curves with the abscissa of the growing period and different wave bands by adopting a sample mean value according to the sampling rule of Neuman (2011).
The cotton spectral characteristic curve construction steps are as follows:
1) selecting cotton sample S at random according to Neuman (2011) sampling rulejN, n is the number of cotton squares selected.
2) Calculating cotton sample size S in the image band b of the remote sensing image at the cotton growth stage pjThe mean of the DN values or the mean of the spectra,
p ∈ [ unsewn, sown, seedling, bud, boll opening, mature harvest]。
3) For all cotton samples S of each cotton growth stage p and each image band bjMean value of DN values or mean value of spectra
Taking the mean value again to obtain the values of all cotton samples in each cotton growth stage p and each image wave band b, and recording the values as
4) Drawing the abscissa as b and the ordinate as
P cotton spectral characteristic curves; plotting the abscissa as p and the ordinate as
B cotton spectral characteristic curves.
1.3, establishing a remote sensing image interpretation mark of the cotton growth stage by combining the analysis of the cotton spectral characteristic curves of different growth stages and based on the visual interpretation of the remote sensing image; the specific process is as follows:
the cotton is in a sowing period and a seedling emergence period, and the remote sensing images show bare land information; the cotton is in the bud stage and in the primary stage of growth, vegetation can cover the earth surface, the remote sensing image is pink, and the ground characteristics of the cotton vegetation in the initial stage are presented; the cotton is in the boll stage and the boll opening stage, the growth is vigorous, the remote sensing image is rosy, and the vegetation ground characteristics of the flourishing cotton are presented; when the cotton is in the harvesting period, the remote sensing image is light red, vegetation ground characteristics of withered cotton are presented, and the vegetation ground characteristics are matched with the spectral analysis.
1.4, gradient division;
based on the spectral characteristics of the cotton at different growth stages and the surface characteristics presented by the interpretation marks, the cotton identification key stage is divided into four gradients, and the four gradients are divided and the surface characteristics are as follows:
| gradient partitioning | Phenological stage | Surface features |
| First gradient | Non-sowing, sowing period and seedling period | Characteristics of bare land |
| Second gradient | Bud stage, boll opening stage | Features of vegetation |
| Third gradient | At the mature harvest stage | Features of vegetation |
| Fourth gradient | After the mature harvest period | Characteristics of bare land |
The crop species within each gradient were determined as follows:
the crop species within the first gradient comprise cotton and all interfering crops;
the crop species in the second, third, and fourth gradients do not contain interfering crops that are opposite to the cotton surface characteristics at the time of the cotton identification key in the previous gradient.
And 1.5, selecting a key time phase for cotton identification in each gradient range by adopting a moving window method.
The specific process is as follows:
and in each gradient, moving a window in the phenological table by taking ten days as a step length, and determining a time phase when the number of the interfering crop species opposite to the surface characteristics of the cotton surface is maximum as a key time phase for identifying the cotton in the gradient.
And 1.6, determining the cotton remote sensing monitoring key time phase in each gradient in the cotton identification key time phase in each gradient by combining the availability of remote sensing data.
And step two, integrating the influence of the forest land on crop extraction, increasing the image of the vegetation growth period in the first period, improving the cotton extraction precision, preferably selecting the remote sensing image of the vegetation growth period between the time phase of the boll period and the time phase of the boll opening period of the cotton, and preferably selecting the remote sensing image of the vegetation growth period between the time phase of the boll period and the time phase of the boll opening period of the cotton.
Acquiring a key time phase remote sensing image, and preprocessing the key time phase remote sensing image;
and taking the remote sensing image of one time phase as a reference, and carrying out image registration on the remote sensing images of other time phases.
Step four, performing multi-scale segmentation on the multi-temporal image, constructing classifiers in a layering manner, and preliminarily extracting cotton;
the specific process is as follows:
and 4.1, performing multi-scale segmentation on the remote sensing images of a plurality of cotton identification key time phases, wherein the optimal segmentation scale is the average value of segmentation scale values of each time phase corresponding to the time phase when the spectrum segmentation index of the remote sensing image object of each time phase is the peak value.
4.2 constructing different image information layers LtT is the number of different used image periods, and a layered classification system for remote sensing monitoring of cotton is shown in FIG. 2;
4.2.1, constructing the layer 1 of the classification layer as L based on the remote sensing image of the vegetation in the vigorous growth period1At the classification level L1In (1), a multi-temporal spectral feature classifier f (L) is established11) Extracting all vegetation land features in the remote sensing image, wherein the vegetation land features comprise garden lands and crops, and the crops comprise cotton and interfering crops; at the classification level L1In the method, a multi-temporal spectral feature classifier f (L) is constructed12) And removing all garden lands in the remote sensing image, wherein the vegetation land feature information left in the remote sensing image is crops, namely cotton and interfering crops.
4.2.2, selecting the remote sensing image of the first gradient cotton identification key time phase, and constructing the 2 nd layer of the classification layer as L2At the classification level L2In (1), a multi-temporal spectral feature classifier f (L) is established2) And removing interfering crops with vigorous growth in spring, wherein the interfering crops with vigorous growth in spring, such as winter wheat in southern city of Hebei province, and the vegetation land feature information left in the remote sensing image comprises cotton, corn and soybean.
4.2.3, selecting the remote sensing image of the second gradient cotton identification key time phase, and constructing the 3 rd layer of the classification layer as L3At the classification level L3In (1), a multi-temporal spectral feature classifier f (L) is established3) Interfering crops such as corn and soybean with bare land characteristics are removed, and the residual vegetation land and feature information in the remote sensing image is cotton and peanut.
4.2.4, selecting the remote sensing image of the third gradient cotton identification key time phase, and constructing the 4 th layer L of the classification layer4At the classification level L4In (1), a multi-temporal spectral feature classifier f (L) is established4) Interfering crops peanuts with bare land characteristics are removed, and the residual vegetation land feature information in the remote sensing image is a preliminary information result of cotton.
In the formula, VndviNormalizing index value, V, for vegetationBrightnessIs the band average brightness value, VGLCMContrast(all dir)As texture contrast value, VBlue_MeanIs the average value of blue band, VMax.diffAnd a, b, c, d, e, f, g and h are threshold values.
And step five, performing secondary multi-scale segmentation on the primary information result of the cotton in the remote sensing image obtained in the step three and other interfering crop information possibly mixed into the cotton, constructing a classifier, and finely extracting the cotton information.
And 5.1, performing secondary multi-scale segmentation on the primary information result of the cotton in the remote sensing image obtained in the fourth step and other interfering crop information which may be mixed into the cotton.
The interfering crops mixed with cotton are peanuts and soybeans; and the remote sensing image based on the secondary multi-scale segmentation is the remote sensing image corresponding to the last gradient cotton identification key time phase.
And (3) selecting optimal scales of secondary multi-scale segmentation: and taking the secondary multi-scale segmentation object as a target area, taking the minimum remote sensing image near-infrared band mean variance interfering crops as an evaluation parameter, taking 3 as a lower segmentation scale limit, taking 2 as a step length, taking the optimal segmentation scale of the primary segmentation in the third step as an upper limit, and selecting the segmentation scale corresponding to the maximum mean variance value as the optimal segmentation scale of the secondary multi-scale segmentation.
And 5.2, constructing a classifier. And (3) constructing a classifier f (L) by combining the cotton spectral characteristic curve and the cotton texture characteristic, removing the interfering crops mixed with the cotton from the cotton primary information result obtained in the step four as shown in the following formula, extracting the information of the cotton subclass 1, and respectively extracting the information of the missed cotton subclass 2 and the information of the cotton subclass 3 from the peanuts and soybeans of other interfering crops.
Wherein, VndviNormalizing index value, V, for vegetationBrightnessIs the band average brightness value, VGLCM Contrast(all dir)As texture contrast value, VMean4Is the average value of the wave band of the near infrared wave band, and k, l, m, n, o and p are threshold values.
And step six, merging the extracted various cotton information, and outputting a drawing.
The present invention will be described in further detail with reference to the drawings and the detailed description thereof, taking the southern city of Hebei province as an example.
The cotton remote sensing monitoring method based on the phenological analysis is suitable for shooting multi-temporal remote sensing images by the same satellite, the specific implementation flow of the cotton remote sensing monitoring method is shown in a figure 1, and the cotton remote sensing monitoring method based on the phenological analysis comprises the following steps:
combining regional cotton spectral characteristic analysis and main crop phenological analysis to determine a key time phase of cotton remote sensing monitoring;
1.1, determining the type and the phenological climate of the main interfering crops in the cotton planting period, and making a phenological table;
the interfering crop types in south China mainly comprise winter wheat, corn, peanut and soybean, and the phenological analysis table is shown in figure 3.
1.2, combining a phenological table, drawing a cotton spectral characteristic curve, and performing characteristic analysis on the cotton spectral characteristic;
in this embodiment, there are 292 cotton plots in the city of the south palace, as shown in fig. 4a, based on the Neuman (2011) sampling rule, the sampling ratio is 0.3, i.e., a cotton plot 292 × 0.3 ≈ 88 is selected, as shown in fig. 4b, and the distribution of the existing and selected cotton plots is shown in fig. 4; according to a cotton phenological analysis table, selecting the time phases of different growth periods of cotton samples, namely, a first-higher multispectral image with the 6 periods of 4 months 21 days, 5 months 16 days, 6 months 22 days, 8 months 27 days, 9 months 8 days and 10 months 14 days, and calculating the average value of DN values of 4 wave bands of the area image corresponding to each cotton sample:
taking the mean value of DN average values of all the cotton samples in each growing period again to obtain the value of each band of the cotton samples in each growing period,
the abscissa is drawn as different wave bands and the ordinate is
The 6 cotton characteristic curves of (a), as shown in fig. 5; the abscissa is drawn as different time phases and the ordinate is
4 cotton characteristic curves; as shown in fig. 6.
The spectral characteristics of the sowing time of 21 days in 4 months and the emergence time of 16 days in 5 months accord with the spectral characteristics of bare land; the 6-month 22-day bud period belongs to the vegetation growth initial stage and accords with the spectral characteristics of basic vegetation; the transition stage of the flowering period and the boll opening period of 27 days in 8 months and the boll opening period of 8 days in 9 months belong to the vegetation growth vigorous stage, and have typical vegetation spectral characteristics; cotton begins to wither in the mature harvest period of 10 months and 14 days, vegetation characteristics gradually weaken, but basic vegetation spectral characteristics are still provided;
1.3, establishing an interpretation mark of the cotton growth stage by combining spectral curve characteristic analysis of different growth stages of cotton and based on remote sensing image visual interpretation, wherein a schematic diagram of the interpretation mark of the cotton growth stage is given in fig. 7, (a) cotton in the sowing stage at 21 days in 4 months, (b) cotton in the seedling stage at 16 days in 5 months, and the images show bare land information; (c) cotton in 6 months and 22 days is in a bud stage and is in a primary stage of growth, vegetation can cover the ground surface and is pink, and images show the ground characteristics of the cotton vegetation in the initial stage; (d) the cotton in 8 months and 27 days is the transition stage of the boll stage and the boll opening stage, the growth is vigorous, the color is rosy, and the image shows the flourishing vegetation ground characteristics of the cotton; (e) and (4) 10 months and 14 days, wherein the cotton is in the harvest stage, but the leaves and the like are dark green, the images are light red, vegetation ground characteristics of withered cotton are shown, and the characteristics are matched with the spectral analysis.
1.4, gradient division and gradient internal cotton identification key time phase selection;
a first gradient: the types of crops needing to be considered in time phase screening are as follows: cotton, winter wheat, corn, peanut, soybean. The gradient time phase coverage range is 4 middle-5 late months, the number of the types of the interfering crops opposite to the surface characteristics of the cotton is at most 1, the key time phase values of cotton identification are respectively 4 middle and late months, and 5 upper, middle and lower months, as shown in fig. 8a, the winter wheat is opposite to the expression characteristics of the cotton corresponding to the time phases, so the winter wheat is removed, and the types of the second gradient crops are corn, cotton, soybean and peanut.
A second gradient: in the bud period, the boll period and the boll opening period, the coverage range of the gradient time phase is from 6 last days of month to 9 last days of month, the number of the types of the interfering crops opposite to the surface characteristics of the cotton is at most 2, and the value of the key time phase for identifying the cotton is from 6 middle days of month to 6 last days of month, as shown in fig. 8 b. The corresponding ones at this time are corn and soybean, as opposed to cotton performance characteristics, thus eliminating corn and soybean, and the third gradient crop type is cotton, peanut.
A third gradient: in the mature harvest period, the coverage range of the gradient time phase is from 10 to 11 days, the number of the types of the interfering crops opposite to the surface characteristics of the cotton is at most 1, and the values of the key time phases of cotton identification are respectively 10 days, middle days, late days and 11 days, as shown in fig. 8 c. The corresponding crop type at this time is peanut, as opposed to cotton, so the fourth gradient crop type is cotton.
Therefore, by combining the cotton spectral characteristic analysis and the regional phenological analysis and combining the availability of remote sensing data, the finally determined key time phases of the cotton remote sensing monitoring are 4 months, 21 days, 6 months, 22 days and 10 months, 14 days.
And secondly, selecting a remote sensing image of a period with vigorous vegetation growth in the first stage between the time phases of the boll stage and the boll opening stage of the cotton. And in combination with the availability of remote sensing data, the key time phase of the added first-stage remote sensing monitoring is 8 months and 27 days.
Acquiring a key time phase remote sensing image and preprocessing the key time phase remote sensing image;
in the embodiment, multispectral images of high score one in the city of south uterus of 4 months 21 days, 6 months 22 days, 8 months 27 days and 10 months 14 days are obtained, image data are preprocessed, the multispectral image wave bands are four wave bands of near infrared, red, green and blue, the Landsat/TM30 meter orthographic image is used as a reference image for the image, geometric fine correction is carried out, and the correction precision is 0.5-1 pixel in a plain area; and then, taking the 4-month 21-day image as a reference, and carrying out image registration on other 3 time phase images to ensure that the 4-phase image has no dislocation.
Step four, performing multi-scale segmentation on the multi-temporal image, constructing classifiers in a layering manner, and preliminarily extracting cotton;
multi-scale segmentation and optimal segmentation ruler for multi-temporal imagesThe degree is the average value of all stages of segmentation scale values corresponding to the peak value of the spectrum segmentation index of the image object in each stage, and the optimal segmentation scale of the multi-scale segmentation of the image in the 4 stages in south China city is 15 through calculation. L is18 months and 27 days of image, L24 months and 21 days of image, L36 months and 22 days of image, L 410 months and 14 days. The classifier for each layer is specifically as follows:
wherein, VndviNormalizing index value, V, for vegetationBrightnessIs the band average brightness value, VGLCMContrast(all dir)As texture contrast value, VBlue_MeanIs the average value of blue band, VMax.diffThe maximum heterogeneity index value.
Step five, performing secondary segmentation on the primary cotton result and the possible mixed cotton objects, constructing a classifier by combining spectral features and textural features of cotton, removing mixed cotton, and supplementing missing cotton information;
(1) and selecting a secondary segmentation object. The secondary segmentation object selects the primary extraction result of cotton in Nangong city and interference crops peanut and soybean which are easily mixed with the cotton in other crops in Nangong city, the total number of the interference crops is 3, the image data selects the image of the period with the least interference crops when the cotton is extracted, and the image of the first high-grade 16 m 10 months and 14 days is selected.
(2) And selecting optimal scales for quadratic segmentation. The selected primarily extracted cotton, peanut and soybean are taken as targets, the near-infrared band mean variance of 14-day images in 10 months is taken as a secondary segmentation scale evaluation parameter, 3 is taken as a segmentation scale lower limit, 2 is taken as a step length, the primary segmentation optimal scale 15 is taken as an upper limit, and the secondary segmentation optimal scale is 5.
(3) And constructing a classifier. Combining the cotton spectral characteristic curve in the city of southern palace, as shown in fig. 6, and cotton texture characteristics to construct a classifier for fine extraction of cotton, extracting cotton subclasses 1 from the primarily extracted cotton, cotton subclasses 2 from the peanuts, and cotton subclasses 3 from the peanuts.
Wherein, VndviNormalizing index value, V, for vegetationBrightnessIs the band average brightness value, VGLCM Contrast(all dir)As texture contrast value, VMean4The average value of the near infrared wave band is shown.
Step six, merging and outputting various types of cotton information extracted in a layered mode to be mapped
And (4) drawing, namely determining a scale by using ArcGIS software, adding a legend and a north pointer, and completing cotton distribution drawing, as shown in fig. 9.
In order to effectively verify and evaluate the method, the extracted cotton information in Nangong City is subjected to precision verification by using cotton sample data.
The attribute precision calculation method for cotton remote sensing monitoring comprises the following steps: firstly, converting the sample data for verification from a vector (. shp format) into raster data, wherein the spatial resolution of the raster data is the same as that of the spatial distribution data of main crops; then converting the grid sample data into point data, wherein the position of the point is the central position of each pixel; and then, performing superposition analysis on the point data and the extracted cotton information vector to calculate the attribute precision of the cotton remote sensing monitoring, wherein the calculation formula is as follows:
in the formula, SX _ accuracycacy is attribute accuracy, NoverlayNumber of points, N, where cotton sample squares intersect with cotton remote sensing monitoring vectorssampleThe number of the transformed points is the cotton sample. Through calculation, the data of the southern palace market are converted into 4448 points, and the points which all fall on the cotton remote sensing monitoring vector are 4194, so that the attribute precision is 94.3%.
The area precision of the cotton remote sensing monitoring is calculated as follows: superposing the verification sample square vector and the cotton remote sensing monitoring vector, counting the crossed area of the verification sample square vector and the cotton remote sensing monitoring vector, wherein the ratio of the area to the total area of the cotton sample square is the area precision, and the calculation formula is as follows:
wherein MJ _ accuracy is area accuracy, AoverlapIs the area of the intersection region of the cotton sample square vector and the cotton remote sensing monitoring vector, AsampleThe area of the cotton sample space vector is calculated to be 1078165.5m2The total area of the cotton sample prescription in Nangong City is 1138033.8m2Therefore, the area accuracy is 94.7%.
The attribute precision is 94.3 percent, and the area precision is 94.7 percent, thereby meeting the requirements of practical application.
The embodiments of the present invention have been described in detail with reference to the accompanying drawings, but the present invention is not limited to the above embodiments, and various changes can be made thereto within the knowledge of those skilled in the art.
Claims (2)
1. A cotton remote sensing monitoring method based on phenological analysis is characterized by comprising the following steps:
firstly, combining the analysis of the main crop phenology and the cotton spectral characteristic analysis in the area to determine the key time phase of the cotton remote sensing monitoring;
1.1, determining the types and the phenological conditions of main interfering crops in the cotton planting period, and making a phenological table, wherein the main interfering crops comprise winter wheat, corn, peanut and soybean;
1.2, drawing a cotton spectral characteristic curve;
1.3, establishing a remote sensing image interpretation mark of the cotton growth stage by combining the analysis of the cotton spectral characteristic curves of different growth stages and based on the visual interpretation of the remote sensing image;
the specific process is as follows:
the cotton is in a sowing period and a seedling emergence period, and the remote sensing images show bare land information; when the cotton is in the bud stage, the remote sensing image is pink, and the ground characteristics of the cotton at the initial vegetation stage are presented; the cotton is in the boll stage and boll opening stage, the remote sensing image is rosy, and the vegetation ground characteristic of flourishing cotton is presented; when the cotton is in the harvesting period, the remote sensing image is light red, and vegetation ground characteristics of withered cotton are presented;
1.4, gradient division;
dividing a key cotton identification period into four gradients, wherein the first gradient is an ungrooved period, a sowing period and a seedling period, and the surface characteristic is a bare land characteristic; the second gradient is a bud period, a flower bell period and a boll opening period, and the ground surface characteristic is a vegetation characteristic; the third gradient is a mature harvest period, and the ground surface characteristic is a vegetation characteristic; the fourth gradient is that after the mature harvest period, the surface characteristic is a bare land characteristic;
1.5, selecting a key time phase for cotton identification in each gradient range by adopting a moving window method;
the specific process is as follows:
in each gradient, moving a window in a phenological table by taking ten days as a step length, and determining a time phase when the number of the interfering crop species opposite to the surface characteristics of the cotton surface is maximum as a key time phase for identifying the cotton in the gradient;
1.6, determining a cotton remote sensing monitoring key time phase in each gradient cotton identification key time phase by combining the availability of remote sensing data;
selecting a remote sensing image of a period with vigorous vegetation growth in the first stage between the time phases of a boll stage and a boll opening stage of cotton;
acquiring a key time phase remote sensing image, and preprocessing the key time phase remote sensing image;
taking the remote sensing image of one time phase as a reference, and carrying out image registration on the remote sensing images of other time phases;
step four, carrying out multi-scale segmentation on the multi-temporal remote sensing image, constructing classifiers in a layering manner, and preliminarily extracting cotton;
the specific process is as follows:
4.1, performing multi-scale segmentation on the remote sensing images of a plurality of cotton identification key time phases, wherein the optimal segmentation scale is the average value of segmentation scale values of each time phase corresponding to the time phase when the spectrum segmentation index of the remote sensing image object of each time phase is the peak value;
4.2 constructing different image information layers LtT is the number of different image periods used;
4.2.1, constructing the layer 1 of the classification layer of the remote sensing image of the selected vegetation in the vigorous growth period as L1At the classification level L1In (1), a multi-temporal spectral feature classifier f (L) is established11) Extracting all vegetation land features in the remote sensing image, wherein the vegetation land features comprise garden lands, cotton and interfering crops; at the classification level L1In the method, a multi-temporal spectral feature classifier f (L) is constructed12) Removing all garden lands in the remote sensing image, wherein the vegetation land information left in the remote sensing image comprises cotton and interfering crops;
4.2.2, selecting the remote sensing image of the first gradient cotton identification key time phase, and constructing the 2 nd layer of the classification layer as L2At the classification level L2In (1), a multi-temporal spectral feature classifier f (L) is established2) Eliminating interfering crops which grow vigorously in spring;
4.2.3, selecting the remote sensing image of the second gradient cotton identification key time phase, and constructing the 3 rd layer of the classification layer as L3At the classification level L3In (1), a multi-temporal spectral feature classifier f (L) is established3) Interfering crops such as corn and soybean with bare land characteristics are removed, and the residual vegetation land feature information in the remote sensing image is cotton and peanut;
4.2.4, selecting the remote sensing image of the third gradient cotton identification key time phase, and constructing the 4 th layer L of the classification layer4At the classification level L4In (1), a multi-temporal spectral feature classifier f (L) is established4) Interfering crop peanuts with bare land characteristics are removed, and the residual vegetation land feature information in the remote sensing image is a preliminary information result of cotton;
in the formula, VndviNormalizing index value, V, for vegetationBrightnessIs the band average brightness value, VGLCMContrast(all dir)As texture contrast value, VBlue_MeanIs the average value of blue band, VMax.diffThe value of the maximum heterogeneity index, a, b, c, d, e, f, g and h are thresholds;
step five, performing secondary multi-scale segmentation on the primary information result of the cotton in the remote sensing image obtained in the step four and other interfering crop information which may be mixed into the cotton, constructing a classifier, and finely extracting the cotton information;
5.1, performing secondary multi-scale segmentation on the preliminary information result of the cotton in the remote sensing image obtained in the fourth step and other interfering crop information which may be mixed into the cotton; the interfering crops mixed with cotton are peanuts and soybeans; the remote sensing image based on the secondary multi-scale segmentation is the remote sensing image corresponding to the last gradient cotton identification key time phase;
and (3) selecting optimal scales of secondary multi-scale segmentation: taking a secondary multi-scale segmentation object as a target area, taking the mean variance of the near-infrared wave band of the minimum remote sensing image interfering crops as an evaluation parameter, taking 3 as the lower limit of the segmentation scale, taking 2 as the step length, taking the optimal segmentation scale of the primary segmentation in the fourth step as the upper limit, and selecting the segmentation scale corresponding to the maximum mean variance value as the optimal segmentation scale of the secondary multi-scale segmentation;
5.2, constructing a classifier;
constructing a classifier f (L) by combining a cotton spectral characteristic curve and textural characteristics, removing interfering crops mixed with cotton from the primary cotton information result after secondary multi-scale segmentation as shown in the following formula, extracting cotton subclass 1 information, and respectively extracting missed cotton subclass 2 information and cotton subclass 3 information from peanuts and soybeans of other interfering crops;
wherein, VndviNormalizing index value, V, for vegetationBrightnessIs the band average brightness value, VGLCM Contrast(all dir)As texture contrast value, VMean4The average value of the wave bands of the near infrared wave bands is shown, and k, l, m, n, o and p are threshold values;
and step six, merging the extracted various cotton information, and outputting a drawing.
2. The method for remotely sensing and monitoring cotton based on phenological analysis as claimed in claim 1, characterized in that in said first step, the cotton spectral characteristic curve construction step is as follows:
1) the cotton sample is randomly selected according to Neuman, 2011, sampling rulejJ is 1,2,3, … … n, n is the number of selected cotton squares;
2) calculating cotton sample size S in the cotton growth stage p and the image band b of the remote sensing imagejThe mean of the DN values or the mean of the spectra,
p ∈ [ unsewn, sown, seedling, bud, boll opening, mature harvest]；
3) For all cotton samples S of each cotton growth stage p and each image band bjMean value of DN values or mean value of spectra
Taking the mean value again to obtain the values of all cotton samples in each cotton growth stage p and each image wave band b, and recording the values as
4) Drawing the abscissa as b and the ordinate as
P cotton spectral characteristic curves; plotting the abscissa as p and the ordinate as
B cotton spectral characteristic curves.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711500930.3A CN108363949B (en) | 2017-12-27 | 2017-12-27 | Cotton remote sensing monitoring method based on phenological analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711500930.3A CN108363949B (en) | 2017-12-27 | 2017-12-27 | Cotton remote sensing monitoring method based on phenological analysis |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108363949A CN108363949A (en) | 2018-08-03 |
| CN108363949B true CN108363949B (en) | 2020-08-28 |
Family
ID=63010598
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711500930.3A Active CN108363949B (en) | 2017-12-27 | 2017-12-27 | Cotton remote sensing monitoring method based on phenological analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108363949B (en) |
Families Citing this family (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109409261B (en) * | 2018-10-11 | 2020-09-15 | 北京师范大学 | Crop classification method and system |
| CN109614891A (en) * | 2018-11-27 | 2019-04-12 | 北京师范大学 | Crops recognition methods based on phenology and remote sensing |
| CN109829464B (en) * | 2018-12-24 | 2021-01-05 | 核工业北京地质研究院 | Method for screening red Fuji apples by using spectral data |
| CN110598514B (en) * | 2019-06-10 | 2023-03-24 | 南京大学 | Method for monitoring plot scale crop seeding area of land reclamation project area |
| CN110717413B (en) * | 2019-09-24 | 2022-11-25 | 云南师范大学 | Unmanned aerial vehicle visible light image-oriented water body information extraction method and device |
| 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 |
| CN111345214A (en) * | 2020-03-13 | 2020-06-30 | 武汉禾大科技有限公司 | Xinjiang cotton region identification method and system based on satellite image data |
| CN112800973B (en) * | 2021-01-29 | 2021-08-27 | 宁波大学 | Spartina alterniflora extraction method based on vegetation phenological feature decision |
| CN113743819B (en) * | 2021-09-15 | 2024-03-26 | 二十一世纪空间技术应用股份有限公司 | Crop yield estimation method, device, electronic equipment and storage medium |
| CN114782832B (en) * | 2022-06-24 | 2022-11-04 | 中国农业科学院农业资源与农业区划研究所 | Crop extraction method and device, readable storage medium and electronic equipment |
| CN116563721A (en) * | 2023-07-12 | 2023-08-08 | 浙江领见数智科技有限公司 | Tobacco field extraction method based on layered classification thought |
| CN117788200A (en) * | 2024-02-28 | 2024-03-29 | 杨凌职业技术学院 | Agricultural product maturity prediction system based on multisource remote sensing data |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8189877B2 (en) * | 2005-10-21 | 2012-05-29 | Carnegie Institution Of Washington | Remote sensing analysis of forest disturbances |
| CN103761717A (en) * | 2014-01-26 | 2014-04-30 | 河海大学 | City water extraction method based on panchromatic remote sensing image |
| CN105447494A (en) * | 2015-12-01 | 2016-03-30 | 二十一世纪空间技术应用股份有限公司 | Vegetable field monitoring method based on multi-source multi-temporal remote sensing image data |
| CN106846325A (en) * | 2017-01-16 | 2017-06-13 | 河南理工大学 | The automatic mode that a kind of remote sensing image optimum segmentation result determines |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9589210B1 (en) * | 2015-08-26 | 2017-03-07 | Digitalglobe, Inc. | Broad area geospatial object detection using autogenerated deep learning models |
-
2017
- 2017-12-27 CN CN201711500930.3A patent/CN108363949B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8189877B2 (en) * | 2005-10-21 | 2012-05-29 | Carnegie Institution Of Washington | Remote sensing analysis of forest disturbances |
| CN103761717A (en) * | 2014-01-26 | 2014-04-30 | 河海大学 | City water extraction method based on panchromatic remote sensing image |
| CN105447494A (en) * | 2015-12-01 | 2016-03-30 | 二十一世纪空间技术应用股份有限公司 | Vegetable field monitoring method based on multi-source multi-temporal remote sensing image data |
| CN106846325A (en) * | 2017-01-16 | 2017-06-13 | 河南理工大学 | The automatic mode that a kind of remote sensing image optimum segmentation result determines |
Non-Patent Citations (2)
| Title |
|---|
| Forest change detection by statistical object-based method;Baudouin Desclee;《Remote Sensing of Environment》;20060530;第102卷(第1-2期);1-11 * |
| 基于遥感图像分析的北京植被状态与变化研究;王佃来;《中国博士学位论文全文数据库农业科技辑》;20140515(第05期);D050-10 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108363949A (en) | 2018-08-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108363949B (en) | Cotton remote sensing monitoring method based on phenological analysis | |
| CN109345555B (en) | Method for identifying rice based on multi-temporal multi-source remote sensing data | |
| George et al. | Forest tree species discrimination in western Himalaya using EO-1 Hyperion | |
| CN105740759B (en) | Semilate rice information decision tree classification approach based on feature extraction in multi-temporal data | |
| CN107527014A (en) | Crops planting area RS statistics scheme of sample survey design method at county level | |
| CN110472184A (en) | A kind of cloudy misty rain area rice recognition methods based on Landsat remotely-sensed data | |
| CN112800973B (en) | Spartina alterniflora extraction method based on vegetation phenological feature decision | |
| CN108195767B (en) | Estuary wetland foreign species monitoring method | |
| GB2598012A (en) | System and method for crop monitoring | |
| CN111209871B (en) | Rape planting land remote sensing automatic identification method based on optical satellite image | |
| CN114387528A (en) | Pine nematode disease monitoring space-air-ground integrated monitoring method | |
| CN113221806B (en) | Cloud platform fusion multi-source satellite image and tea tree phenological period based automatic tea garden identification method | |
| CN105404873B (en) | A kind of winter wheat recognition methods based on the conversion of NDVI time series coordinates | |
| CN104951772B (en) | A kind of winter wheat extracting method based on NDVI time-serial positions integration | |
| CN109977991A (en) | Forest resourceies acquisition method based on high definition satellite remote sensing | |
| CN108710864A (en) | Winter wheat Remotely sensed acquisition method based on various dimensions identification and image noise reduction processing | |
| CN116188793A (en) | Astragalus sinicus planting area monitoring method based on satellite remote sensing image | |
| CN112329733B (en) | Winter wheat growth monitoring and analyzing method based on GEE cloud platform | |
| CN115641504A (en) | Automatic remote sensing extraction method for field boundary based on crop phenological characteristics and decision tree model | |
| CN113554675A (en) | Edible fungus yield estimation method based on unmanned aerial vehicle visible light remote sensing | |
| Deng et al. | Tree crown recognition algorithm on high spatial resolution remote sensing imagery | |
| Luo et al. | Staple crop mapping with Chinese GaoFen-1 and GaoFen-6 satellite images: A case study in Yanshou County, Heilongjiang Province, China | |
| ZHANG et al. | Object-oriented crop classification based on UAV remote sensing imagery | |
| Feng et al. | Estimating paddy rice area in Southren China with multi-temporal MODIS data | |
| CN117347974B (en) | Method and device for estimating aboveground biomass of bush vegetation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| EE01 | Entry into force of recordation of patent licensing contract | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20180803 Assignee: Zhongguancun Technology Leasing Co.,Ltd. Assignor: TWENTY FIRST CENTURY AEROSPACE TECHNOLOGY Co.,Ltd. Contract record no.: X2023980040209 Denomination of invention: A Remote Sensing Monitoring Method for Cotton Based on Phenological Analysis Granted publication date: 20200828 License type: Exclusive License Record date: 20230821 |
|
| PE01 | Entry into force of the registration of the contract for pledge of patent right | ||
| PE01 | Entry into force of the registration of the contract for pledge of patent right |
Denomination of invention: A Remote Sensing Monitoring Method for Cotton Based on Phenological Analysis Effective date of registration: 20230822 Granted publication date: 20200828 Pledgee: Zhongguancun Technology Leasing Co.,Ltd. Pledgor: TWENTY FIRST CENTURY AEROSPACE TECHNOLOGY Co.,Ltd. Registration number: Y2023980053341 |