CN110598514A - Method for monitoring plot scale crop seeding area of land reclamation project area - Google Patents

Abstract

The invention relates to a method for monitoring the planting area of crops at a plot scale in a land remediation project area, and belongs to the technical field of land monitoring. This method uses high temporal resolution images and high spatial resolution images, combined with vector land use data, to obtain high temporal and spatial resolution NDVI of agricultural land plots in the land consolidation project area; and introduces Google Earth's RGB image spectral features and texture features to extract The characteristic space of crop planting in paddy field, dry land and irrigated land; secondly, the quadratic difference algorithm and the method of plot statistics are used to realize the extraction of the multi-cropping level at the plot scale in the land consolidation project area; finally, the plot in the land consolidation project area is realized according to the spatial statistical method. Extraction of scale crop sown area and its variation characteristics. The invention not only has low cost and convenient processing, but also is not easily affected by subjective factors, and the results are objective and reliable.

Description

A method for monitoring the sown area of crops at the plot scale in the land consolidation project area

technical field

The invention relates to a method for monitoring the planting area of crops on a plot scale in a land remediation project area, and belongs to the technical field of land monitoring.

Background technique

Land consolidation, firstly, by flattening the uneven arable land, so that the arable land is concentrated and contiguous, providing convenience for all kinds of agricultural machinery for planting activities such as transplanting, harvesting, using pesticides and applying chemical fertilizers; secondly, by repairing dilapidated ditches Or build new irrigation and drainage facilities, etc., to provide support for the realization of drought and flood protection; the third is to build field roads to facilitate the entry and exit of agricultural machinery, and farmers' daily care and management; the fourth is to build a field protection network, and take relevant measures to prevent landslides and wind damage, etc. . The important goal of land consolidation is to improve production conditions such as water distribution in the field, and to realize the convenience of management, thereby changing the quality of agricultural land, the level of multi-cropping (the number of crops planted on a certain agricultural land in a year), and the sown area of crops, etc., so as to improve agricultural use. productivity and agricultural income levels.

Land consolidation is an important measure to improve the production capacity and production stability of agricultural land (land used to grow crops), and an important means to build high-standard farmland. In recent years, it has become a key content of national strategic deployment. However, it is necessary to quantitatively identify how agricultural land production has changed in the land consolidation project area.

The sown area of crops is one of the important contents to be investigated for changes in agricultural production in the land consolidation project area. A plot of agricultural land is a small area of agricultural land formed by dividing agricultural land by other land or linear features in a certain area. Production conditions such as water and soil are relatively close. Plot-scale crop sown area refers to the total area of crops planted in each agricultural land plot within the year. Finding out the plot-scale crop sown area in the land remediation project area is important for disaster response, fine management and protection of agricultural land, zoning utilization and improvement. very necessary.

At present, at the practical level, the sown area of crops on agricultural land is mainly carried out by sampling survey method; while at the research level, the sown area of a specific crop is mainly extracted or extracted by remote sensing images with low spatial resolution. However, the method of extracting the sown area for specific crops is difficult to effectively monitor the sown area at the plot scale, because 1) for different pixels in the same plot in the first half (or second half) When the planted crop a is in the period of turning green and jointing, while the crop b planted by another pixel is in the period of sowing and germination, that is, it is not in the same growth period. In area extraction, a crop will obviously be missed, and the selection of the phase of the image also depends on the type of crop, but the type of crop is determined according to the farming habits, crop rotation, fallow, etc. may occur in the land consolidation project area. This makes it difficult to conduct annual dynamic monitoring. 2) For the same pixel, if different crops are planted in the first half and second half of the year, it is necessary to investigate the type of crops planted and prior knowledge to select the phase every year, which will cause monitoring difficulties; A detailed grasp of the phenological period and crop growth characteristics can only enable monitoring of that specific crop. Therefore, for the same plot, it is difficult to achieve effective monitoring of the sown area of crops in the plot by selecting several time-phase monitoring specific crop sown areas to superimpose.

In addition, the remote sensing extraction method with low spatial resolution cannot achieve the monitoring at the plot scale. The reason is that when a pixel covers the entire plot, the interior of the plot cannot be described in detail. When planting different crops, the value of the pixel is the result of the synthesis of each crop. If the proportion of various crops planted in different areas in each block is inconsistent, it is difficult to set the peak true value judgment rule, and if there are crops planted in the block all the time It is difficult to effectively extract the sown area. For example, months a-d in half of the area are wheat development periods, while months c-e in half areas are rice development periods, so it is difficult for months a-e to form the required "increase-decrease" inverted V at the plot scale. type crop growth curve; if one pixel covers multiple plots, it is even more impossible to separate the sown area among the plots.

To sum up, the existing technology has the following limitations: 1) The sampling survey method cannot reflect the characteristics of global spatial heterogeneity, cannot obtain historical information, the sample selection is easily affected by subjective factors, and the time, human and financial costs are high. 2) The method of extracting the sown area of a specific crop is difficult to obtain the total area of all crops in the year, which is especially prominent when multiple crops are intercropped and the growing seasons overlap in certain periods. 3) The method of extracting crop sown area using remote sensing images with low spatial resolution cannot be applied to the plot scale, because most agricultural land plots in my country are relatively small, especially in mountainous, hilly and other areas, where low spatial resolution is used. Remote sensing images will make it difficult to segment each block, and the probability of misjudgment is extremely high; and if high-precision images are used to extract crop sown area, there are problems that are difficult to be widely used due to high price, large amount of data, and very difficult processing. Such images are provided for recently launched satellites, but they are insufficient in long-term dynamic monitoring, especially in historical information acquisition.

SUMMARY OF THE INVENTION

The technical problem to be solved by the present invention is to provide a method for monitoring the sown area of crops at a plot scale in a land remediation project area, which is low in cost, convenient in handling, and not easily affected by subjective factors.

In order to solve the above-mentioned technical problems, the technical solution proposed by the present invention is: a method for monitoring the sown area of crops on a plot scale in a land remediation project area, comprising the following steps:

Step 1: Obtain high temporal resolution images, high spatial resolution images and vector land use data of the year to be monitored covering the land remediation project area, where the vector land use data includes the boundaries of arable land and land use types;

High temporal resolution NDVI and high spatial resolution NDVI are obtained from high temporal resolution images and high spatial resolution images respectively, and high spatial and temporal resolution NDVI is obtained by fusing high temporal resolution NDVI and high spatial resolution NDVI by ESTARFM method. ;

Step 2: Using the cultivated land plot boundary as a constraint condition, segment the high temporal and spatial resolution NDVI to obtain a high temporal and spatial resolution image NDVI of the agricultural land in the land consolidation project area;

According to the boundary of cultivated land, combined with the vegetation planting areas identified by Google Earth high-score images based on spectral and texture features, the typical crop planting areas in the land remediation project area were obtained;

According to the land use type, the typical crop planting area is divided into three types: paddy field, dry land and irrigated land, and the high temporal and spatial resolution NDVI time series features of these three types of crop planting typical areas are extracted respectively. The time series features include the shape of the curve formed by the high temporal and spatial resolution NDVI time series data, the peak value of the curve, and the time distance between the peaks;

Step 3: According to the high temporal and spatial resolution NDVI time series characteristics of the typical areas of three types of crop planting, combined with the quadratic difference algorithm to extract the peak value of each pixel in the high temporal and spatial resolution NDVI time series characteristics, according to the peak size and the distance between the peaks Set the preset peak filtering rules, and the number of peaks after filtering is the multiple seeding level of the pixel;

Step 4: Combine the cultivated land plot boundaries in the vector land use data to carry out plot-by-plot statistics on the multi-cropping level of the pixels, and use the area ratio analysis method to calculate the average multi-cropping level of each cultivated land plot;

Step 5: Estimate the crop sown area of each arable land plot by the product of the area of each arable land plot and the average multiple cropping level, and count the sown area of crops in the land consolidation project area by means of spatial statistical summation.

Step 6. Use the interannual difference and spatial autocorrelation analysis methods to obtain the spatiotemporal changes of the crop sown area in the land consolidation project area on the plot scale, and separate and output the high-value cluster area and the low-value cluster area. The value agglomeration area represents a concentrated area of plots with increased crop sown area, and the low value agglomerated area represents an area of concentrated plots with a decreased crop sown area.

It should be noted that the high temporal resolution image in the present invention refers to the time series data with better quality that can be formed at least once every 16 days after processing. For example, a satellite provides one period of data every 8 days, and two periods of data every 16 days. , if one of the phases is of good quality, it will meet the requirements. Generally speaking, the higher the temporal resolution, the higher the probability of having a period of good quality time series data in 16 days. The high spatial resolution image means that the pixel size of the image is smaller than the size of the plot. Generally speaking, the higher the spatial resolution, the better, generally less than 30m can meet the requirements. In terms of temporal resolution, high spatial resolution High-resolution images are required to meet at least two high-quality images a year. NDVI stands for Normalized Vegetation Index, which is the prior art.

In the land consolidation project area, due to the large number of fields and the different farming conditions among the fields, the traditional field survey method is difficult to meet the needs of dynamic monitoring of crop sown area; Frequently, the growing seasons of different crops overlap. Based on the phenological period and growth characteristics of a certain type of crops, the use of a few images can only monitor the sown area of this type of crops, and it is difficult to monitor the sown area of each crop in the year. Effective monitoring of the sown area of crops; moreover, the area of cultivated land in my country is relatively small, and the low spatial resolution images traditionally used at the regional scale cannot be directly applied to the plot scale due to the problem of mixed pixels. The method of the present invention can avoid the aforementioned shortcomings, and fill the gap in the monitoring of the crop sown area at the plot scale in the land remediation project area.

The invention is based on multi-source remote sensing data and by means of a fusion algorithm, and lays a data basis for monitoring the planting area of crops on a plot scale. If there is no such data basis, it is difficult to effectively obtain the sown area of crops at the plot scale in the historical period, because the remote sensing images generally used in the historical period have great limitations: when the temporal resolution of the image is high, its spatial resolution is low; When the spatial resolution of the image is low and the temporal resolution is high, without data fusion, it is difficult to provide data that conforms to the monitoring of the plot scale, and it is even more difficult to obtain the information on the sown area of crops in the historical period.

On the basis of the fusion data, the peak value of each pixel can be obtained by using the quadratic difference algorithm to extract the peak value, but the peak value extraction is often prone to errors, so it is necessary to pay special attention to the true and false peak values when using the plot scale. Based on the RGB image of Google Earth, the typical crop feature space is extracted to identify the true and false peaks, which can improve the peak recognition accuracy, and the feature space is easy to extract and use. Since the size of the fused pixel is generally smaller than the size of the plot, the present invention proposes to use the method of block statistics and area ratio to obtain the average multi-cropping level of the plot scale on the basis of the multi-cropping level of the pixel. The product of the plot area is used to obtain the plot-scale crop sown area. In order to cooperate with land remediation planning and fine management, the present invention proposes to use spatial statistical method to detect the characteristic area of dynamic change of crop sown area on the basis of plot-scale crop sown area extraction. The invention can simply and accurately obtain the crop sown area and its variation characteristics at the plot scale.

The invention utilizes the respective advantages of high temporal resolution images and high spatial resolution images, and combines vector land use data to obtain the results of high temporal and spatial resolution NDVI of agricultural land plots in the land remediation project area; The texture features are used to extract the crop planting feature space of paddy field, dry land and irrigated land. Then, the quadratic difference algorithm and the statistical method of the plot are used to realize the extraction of the multi-cropping level of the plot scale in the land consolidation project area. Finally, the land consolidation is realized according to the spatial statistical method. Extraction of plot-scale crop sown area and its variation characteristics in the project area.

The invention extracts the sown area of crops at the plot scale in the land remediation project area and its variation characteristics by a creative method, which can greatly save time, labor and cost, and is beneficial to the fine management of agricultural land in the land remediation project area and the improvement of land remediation planning.

Description of drawings

The present invention will be further described below in conjunction with the accompanying drawings.

Figure 1 is the RGB map of Google Earth in the land consolidation project area.

Figure 2 is a map of vector land use types in the land consolidation project area.

Figure 3 is a high spatiotemporal resolution NDVI time series feature map of typical crop planting areas.

Figure 4 shows the multiple cropping level of the plot-scale crops in the land consolidation project area in 2001.

Figure 5 shows the multiple cropping level of plot-scale crops in the land consolidation project area in 2010.

Figure 6 is the plot-scale crop sown area in the land consolidation project area in 2001.

Figure 7 is the plot-scale crop sown area in the land consolidation project area in 2010.

Figure 8 is a graph of the change in the sown area of crops at the plot scale in the land consolidation project area.

Fig. 9 is a map showing the variation characteristics of the sown area of crops at the plot scale in the land consolidation project area.

Detailed ways

Example

The method for monitoring the sown area of crops on a plot scale in a land remediation project area of the present embodiment includes the following steps:

Step 1. Obtain the high temporal resolution images, high spatial resolution images and vector land use data of the year to be monitored covering the land remediation project area. The vector land use data includes the boundaries of arable land and land use types, as shown in Figure 2. Show.

In this embodiment, the land consolidation project area is located in Nantong City, Jiangsu Province. The images used are free images from 2001 and 2010, of which the high temporal resolution images are Terra MODIS images, and the high spatial resolution images are LandsatOLI images. The Terra MODIS image data is 1 image per day with a spatial resolution of 250 meters; the Landsat OLI data selects two cloud-free images per year with a spatial resolution of 30 meters; the scale of the vector land use data is 1:5000, and the accuracy can meet the requirements.

Firstly, data preprocessing is performed on high temporal resolution images and high spatial resolution images, that is, radiometric correction, atmospheric correction and geometric correction are performed, and projection conversion is performed based on the projection space of vector land use data.

The high temporal resolution NDVI and the high spatial resolution NDVI are obtained from the high temporal resolution image and the high spatial resolution image respectively through the band operation. This embodiment is based on the near-infrared band reflectance and red light in the image data in ENVI software The measurement and analysis of the reflectivity of the band is obtained, and the calculation formula is:

NDVI=(NIR-RED)/(NIR+RED)

where NIR is the reflectance in the near-infrared band, and RED is the reflectance in the red band.

The high temporal resolution image Terra MODIS data is combined with the maximum value and S-G filtering algorithm to remove noise interference such as clouds and water vapor to form a real NDVI data every 16 days.

The high temporal and spatial resolution NDVI is obtained by fusing the high temporal resolution NDVI and the high spatial resolution NDVI by the ESTARFM method, and the high temporal and spatial resolution image NDVI time series data with a spatial resolution of 30 meters is obtained every 16 days.

Radiometric correction, atmospheric correction, geometric correction, projection conversion, band operation, maximum value synthesis, S-G filtering algorithm, ESTARFM (fusion) method are existing technologies, and you can refer to books or literature related to remote sensing image processing or software operation guide.

Step 2: Using the cropping tool in ArcGIS software as a constraint condition, the high spatial and temporal resolution NDVI is segmented to obtain a high spatial and temporal resolution image NDVI of the farmland in the land remediation project area, which is the subsequent farmland. The typical regional characteristics of crops in dry land and irrigated land provide a reasonable range, and make the subsequently extracted multi-cropping level statistical unit boundaries consistent with the arable land plots in the existing vector land use data.

According to the boundaries of the cultivated land, combined with the high-scoring images (RGB images) in Google Earth, and the vegetation planting areas identified by spectral and texture features, the typical crop planting areas in the land remediation project area are obtained. The RGB images are shown in Figure 1. It can be downloaded from the Google Earth software platform.

According to the land use type, the typical crop planting area is divided into three types: paddy field, dry land and irrigated land, and the high temporal and spatial resolution NDVI time series features of these three types of crop planting typical areas are extracted respectively. The time series features include the shape of the curve formed by the high temporal and spatial resolution NDVI time series data, the peak value of the curve, and the temporal distance between the peaks.

In this embodiment, feature space extraction is performed based on RGB images in Google Earth, and then the true value judgment rule is set: judge whether there is vegetation through RGB images, and extract corresponding land use types in combination with land use types (irrigated land, dry land, paddy field). The typical vegetation area on the type, combined with high temporal and spatial resolution NDVI to extract crop planting time series features (feature space). Through the feature space extraction method, the real value screening rules are set instead of field investigation to obtain empirical values, which is highly practical and convenient. Among them, the identification of vegetation planting areas based on spectral and texture features through RGB images in Google Earth is an existing technology, and you can refer to "The Method of Making Forestry Maps Based on Google Earth" (Author: Zhang Zhichao, Inner Mongolia Forestry Survey Design Vol.41 No.3 , May 2018), or refer to "Remote Sensing Analysis of Land Use Types and Terrain Distribution of Pingtan Island Based on Google Earth Images" (Author: Gao Wei et al., Shelter Forest Science and Technology Issue 7 (Total Issue 178), July 2018) .

Step 3: According to the high temporal and spatial resolution NDVI time series characteristics of typical areas of three types of crop planting, combined with the quadratic difference algorithm to extract the peak value of each pixel in the high temporal and spatial resolution NDVI time series characteristics, as shown in Figure 3, according to the peak size As well as the distance between the peaks, set the preset peak filtering rules, and the number of peaks after filtering is the multiple seeding level of the pixel. The quadratic difference algorithm is the prior art and can be implemented by IDL software.

In this embodiment, when extracting the peak value of each pixel in the high temporal and spatial resolution NDVI time series feature, the peak value is judged true and false: if a certain peak value is lower than 95% of the minimum NDVI peak value of paddy field, agricultural land, and irrigated land, Or if the time interval between the peak and other peaks is less than 5 periods of data, the peak is considered to be a false peak and the peak is eliminated.

Step 4: Combining the cultivated land plot boundaries in the vector land use data, the multi-cropping level of the pixel is counted by plot, and the average multi-cropping level of each cultivated land plot is calculated by using the area ratio analysis method. Sub-plot statistics and area ratio analysis were completed in ArcGIS Zonal tool and Map Algebra tool respectively. The obtained plot-scale crop multiple cropping levels in 2001 and 2010 are shown in Figure 4 and Figure 5, respectively.

In order to verify the accuracy of the multi-cropping level results at the plot scale, field investigations were conducted on many of the plots, and the results showed that the accuracy met the requirements.

Step 5: Estimate the crop sown area of each arable land plot by the product of the area of each arable land plot and the average multiple cropping level, and count the sown area of crops in the land consolidation project area by means of spatial statistical summation. The obtained plot-scale crop sown areas in 2001 and 2010 are shown in Figure 6 and Figure 7, respectively.

Comparing the sown area in 2001 and 2010, the changes in the sown area of crops in the land consolidation project area and the extraction of characteristic areas are carried out. The results are shown in Figure 8 and Figure 9, respectively. The specific method is as follows: using the Map Algebra tool to calculate the change of the crop sown area at the plot scale, using the local spatial autocorrelation analysis method to obtain the characteristic area of the change, and separating and outputting the high-value cluster area and the low-value cluster area. High-value agglomeration areas represent the concentrated areas of plots with increased sown area; low-value agglomeration areas represent the concentrated areas of plots with decreased sown area.

The present invention is not limited to the specific technical solutions described in the foregoing embodiments, and in addition to the foregoing embodiments, the present invention may also have other embodiments. For those skilled in the art, any modifications, equivalent replacements, improvements and other technical solutions formed within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims (5)

1. A method for monitoring the plot scale crop seeding area of a land reclamation project area comprises the following steps:

acquiring a high-time resolution image, a high-time resolution image and vector land utilization data of a year to be monitored, which covers a land improvement project area, wherein the vector land utilization data comprises a cultivated land plot boundary and a land utilization type;

respectively obtaining high time resolution NDVI and high spatial resolution NDVI from the high time resolution image and the high spatial resolution image, and fusing the high time resolution NDVI and the high spatial resolution NDVI by an ESTARFM method to obtain high spatial resolution NDVI;

secondly, dividing the high spatial-temporal resolution NDVI by taking the farmland block boundary as a constraint condition to obtain a high spatial-temporal resolution image NDVI of the agricultural land in the land reclamation project area;

according to the boundary of a cultivated land plot, combining a Google Earth high-resolution image to identify a vegetation planting area according to spectral and textural features, and obtaining a crop planting typical area of a land improvement project area;

dividing the crop planting typical area into a paddy field, a dry land and a water-irrigated land according to the land utilization type, and respectively extracting high-space-time resolution NDVI time sequence characteristics of the three types of crop planting typical areas, wherein the high-space-time resolution NDVI time sequence characteristics comprise a curve shape, a curve peak value and a time distance between wave peaks, and the curve shape, the curve peak value and the time distance are formed by high-space-time resolution NDVI time sequence data;

step three, extracting the peak value of each pixel in the high space-time resolution NDVI time sequence characteristics by combining a secondary difference algorithm according to the high space-time resolution NDVI time sequence characteristics of the typical planting areas of the three types of crops, setting a preset peak value screening rule according to the size of the peak value and the distance between the peak values, wherein the number of the screened peak values is the multiple planting level of the pixel;

step four, combining cultivated land block boundaries in the vector land utilization data to carry out block division statistics on the multiple planting level of the pixels, and calculating by adopting an area proportion analysis method to obtain the average multiple planting level of each cultivated land block;

and step five, estimating the crop seeding area of each cultivated land block according to the product of the area of each cultivated land block and the average multiple planting level, and counting the crop seeding area of the land improvement project area in a spatial statistics and summation mode.

2. The land improvement project area plot scale crop planting area monitoring method of claim 1, wherein: and step six, acquiring the space-time change of the crop seeding area of the land improvement project area on the land parcel scale by adopting an inter-annual difference value and a space autocorrelation analysis method, and separating and outputting a high-value gathering area and a low-value gathering area, wherein the high-value gathering area represents an area in which the land parcels with increased crop seeding area are gathered, and the low-value gathering area represents an area in which the land parcels with reduced crop seeding area are gathered.

3. The land improvement project area plot scale crop planting area monitoring method of claim 1, wherein: in the first step, firstly, the high-time resolution image and the high-space resolution image are preprocessed, wherein the preprocessing comprises radiation correction, atmospheric correction and geometric correction, and projection conversion is carried out by taking a vector land utilization data projection space as a reference.

4. The land improvement project area plot scale crop planting area monitoring method of claim 1, wherein: in the first step, after the high time resolution NDVI data are obtained, the high time resolution NDVI is reconstructed by adopting a maximum synthesis method and an S-G filtering algorithm, so that adjacent high time resolution NDVI arranged according to a time sequence have the same preset time interval.

5. The land improvement project area plot scale crop planting area monitoring method of claim 1, wherein: in the third step, when the peak value of each pixel in the high space-time resolution NDVI time sequence characteristic is extracted, the true and false judgment is carried out on the peak value, which is specifically as follows: and if a certain peak value is lower than 95% of the minimum value of the NDVI peak values of the paddy field, the agricultural land and the water-pouring land, or the time interval between the peak value and other peak values is smaller than a preset interval, the peak value is considered as a false peak value and the peak value is removed.