CN115861790A - Cultivated land remote sensing image analysis method, device, equipment, storage medium and product - Google Patents
Cultivated land remote sensing image analysis method, device, equipment, storage medium and product Download PDFInfo
- Publication number
- CN115861790A CN115861790A CN202211338255.XA CN202211338255A CN115861790A CN 115861790 A CN115861790 A CN 115861790A CN 202211338255 A CN202211338255 A CN 202211338255A CN 115861790 A CN115861790 A CN 115861790A
- Authority
- CN
- China
- Prior art keywords
- remote sensing
- sensing image
- target
- cultivated land
- historical
- 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.)
- Pending
Links
- 238000003703 image analysis method Methods 0.000 title claims abstract description 25
- 238000003860 storage Methods 0.000 title claims abstract description 14
- 238000012549 training Methods 0.000 claims abstract description 74
- 238000013145 classification model Methods 0.000 claims abstract description 44
- 238000009826 distribution Methods 0.000 claims abstract description 27
- 238000010191 image analysis Methods 0.000 claims abstract description 21
- 238000004458 analytical method Methods 0.000 claims abstract description 19
- 238000013508 migration Methods 0.000 claims description 21
- 230000005012 migration Effects 0.000 claims description 21
- 238000004590 computer program Methods 0.000 claims description 16
- 238000003967 crop rotation Methods 0.000 claims description 11
- 238000003384 imaging method Methods 0.000 claims description 10
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 4
- 238000013526 transfer learning Methods 0.000 abstract description 11
- 238000012545 processing Methods 0.000 abstract description 5
- 238000000034 method Methods 0.000 description 24
- 238000007637 random forest analysis Methods 0.000 description 8
- 238000004891 communication Methods 0.000 description 5
- 238000013528 artificial neural network Methods 0.000 description 4
- 238000012512 characterization method Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 238000002372 labelling Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 230000007547 defect Effects 0.000 description 3
- 238000002472 multi-metal deposition Methods 0.000 description 3
- 238000005070 sampling Methods 0.000 description 3
- 238000012795 verification Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 239000002699 waste material Substances 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000001172 regenerating effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention provides a cultivated land remote sensing image analysis method, a cultivated land remote sensing image analysis device, cultivated land remote sensing image analysis equipment, a storage medium and a product, which relate to the technical field of image processing and comprise the following steps: acquiring a historical remote sensing image set and a current remote sensing image of a target cultivated land area; determining N target historical remote sensing images in the historical remote sensing image set based on a data distribution difference analysis result of the current remote sensing image and the historical remote sensing image set, wherein each pixel in the target historical remote sensing image corresponds to one cultivated land type label, and N is a positive integer; transferring the cultivated land type label in the target historical remote sensing image to the current remote sensing image in a multi-source transfer learning mode to obtain the current remote sensing target image; and training the preset depth forest model based on the N target historical remote sensing images and the current remote sensing target image to obtain a trained cultivated land remote sensing image classification model.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a cultivated land remote sensing image analysis method, a cultivated land remote sensing image analysis device, cultivated land remote sensing image analysis equipment, a cultivated land remote sensing image storage medium and a cultivated land remote sensing image product.
Background
Maintaining the quantity and quality of cultivated land is very important for grain production, and a phenomenon of abandoned land of different degrees occurs in many places, which causes the problems of waste of land resources and reduction of cultivated land quality.
In the related art, it is usually necessary to manually determine and analyze an image of a cultivation area to determine a abandoned cultivated land, which is generally slow in efficiency and has a certain misjudgment rate.
Therefore, how to effectively detect abandoned cultivated land has become an urgent problem to be solved in the industry.
Disclosure of Invention
The invention provides a cultivated land remote sensing image analysis method, a cultivated land remote sensing image analysis device, equipment, a storage medium and a product, which are used for solving the defect of how to effectively detect abandoned cultivated land in the prior art.
The invention provides a cultivated land remote sensing image analysis method, which comprises the following steps:
acquiring a historical remote sensing image set and a current remote sensing image of a target cultivated land area;
determining N target historical remote sensing images in the historical remote sensing image set based on the data distribution difference analysis result of the current remote sensing image and the historical remote sensing image set, wherein each pixel in the target historical remote sensing images corresponds to a cultivated land type label, and N is a positive integer;
transferring the cultivated land type label in the target historical remote sensing image to the current remote sensing image in a multi-source transfer learning mode to obtain a current remote sensing target image;
training a preset depth forest model based on the N target historical remote sensing images and the current remote sensing target image to obtain a trained farmland remote sensing image classification model;
the farmland remote sensing image classification model is used for classifying farmland types of all pixels in the remote sensing image of the target farmland region.
According to the farmland remote sensing image analysis method provided by the invention, after the trained farmland remote sensing image classification model is obtained, the method further comprises the following steps:
inputting the current remote sensing image into the trained farmland remote sensing image classification model, and outputting the farmland type of each pixel in the current remote sensing image;
and acquiring a target pixel with a bare land type cultivated land type in each pixel, and determining a abandoned land in the land corresponding to the target pixel based on the imaging time of the current remote sensing image and crop rotation information of the land corresponding to the target pixel.
According to the farmland remote sensing image analysis method provided by the invention, based on the data distribution difference analysis result of the current remote sensing image and the historical remote sensing image set, N target historical remote sensing images in the historical remote sensing image set are determined, and the method comprises the following steps:
calculating the maximum average difference value information of the current remote sensing image and each historical remote sensing image in the historical remote sensing image set;
and determining a historical remote sensing image subset in which the sum of the weights of the maximum mean difference value information in the historical remote sensing image set is greater than a first preset threshold value and the sum of the image quantity is less than a second preset threshold value, wherein the historical remote sensing image subset comprises N target historical remote sensing images.
According to the cultivated land remote sensing image analysis method provided by the invention, a preset depth forest model is trained based on N target historical remote sensing images and the current remote sensing target image to obtain a trained cultivated land remote sensing image classification model, and the method comprises the following steps:
taking one target historical remote sensing image or current remote sensing target image as a training sample to obtain a plurality of training samples;
and training the preset deep forest model by using a plurality of training samples, and stopping training under the condition of meeting preset training conditions to obtain a trained farmland remote sensing image classification model.
According to the farmland remote sensing image analysis method provided by the invention, the farmland type label comprises the following steps: a water type tag, an artificial ground object type tag, a forest land type tag, a crop type tag, and a bare land type tag.
The invention also provides a cultivated land remote sensing image analysis device, which comprises:
the acquisition module is used for acquiring a historical remote sensing image set and a current remote sensing image of a target cultivated land area;
the determining module is used for determining N target historical remote sensing images in the historical remote sensing image set based on the data distribution difference analysis result of the current remote sensing image and the historical remote sensing image set, wherein each pixel in the target historical remote sensing images corresponds to one cultivated land type label, and N is a positive integer;
the migration module is used for migrating the cultivated land type label in the target historical remote sensing image to the current remote sensing image in a multi-source migration learning mode to obtain a current remote sensing target image;
the training module is used for training a preset depth forest model based on the N target historical remote sensing images and the current remote sensing target image to obtain a trained farmland remote sensing image classification model;
the farmland remote sensing image classification model is used for classifying farmland types of all pixels in the remote sensing image of the target farmland area.
According to the farmland remote sensing image analysis device provided by the invention, the device is also used for:
inputting the current remote sensing image into the trained farmland remote sensing image classification model, and outputting the farmland type of each pixel in the current remote sensing image;
and acquiring a target pixel with a bare land type cultivated land type in each pixel, and determining a abandoned land plot in the land plot corresponding to the target pixel based on the imaging time of the current remote sensing image and the crop rotation information of the land plot corresponding to the target pixel.
According to the farmland remote sensing image analysis device provided by the invention, the device is also used for:
calculating the maximum average difference value information of the current remote sensing image and each historical remote sensing image in the historical remote sensing image set;
and determining a historical remote sensing image subset in which the sum of the weights of the maximum mean difference value information in the historical remote sensing image set is greater than a first preset threshold value and the sum of the image quantity is less than a second preset threshold value, wherein the historical remote sensing image subset comprises N target historical remote sensing images.
According to the farmland remote sensing image analysis device provided by the invention, the device is also used for:
taking one target historical remote sensing image or current remote sensing target image as a training sample to obtain a plurality of training samples;
and training the preset deep forest model by using a plurality of training samples, and stopping training under the condition of meeting preset training conditions to obtain a trained farmland remote sensing image classification model.
According to the farmland remote sensing image analysis device provided by the invention, the farmland type label comprises: a water type tag, an artificial ground object type tag, a forest land type tag, a crop type tag, and a bare land type tag.
The invention also provides electronic equipment which comprises a memory, a processor and a computer program which is stored on the memory and can run on the processor, wherein the processor executes the program to realize the remote sensing image analysis method for the cultivated land.
The present invention also provides a non-transitory computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements a method of remote sensing image analysis of a cultivated land as any one of the above.
The invention also provides a computer program product comprising a computer program which, when executed by a processor, implements the remote sensing image analysis method of cultivated land as described in any one of the above.
According to the cultivated land remote sensing image analysis method, the device, the equipment, the storage medium and the product, the target historical remote sensing image with small feature distribution difference is selected from the historical remote sensing images in a centralized manner through the current remote sensing image to perform subsequent model training, the cultivated land type label in the target historical remote sensing image is migrated into the current remote sensing image in a multi-source migration learning manner, a labeling sample is given to the new image, so that the training sample does not need to be manually acquired on the new image, the automation of model training and land object classification is realized, and finally, the preset deep forest model is trained according to N target historical remote sensing images and the current remote sensing target image to obtain the cultivated land remote sensing image classification model for classifying cultivated land types of all pixels in the remote sensing image of the target cultivated land area, so that abandoned cultivated land is automatically and efficiently identified.
Drawings
In order to more clearly illustrate the technical solutions of the present invention or the prior art, the drawings needed for the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and those skilled in the art can also obtain other drawings according to the drawings without creative efforts.
FIG. 1 is a schematic flow chart of a remote sensing image analysis method for cultivated land provided by the embodiment of the application;
FIG. 2 is a main flow chart of remote sensing image analysis of cultivated land provided by the embodiment of the application;
fig. 3 is a schematic view illustrating a waste identification process according to an embodiment of the present application;
FIG. 4 is a schematic structural diagram of a remote sensing image analysis device for cultivated land described in the embodiments of the present application;
fig. 5 is a schematic structural diagram of an electronic device provided in the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is obvious that the described embodiments are some, but not all embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Fig. 1 is a schematic flow chart of a remote sensing image analysis method for cultivated land provided in an embodiment of the present application, as shown in fig. 1, including:
specifically, the target cultivated land area described in the embodiment of the present application may specifically be a cultivated land on which a abandoned situation analysis needs to be performed, and the target cultivated land area may be set by a user according to a user requirement.
The historical remote sensing image set described in the embodiment of the application may include historical remote sensing images of a plurality of target cultivated land areas, and the historical remote sensing images may be acquired from stored historical data.
The current remote sensing image in the embodiment of the application refers to a recently acquired remote sensing image of a target cultivated land area.
More specifically, because the resolutions of the images acquired by different sensors are different, after an initial historical remote sensing image set and a current remote sensing image are acquired, spatial consistency processing is performed on the images, the spatial resolution of data is unified with that in a historical image library, a ground object distinguishing error caused by inconsistent resolution is reduced, and finally the acquisition of the historical remote sensing image set and the current remote sensing image is realized.
more specifically, remote sensing images of different sensor types at different periods are accumulated in the same cultivated land area, a historical remote sensing image library of the area is formed, and the feature distribution of the same type of ground object on different images has a certain difference, so that the historical remote sensing image can be further selected from the historical sensing images in a centralized manner to be used as a training sample.
More specifically, in the embodiment of the present application, the data distribution difference analysis result of the current remote sensing image and the historical remote sensing image set may specifically be a result of performing maximum mean difference analysis on the current remote sensing image and each historical remote sensing image.
And determining a target historical remote sensing image with lower distribution difference between the historical remote sensing image set and the current remote sensing image through the data distribution difference analysis result.
More specifically, the target history remote sensing image in the embodiment of the present application includes remote sensing images of a plurality of plots, the remote sensing image of each plot is composed of at least one pixel, and in the embodiment of the present application, a cultivated land type label may be marked on the remote sensing image of each plot, that is, each corresponding pixel may also carry a cultivated land type label.
The tillable area type labels described in the embodiments of the present application are used to mark the actual tillable area type of each plot.
specifically, the migration knowledge extracted by the multi-source migration learning method is not limited to a single source domain data set any more, but is derived from two or more source domain data sets, and the information in the existing multi-source data can be fully utilized. The migration learning technology can effectively migrate the cultivated land type labels in the processed target historical remote sensing images to the current remote sensing images to a certain extent, and automatic labeling of the current remote sensing images to a certain extent is achieved.
In the embodiment of the application, part of pixels with rich information content in the current remote sensing image can acquire the cultivated land type labels through transfer learning, that is, after the transfer learning, part of pixels in the current remote sensing target image already carry the cultivated land type labels, however, a lot of pixels still exist in the current remote sensing target image and do not carry the cultivated land type labels, the cultivated land type corresponding to the pixels cannot be determined, and further analysis needs to be performed through a subsequent model.
And (3) performing migration learning on the current remote sensing image according to each historical remote sensing image by adopting a migration component analysis (TCA) algorithm in the migration learning method.
the farmland remote sensing image classification model is used for classifying farmland types of all pixels in the remote sensing image of the target farmland area.
Specifically, in the embodiment of the application, the deep forest model has the characteristics of strong representation learning capacity, fewer hyper-parameters, strong model structure interpretability, low calculation overhead and the like.
After multi-source migration learning is carried out, training is respectively carried out on a preset depth forest model based on N target historical remote sensing images and current remote sensing target images, after a preset training condition is met, the training is stopped, a trained cultivated land remote sensing image classification model is obtained, the model can classify cultivated land types of all pixels in the remote sensing images of the target cultivated land area, and then the abandoned land block in the target cultivated land area is effectively identified.
In the embodiment of the application, a target history remote sensing image with small characteristic distribution difference is selected from the history remote sensing image in a centralized mode through a current remote sensing image to perform subsequent model training, a farmland type label in the target history remote sensing image is migrated to the current remote sensing image in a multi-source migration learning mode, a labeling sample is given to the new image, so that a training sample does not need to be manually collected on the new image, automation of model training and land and feature classification is realized, and finally, a preset depth forest model is trained according to N target history remote sensing images and the current remote sensing target image to obtain a farmland remote sensing image classification model for classifying the farmland types of all pixels in the remote sensing image of a target farmland area, so that abandoned farmland is automatically and efficiently identified.
Optionally, after the trained classification model of remote sensing images of farmland is obtained, the method further includes:
inputting the current remote sensing image into the trained farmland remote sensing image classification model, and outputting the farmland type of each pixel in the current remote sensing image;
and acquiring a target pixel with a bare land type cultivated land type in each pixel, and determining a abandoned land plot in the land plot corresponding to the target pixel based on the imaging time of the current remote sensing image and the crop rotation information of the land plot corresponding to the target pixel.
Specifically, in the embodiment of the application, after the cultivated land type of each pixel in the current remote sensing image is obtained, the multi-classification result weighted voting is performed on each pixel, and the pixel is given to the category with the highest weight, so that the cultivated land category of each land parcel is obtained.
However, since cultivation is greatly affected by climate and cultivation period, some crops planted in a land are planted only in the planting period, and other crops fed back in the remote sensing image in other periods may belong to a bare land type, but this does not represent that the land is really a bare land.
Therefore, in the embodiment of the present application, it is necessary to further combine the imaging time of the current remote sensing image and the crop rotation information of the plot corresponding to the target pixel for judgment.
Firstly, determining whether the crops in the plot belong to a planting period or not based on the crop rotation information of the plot and the imaging time of the current remote sensing image. The plot crop rotation information can be specifically determined according to the type of crops which should be planted in the plot in advance.
Further, if the cultivation type of the pixel is a bare land type, it needs to be further analyzed whether the land parcel corresponding to the pixel is in the planting period, if the land parcel of the pixel is in the planting period, it indicates that the land parcel may be a abandoned land parcel, and if the land parcel of the pixel is in a non-planting period, it indicates that the land parcel is probably not an abandoned land parcel.
More specifically, if a abandoned land parcel is identified, the information is reported, and it is further confirmed manually whether the land parcel is a abandoned land parcel.
In the embodiment of the application, the land parcel of the bare land type output by the model can be further secondarily judged according to the imaging time of the current remote sensing image and the crop rotation information of the land parcel corresponding to the target pixel, so that the identification accuracy of the abandoned land parcel is effectively ensured.
Optionally, determining N target historical remote sensing images in the historical remote sensing image set based on a data distribution difference analysis result of the current remote sensing image and the historical remote sensing image set, including:
calculating the maximum average difference value information of the current remote sensing image and each historical remote sensing image in the historical remote sensing image set;
and determining a historical remote sensing image subset in which the sum of the weights of the maximum mean difference value information in the historical remote sensing image set is greater than a first preset threshold value and the sum of the image quantity is less than a second preset threshold value, wherein the historical remote sensing image subset comprises N target historical remote sensing images.
Specifically, in the present invention, a Maximum Mean Difference (MMD) is used to measure a distribution difference between different images, where MMD is one of the most commonly used difference measurement indexes in a migration learning study, and the MMD maps source domain data and target domain data into a Regenerative Kernel Hilbert Space (RKHS) and calculates a distance between the Mean values of the two sets of data. The MMD is calculated as:
wherein VMMD represents the value of MMD, XS and XT represent the source domain data and the target domain data, respectively; nS and nT respectively represent the quantity of the source domain data and the target domain data; ψ (-) represents a kernel function. The larger the value of the MMD, the larger the difference in the distribution of the two sets of data, and when the distribution of the two sets of data is completely consistent, the value of the MMD is equal to 0.
Firstly, respectively calculating MMDs of all images in a historical remote sensing image set and a current remote sensing image, then normalizing all MMDs, taking out a group of historical images with the sum of weighted values of the MMDs being large enough and the number of the images being small as much as possible in a self-adaptive dynamic selection mode, using the historical images as target historical remote sensing images, and further obtaining a historical remote sensing image subset.
In the embodiment of the application, the maximum mean difference value information of the current remote sensing image and each historical remote sensing image in the historical remote sensing image set is calculated, so that the historical remote sensing image with the smallest difference as possible from the current remote sensing image is effectively screened from the historical remote sensing image set as a training sample, and the model for subsequent training can be effectively ensured to be capable of better identifying the current cultivated land type of the target cultivated land area.
Optionally, training a preset depth forest model based on N target historical remote sensing images and the current remote sensing target image to obtain a trained classification model of remote sensing images of cultivated land, including:
taking one target historical remote sensing image or current remote sensing target image as a training sample to obtain a plurality of training samples;
and training the preset deep forest model by using a plurality of training samples, and stopping training under the condition of meeting preset training conditions to obtain a trained farmland remote sensing image classification model.
Specifically, after multi-source transfer learning is carried out, models are trained respectively based on transferred target historical remote sensing images, and trained farmland remote sensing image classification models are obtained.
The method uses the deep forest as a classification model of remote sensing images of the cultivated land.
The deep forest has the characteristics of strong representation learning capability, less hyper-parameters, strong model structure interpretability, small calculation overhead and the like, and although the deep neural network has strong performance, the deep forest has the defects of dependence on a large number of training samples, complex model, poor interpretability, more hyper-parameters, high parameter adjustment difficulty and the like. If a strong characterization learning capability of a Deep neural network is introduced into a proper learning model, the performance of the model can be enough to be similar to that of the Deep neural network, and the defects can be avoided, so that a Deep Forest (DF) is provided. DF is a model for decision tree integration, mainly comprising two parts, multi-granularity scanning and cascading forest, which will be briefly described below.
Multi-granularity scanning is used to enhance the characterization learning capabilities of the model. The multi-granularity scanning process of the sequence data is as shown in the above figure, and the original feature dimension of the input sample is equal to 400, and the number of categories is equal to 3. When the size of the sampling window is 100 dimensions and the sliding distance is 1, sliding sampling on the sample can obtain 301 sub-samples, and the characteristic dimension of each sub-sample is equal to 100. Inputting one sub-sample into a random forest to obtain prediction probabilities of 3 categories, so that inputting 301 sub-samples into 2 random forests can obtain 301 × 3 × 2=1806 category prediction probabilities in total, and splicing the category prediction probabilities together to serve as a multi-granularity feature of a sample. In addition, the multiple granularity scanning is performed by using a plurality of sampling windows with different sizes, so that the dimensionality of the multiple granularity characteristic can be further increased. Because each subsample represents the local characteristics of the original sample, the multi-granularity scanning is equivalent to structured upsampling on the original characteristics, and the method enhances the characterization learning capability of the model and is beneficial to improving the performance of the model.
The characterization learning ability of the deep neural network mainly derives from the layer-by-layer processing of the original features, and based on the inspiration, the cascade forest also consists of cascade layers which are progressive layer by layer, as shown in the figure. To enhance diversity, two fully random forests and two random forests (hereinafter collectively referred to as random forests) are used per cascaded layer. And taking the class prediction probability output by the random forest as the enhancement feature of the sample, outputting 12 class prediction probabilities by 4 random forests as the enhancement feature of the sample when the class number is 3 for one sample, and splicing the enhancement feature and the multi-granularity feature together to serve as the input feature of the next cascade layer.
In each level of cascade layer of the cascade forest, training samples are divided into a growth set and a verification set, the growth set is used for training a model, the verification set is used for verifying the performance of the model, and when the prediction precision of the model to the verification set is not obviously improved any more, the generation of a new cascade layer is stopped. The method can avoid overfitting of the model, and can adjust the complexity of the model in a self-adaptive manner, so that the model can adapt to training data of different scales. And when the cascade forests stop growing, averaging the class prediction probabilities of the 4 random forests in the last cascade layer on the sample, and taking the class with the highest prediction probability as the final prediction result of the sample.
The preset training condition described in the embodiment of the present application may specifically be that a preset number of times of training is satisfied, for example, training is completed for 150 times, and the preset training condition may also be that a preset training event is satisfied, for example, training is completed for 30 minutes.
In the embodiment of the application, the preset deep forest model can be trained through a plurality of training samples, and finally, a model for classifying the cultivation land types of all pixels in the remote sensing image of the target cultivation land area is obtained, so that the abandoned land block can be effectively identified.
The cultivated land remote sensing image analysis device provided by the invention is described below, and the cultivated land remote sensing image analysis device described below and the cultivated land remote sensing image analysis method described above can be referred to correspondingly.
Optionally, fig. 2 is a main flowchart of remote sensing image analysis of a cultivated land provided in the embodiment of the present application, and as shown in fig. 2, the main flowchart includes:
because the resolutions of the images acquired by different sensors are different, the input images are firstly subjected to spatial consistency processing, the spatial resolution of the data is unified with the data in the historical remote sensing image data set, and the ground feature distinguishing error caused by the inconsistent resolution is reduced.
The feature distribution of the same type of ground object on different images has certain differences, and the differences are called distribution differences among domains in transfer learning. The difference in distribution between domains affects the effect of transfer learning, and the smaller the difference between the two domains, the easier it is to transfer knowledge between the two domains.
Since remote sensing images of different sensor types at different periods are accumulated in the same region, a historical source domain remote sensing image library of the region is formed, and a group of most suitable source domain image sets are selected from the historical source domain remote sensing image library according to data distribution differences to obtain a low-difference historical remote sensing image set.
The method comprises the steps of carrying out multi-source migration learning according to a low-difference historical remote sensing image set and an input remote sensing image, then carrying out model training according to a current remote sensing target image after migration and the low-difference historical remote sensing image set, using a deep forest model in the model training process, obtaining a trained cultivated land remote sensing image classification model after completing the model training, further analyzing each pixel in the input remote sensing image according to the classification model, extracting an unplanted area in a cultivated land range by combining cultivated land pattern spots, and then extracting a abandoned cultivated land by combining the region crop phenological information.
Fig. 3 is a schematic view of a abandoned land recognition process in the embodiment of the present application, and as shown in fig. 3, it is determined whether the current land is planted according to the image imaging time of the current remote-sensing image and crop rotation information of the land, if the land is in the planting stage, it is determined whether there is a crop planted by combining with the land category information, and if there is no crop planted, it is determined that the land is a suspected abandoned land.
Fig. 4 is a schematic structural diagram of the remote sensing image analysis device for cultivated land described in the embodiment of the present application, as shown in fig. 4, including: an acquisition module 410, a determination module 420, a migration module 430, and a training module 440;
the obtaining module 410 is configured to obtain a historical remote sensing image set and a current remote sensing image of a target cultivated land area;
the determining module 420 is configured to determine N target historical remote sensing images in the historical remote sensing image set based on a data distribution difference analysis result of the current remote sensing image and the historical remote sensing image set, where each pixel in the target historical remote sensing image corresponds to one cultivated land type label, and N is a positive integer;
the migration module 430 is configured to migrate the cultivated land type label in the target historical remote sensing image to the current remote sensing image in a multi-source migration learning manner to obtain a current remote sensing target image;
the training module 440 is configured to train a preset depth forest model based on the N target historical remote sensing images and the current remote sensing target image to obtain a trained classification model of the farmland remote sensing image;
the farmland remote sensing image classification model is used for classifying farmland types of all pixels in the remote sensing image of the target farmland area.
Optionally, the apparatus is further configured to:
inputting the current remote sensing image into the trained farmland remote sensing image classification model, and outputting the farmland type of each pixel in the current remote sensing image;
and acquiring a target pixel with a bare land type cultivated land type in each pixel, and determining a abandoned land plot in the land plot corresponding to the target pixel based on the imaging time of the current remote sensing image and the crop rotation information of the land plot corresponding to the target pixel.
Optionally, the apparatus is further configured to:
calculating the maximum average difference value information of the current remote sensing image and each historical remote sensing image in the historical remote sensing image set;
and determining a historical remote sensing image subset in which the sum of the weights of the maximum mean difference value information in the historical remote sensing image set is greater than a first preset threshold value and the sum of the image quantity is less than a second preset threshold value, wherein the historical remote sensing image subset comprises N target historical remote sensing images.
Optionally, the apparatus is further configured to:
taking one target historical remote sensing image or current remote sensing target image as a training sample to obtain a plurality of training samples;
and training the preset deep forest model by using a plurality of training samples, and stopping training under the condition of meeting preset training conditions to obtain a trained farmland remote sensing image classification model.
Optionally, the tillable area type tag comprises: a water type tag, an artificial ground object type tag, a forest land type tag, a crop type tag, and a bare land type tag.
In the embodiment of the application, a target historical remote sensing image with smaller feature distribution difference is selected from the historical remote sensing images in a centralized manner through the current remote sensing image to perform subsequent model training, a farmland type label in the target historical remote sensing image is migrated into the current remote sensing image in a multi-source migration learning manner, a labeling sample is given to the new image, so that training samples do not need to be manually collected on the new image, automation of model training and land feature classification is realized, and finally, a preset depth forest model is trained according to N target historical remote sensing images and the current remote sensing target image to obtain a cultivated land remote sensing image classification model for classifying cultivated land types of all pixels in the remote sensing image of the target cultivated land area, so that abandoned cultivated land is automatically and efficiently identified.
Fig. 5 is a schematic structural diagram of an electronic device provided in the present invention, and as shown in fig. 5, the electronic device may include: a processor (processor) 510, a communication Interface (Communications Interface) 520, a memory (memory) 530 and a communication bus 540, wherein the processor 510, the communication Interface 520 and the memory 530 communicate with each other via the communication bus 540. Processor 510 may invoke logic instructions in memory 530 to perform a method of remote sensing image analysis of a tillable area, the method comprising: acquiring a historical remote sensing image set and a current remote sensing image of a target cultivated land area;
determining N target historical remote sensing images in the historical remote sensing image set based on the data distribution difference analysis result of the current remote sensing image and the historical remote sensing image set, wherein each pixel in the target historical remote sensing image corresponds to one cultivated land type label, and N is a positive integer;
transferring the cultivated land type label in the target historical remote sensing image to the current remote sensing image in a multi-source transfer learning mode to obtain a current remote sensing target image;
training a preset depth forest model based on the N target historical remote sensing images and the current remote sensing target image to obtain a trained farmland remote sensing image classification model;
the farmland remote sensing image classification model is used for classifying farmland types of all pixels in the remote sensing image of the target farmland region.
Furthermore, the logic instructions in the memory 530 may be implemented in the form of software functional units and stored in a computer readable storage medium when the software functional units are sold or used as independent products. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk, and various media capable of storing program codes.
In another aspect, the present invention also provides a computer program product, the computer program product comprising a computer program, the computer program being stored on a non-transitory computer-readable storage medium, wherein when the computer program is executed by a processor, the computer is capable of executing the remote sensing image analysis method provided by the above methods, the method comprising: acquiring a historical remote sensing image set and a current remote sensing image of a target cultivated land area;
determining N target historical remote sensing images in the historical remote sensing image set based on the data distribution difference analysis result of the current remote sensing image and the historical remote sensing image set, wherein each pixel in the target historical remote sensing image corresponds to one cultivated land type label, and N is a positive integer;
transferring the cultivated land type label in the target historical remote sensing image to the current remote sensing image in a multi-source transfer learning mode to obtain a current remote sensing target image;
training a preset depth forest model based on the N target historical remote sensing images and the current remote sensing target image to obtain a trained farmland remote sensing image classification model;
the farmland remote sensing image classification model is used for classifying farmland types of all pixels in the remote sensing image of the target farmland area.
In yet another aspect, the present invention also provides a non-transitory computer-readable storage medium having stored thereon a computer program, which when executed by a processor, implements a method for remote sensing image analysis of cultivated land provided by the above methods, the method comprising: acquiring a historical remote sensing image set and a current remote sensing image of a target cultivated land area;
determining N target historical remote sensing images in the historical remote sensing image set based on the data distribution difference analysis result of the current remote sensing image and the historical remote sensing image set, wherein each pixel in the target historical remote sensing image corresponds to one cultivated land type label, and N is a positive integer;
transferring the cultivated land type label in the target historical remote sensing image to the current remote sensing image in a multi-source transfer learning mode to obtain a current remote sensing target image;
training a preset depth forest model based on the N target historical remote sensing images and the current remote sensing target image to obtain a trained farmland remote sensing image classification model;
the farmland remote sensing image classification model is used for classifying farmland types of all pixels in the remote sensing image of the target farmland area.
The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment may be implemented by software plus a necessary general hardware platform, and may also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the embodiments or some parts of the embodiments.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. A cultivated land remote sensing image analysis method is characterized by comprising the following steps:
acquiring a historical remote sensing image set and a current remote sensing image of a target cultivated land area;
determining N target historical remote sensing images in the historical remote sensing image set based on the data distribution difference analysis result of the current remote sensing image and the historical remote sensing image set, wherein each pixel in the target historical remote sensing image corresponds to one cultivated land type label, and N is a positive integer;
migrating the cultivated land type label in the target historical remote sensing image to the current remote sensing image in a multi-source migration learning mode to obtain a current remote sensing target image;
training a preset depth forest model based on the N target historical remote sensing images and the current remote sensing target image to obtain a trained farmland remote sensing image classification model;
the farmland remote sensing image classification model is used for classifying farmland types of all pixels in the remote sensing image of the target farmland area.
2. The cultivated land remote sensing image analysis method according to claim 1, characterized in that after said trained cultivated land remote sensing image classification model is obtained, further comprising:
inputting the current remote sensing image into the trained farmland remote sensing image classification model, and outputting the farmland type of each pixel in the current remote sensing image;
and acquiring a target pixel with a bare land type cultivated land type in each pixel, and determining a abandoned land plot in the land plot corresponding to the target pixel based on the imaging time of the current remote sensing image and the crop rotation information of the land plot corresponding to the target pixel.
3. The cultivated land remote sensing image analysis method according to claim 1, wherein determining N target historical remote sensing images in the historical remote sensing image set based on the data distribution difference analysis result of the current remote sensing image and the historical remote sensing image set comprises:
calculating the maximum average difference value information of the current remote sensing image and each historical remote sensing image in the historical remote sensing image set;
and determining a historical remote sensing image subset in which the sum of the weights of the maximum mean difference value information in the historical remote sensing image set is greater than a first preset threshold value and the sum of the image quantity is less than a second preset threshold value, wherein the historical remote sensing image subset comprises N target historical remote sensing images.
4. The cultivated land remote sensing image analysis method according to claim 1, wherein training a preset depth forest model based on N target history remote sensing images and the current remote sensing target image to obtain a trained cultivated land remote sensing image classification model comprises:
taking one target historical remote sensing image or current remote sensing target image as a training sample to obtain a plurality of training samples;
and training the preset deep forest model by using a plurality of training samples, and stopping training under the condition of meeting preset training conditions to obtain a trained farmland remote sensing image classification model.
5. The cultivated land remote sensing image analysis method according to claim 1, characterized in that the cultivated land type label comprises: a water type tag, an artificial ground object type tag, a forest land type tag, a crop type tag, and a bare land type tag.
6. A cultivated land remote sensing image analysis device is characterized by comprising:
the acquisition module is used for acquiring a historical remote sensing image set and a current remote sensing image of a target cultivated land area;
the determining module is used for determining N target historical remote sensing images in the historical remote sensing image set based on the data distribution difference analysis result of the current remote sensing image and the historical remote sensing image set, wherein each pixel in the target historical remote sensing images corresponds to one cultivated land type label, and N is a positive integer;
the migration module is used for migrating the cultivated land type label in the target historical remote sensing image to the current remote sensing image in a multi-source migration learning mode to obtain a current remote sensing target image;
the training module is used for training a preset depth forest model based on the N target historical remote sensing images and the current remote sensing target image to obtain a trained farmland remote sensing image classification model;
the farmland remote sensing image classification model is used for classifying farmland types of all pixels in the remote sensing image of the target farmland area.
7. The cultivated land remote sensing image analysis apparatus according to claim 6, characterized in that said apparatus is further adapted to:
inputting the current remote sensing image into the trained farmland remote sensing image classification model, and outputting the farmland type of each pixel in the current remote sensing image;
and acquiring a target pixel with a bare land type cultivated land type in each pixel, and determining a abandoned land in the land corresponding to the target pixel based on the imaging time of the current remote sensing image and crop rotation information of the land corresponding to the target pixel.
8. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the remote sensing image analysis method of cultivated land according to any one of claims 1 to 5 when executing the program.
9. A non-transitory computer-readable storage medium having stored thereon a computer program, wherein the computer program, when executed by a processor, implements the remote sensing image analysis method of cultivated land according to any one of claims 1 to 5.
10. A computer program product comprising a computer program, characterized in that the computer program, when being executed by a processor, implements the remote sensing image analysis method of cultivated land according to any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211338255.XA CN115861790A (en) | 2022-10-28 | 2022-10-28 | Cultivated land remote sensing image analysis method, device, equipment, storage medium and product |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211338255.XA CN115861790A (en) | 2022-10-28 | 2022-10-28 | Cultivated land remote sensing image analysis method, device, equipment, storage medium and product |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115861790A true CN115861790A (en) | 2023-03-28 |
Family
ID=85662056
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211338255.XA Pending CN115861790A (en) | 2022-10-28 | 2022-10-28 | Cultivated land remote sensing image analysis method, device, equipment, storage medium and product |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115861790A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117975292A (en) * | 2024-03-25 | 2024-05-03 | 中国科学院东北地理与农业生态研究所 | Extraction method for temperate forest swamps |
| CN118135311A (en) * | 2024-03-13 | 2024-06-04 | 南京北斗创新应用科技研究院有限公司 | Heterogeneous time sequence image wetland monitoring method and device based on improved cascade forests |
| CN118230188A (en) * | 2024-04-12 | 2024-06-21 | 国家基础地理信息中心 | Cultivated land type change detection method, cultivated land type change detection device, electronic equipment and readable storage medium |
-
2022
- 2022-10-28 CN CN202211338255.XA patent/CN115861790A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118135311A (en) * | 2024-03-13 | 2024-06-04 | 南京北斗创新应用科技研究院有限公司 | Heterogeneous time sequence image wetland monitoring method and device based on improved cascade forests |
| CN117975292A (en) * | 2024-03-25 | 2024-05-03 | 中国科学院东北地理与农业生态研究所 | Extraction method for temperate forest swamps |
| CN118230188A (en) * | 2024-04-12 | 2024-06-21 | 国家基础地理信息中心 | Cultivated land type change detection method, cultivated land type change detection device, electronic equipment and readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Kukreja et al. | A Deep Neural Network based disease detection scheme for Citrus fruits | |
| Alvarez et al. | Body condition estimation on cows from depth images using Convolutional Neural Networks | |
| CN113392775B (en) | Sugarcane seedling automatic identification and counting method based on deep neural network | |
| CN115861790A (en) | Cultivated land remote sensing image analysis method, device, equipment, storage medium and product | |
| Liu et al. | A vision-based robust grape berry counting algorithm for fast calibration-free bunch weight estimation in the field | |
| CN113919442B (en) | Tobacco maturity state identification method based on convolutional neural network | |
| CN112365497A (en) | High-speed target detection method and system based on Trident Net and Cascade-RCNN structures | |
| Dhiman et al. | A general purpose multi-fruit system for assessing the quality of fruits with the application of recurrent neural network | |
| CN112819821A (en) | Cell nucleus image detection method | |
| CN110852358A (en) | Vehicle type distinguishing method based on deep learning | |
| CN116523352B (en) | Forest resource information management method and system | |
| CN110503140A (en) | Classification method based on depth migration study and neighborhood noise reduction | |
| CN114782837B (en) | Plant estimation method, plant estimation device, electronic equipment and storage medium | |
| CN115908590A (en) | Data intelligent acquisition method and system based on artificial intelligence | |
| Ferrer-Ferrer et al. | Simultaneous fruit detection and size estimation using multitask deep neural networks | |
| Asriny et al. | Transfer learning VGG16 for classification orange fruit images | |
| Heryanto et al. | Classification of Coffee Beans Defect Using Mask Region-based Convolutional Neural Network | |
| CN112465821A (en) | Multi-scale pest image detection method based on boundary key point perception | |
| CN112184627A (en) | Citrus fresh-keeping quality detection method based on image processing and neural network and application | |
| CN116188834B (en) | Full-slice image classification method and device based on self-adaptive training model | |
| Patel et al. | An intelligent grading system for automated identification and classification of banana fruit diseases using deep neural network | |
| CN114067314B (en) | Neural network-based peanut mildew identification method and system | |
| CN109376619A (en) | A kind of cell detection method | |
| Mirra et al. | Classification of Fruits using Deep Learning Algorithms | |
| Liu et al. | Peaks fusion assisted early-stopping strategy for overhead imagery segmentation with noisy labels |
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 |