Detailed Description
In order to make the objects, technical solutions and advantages of the present application more clearly understood, the present application is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of and not restrictive on the broad application.
The feature identification method provided by the embodiment of the application can be applied to the application environment shown in fig. 1. Fig. 1 provides a computer device, which may be a server, and its internal structure diagram may be as shown in fig. 1. The computer device includes a processor, a memory, and a network interface connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The database of the computer device is used for storing image data. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a method of feature recognition.
In one embodiment, as shown in fig. 2, a method for recognizing a ground object is provided, which is exemplified by the method applied to the computer device in fig. 1, and includes the following steps:
s201, acquiring homogeneity information and heterogeneity information of each sub-pixel in the initial remote sensing image according to the initial remote sensing image.
The remote sensing image is a picture for recording the size of electromagnetic waves of various ground objects, and the remote sensing image can comprise information of various ground objects; the sub-pixel refers to a minimum imaging unit in the remote sensing image; the homogeneity of each sub-pixel in the remote sensing image refers to the space attraction among particles of the same ground feature type, namely the property that the particles are of the same ground feature type, and the heterogeneity of each sub-pixel in the remote sensing image is used for describing the correlation relationship existing when the neighborhood pixels and the sub-pixels are of different types.
In general, the homogeneity information of the sub-pixels determines to some extent the ground object class of the sub-pixels. In this embodiment, homogeneity information of the sub-pixels and the neighborhood pixels can be obtained by using the homogeneity property between the sub-pixels and the neighborhood pixels, but since the homogeneity of different types of the sub-pixels may be the same, the determination of the sub-pixel type is affected, and therefore the determination of the sub-pixel type can be performed by combining the heterogeneity of the correlation relationship which describes the different types of the neighborhood pixels and the sub-pixels. In the embodiment, the attribute relationship between the sub-pixel and the neighborhood pixel is described together by introducing the homogeneity information and the heterogeneity information of the sub-pixel, so that the influence of the whole neighborhood pixel on the sub-pixel is quantized by the description, and the spatial association of part of the ground objects of the neighborhood pixel on the sub-pixel is avoided, so that the attribute characteristics of the sub-pixel are more obvious, the ground object attribute of the sub-pixel is better analyzed, and the positioning precision is improved.
Optionally, in this embodiment, the homogeneity information of the sub-pixels in the initial remote sensing image may include a homogeneity size between the sub-pixels and the neighborhood pixels, and the heterogeneity information may include a heterogeneity size between the sub-pixels and the neighborhood pixels. Optionally, the same sub-pixel may have a plurality of neighborhood pixels, and an average value of the homogeneity size between the sub-pixel and each neighborhood pixel may be used as the homogeneity information of the sub-pixel and the neighborhood pixel, or the maximum homogeneity size between the sub-pixel and each neighborhood pixel may be used as the homogeneity information of the sub-pixel and the neighborhood pixel; similarly, the average value of the heterogeneity between the sub-pixel and each neighborhood pixel may be used as the heterogeneity information of the sub-pixel and the neighborhood pixel, or the maximum heterogeneity between the sub-pixel and each neighborhood pixel may be used as the heterogeneity information of the sub-pixel and the neighborhood pixel.
In this embodiment, the homogeneity of each sub-pixel in the initial remote sensing image when the sub-pixels are of different ground object types may be sequentially calculated to obtain a first K column vector, where K is the number of ground object types, and the greater the homogeneity of a certain class of sub-pixels is, the more likely the sub-pixels belong to the class. And calculating the heterogeneity of the sub-pixels in different ground object types in sequence to obtain a second K-column vector.
It can be understood that the hyperspectral image is used as one of the remote sensing images, the initial remote sensing image in this embodiment may be a hyperspectral image, or the initial remote sensing image in this embodiment may also be another type of remote sensing image.
S202, obtaining a sub-pixel remote sensing image corresponding to the initial remote sensing image according to the initial remote sensing image; and the resolution ratio of the sub-pixel remote sensing image is higher than that of the initial remote sensing image.
Optionally, in this embodiment, the initial remote sensing image may be processed by using a ground feature distribution condition in the initial remote sensing image to obtain a sub-pixel remote sensing image corresponding to the remote sensing image, for example, the initial remote sensing image may be interpolated to obtain the sub-pixel remote sensing image. It can be understood that the resolution of the sub-pixel remote sensing image is higher than that of the original remote sensing image, and the higher the resolution of the remote sensing image is, the more details can be shown, that is, the sub-pixel remote sensing image includes more abundant spectral information.
And S203, obtaining the recognition result of various ground objects in the initial remote sensing image according to the sub-pixel remote sensing image, the homogeneity information, the heterogeneity information and the preset positioning model.
In this embodiment, the preset positioning model is a pre-trained model, and the positioning model can identify various ground features in the initial remote sensing image to obtain an identification result of various ground features in the initial remote sensing image.
Optionally, in this embodiment, the preset positioning model may include a plurality of sub models, the operation executed by each sub model may be different, and the positioning model including the plurality of sub models may be trained when the positioning model is trained, so that the positioning model may recognize the feature in the initial remote sensing image according to the sub-pixel remote sensing image, the homogeneity information of each sub-pixel in the initial remote sensing image, and the heterogeneity information of each sub-pixel. Optionally, in this embodiment, when the positioning model is trained, 25% of data in various ground feature data may be respectively selected as training samples to be trained, so as to obtain the trained positioning model, and store the weight of the trained positioning model.
In the ground object identification method, the identification results of various ground objects in the initial remote sensing image can be obtained according to the sub-pixel remote sensing image, the homogeneity information of each sub-pixel in the initial remote sensing image, the heterogeneity information of each sub-pixel in the initial remote sensing image and a preset positioning model by obtaining the homogeneity information and the heterogeneity information of each sub-pixel in the initial remote sensing image and the sub-pixel remote sensing image of the initial remote sensing image. Because the resolution ratio of the sub-pixel remote sensing image is higher than that of the initial remote sensing image, the spectral information in the sub-pixel remote sensing image can be fully learned through the positioning model, and various ground features in the initial remote sensing image can be accurately identified by combining the homogeneity information and the heterogeneity information of each pixel in the initial remote sensing image, so that the accuracy of the identification result of various ground features in the initial remote sensing image is improved.
In the scene of obtaining the recognition result of various ground objects in the initial remote sensing image according to the sub-pixel remote sensing image, the homogeneity information of each sub-pixel in the initial remote sensing image, the heterogeneity information of each sub-pixel in the initial remote sensing image and a preset positioning model, the positioning model can comprise a first back propagation neural network, a second back propagation neural network, a long-short term memory network and a classification network. In one embodiment, as shown in fig. 3, S203 includes:
and S301, inputting the homogeneity information into the first back propagation neural network to obtain the homogeneity characteristic.
Among them, the Back Propagation (BP) neural network is a multi-layer feedforward neural network trained according to an error back propagation algorithm, which uses a gradient search technique to minimize the mean square error between the actual output value and the expected output value of the network. In this embodiment, homogeneity information of each sub-pixel in the initial remote sensing image is input into the first BP neural network, and the homogeneity information of each sub-pixel is processed by the first BP neural network, so that homogeneity characteristics corresponding to each sub-pixel are obtained.
Optionally, in this embodiment, the first BP neural network model may include an input layer, a hidden layer, and an output layer, where the input layer may include K neurons, where K is the number of sub-pixels in the initial remote sensing image, the input layer is used to input homogeneity information of each sub-pixel in the initial remote sensing image, and the homogeneity characteristic of a high dimension is obtained in the output layer through processing by the hidden layer.
And S302, inputting the heterogeneity information into a second back propagation neural network to obtain heterogeneity characteristics.
In this embodiment, the heterogeneity information of each sub-pixel in the initial remote sensing image is input into the second BP neural network, and the heterogeneity information of each sub-pixel is processed by the second BP neural network, so as to obtain the heterogeneity characteristic corresponding to each sub-pixel.
Optionally, in this embodiment, the second BP neural network model may include an input layer, a hidden layer, and an output layer, where the input layer may include K neurons, where K is the number of sub-pixels in the initial remote sensing image, the input layer is used to input heterogeneity information of each sub-pixel in the initial remote sensing image, and high-dimensional heterogeneity characteristics are obtained in the output layer through processing by the hidden layer.
And S303, inputting the sub-pixel remote sensing image into a long-term and short-term memory network to obtain spectral characteristic information corresponding to the sub-pixel remote sensing image.
Among them, the Long Short-Term Memory network (LSTM) is a time-cycle neural network suitable for processing and predicting important events with very Long interval and delay in time sequence.
In this embodiment, a double-layer LSTM network may be used, 128 hidden nodes may be set, and spectral feature information corresponding to the sub-pixel remote sensing image is fully extracted through the double-layer LSTM network.
And S304, inputting the homogeneity characteristic, the heterogeneity characteristic and the spectral characteristic information into a classification network to obtain an identification result.
Optionally, the classification network may be a model with a classification function, and in this embodiment, a Softmax classifier may be used, which is a generalized generalization of a logistic regression classifier facing multiple classifications.
Exemplarily, in this embodiment, as shown in fig. 4, the first back propagation neural network, the second back propagation neural network, and the long-term and short-term memory network are three sub-models of the positioning model, further, the homogeneity characteristic, the heterogeneity characteristic, and the spectral characteristic obtained from the three sub-models may be input into the classification network of the positioning model, and the spectral characteristic, the homogeneity characteristic, and the heterogeneity characteristic are further subjected to fusion classification by the classification network, so as to obtain the recognition result of various types of ground features in the initial remote sensing image.
Optionally, as an optional implementation, the classification network of the positioning model may include a feature fusion layer and a classification layer; the homogeneity characteristic, the heterogeneity characteristic and the spectral characteristic information can be subjected to characteristic fusion through the characteristic fusion layer to obtain fused characteristics; and then classifying the fused features by utilizing a classification layer to obtain the recognition results of various ground objects in the initial remote sensing image. The feature fusion is to extract different feature vectors from the same mode to perform optimized combination, and has serial and parallel processing modes.
Optionally, in this embodiment, indian Pines data may be used to verify the ground feature identification method, where the size of an image in the Indian Pines data is 145 × 145 × 200, the data is composed of 16 ground feature categories (excluding a background), a low-resolution image is generated by performing downsampling on the data, a scaling factor s =2 is set, the data is classified by a svm classifier, and sub-pixel positioning is performed by the SPSAM algorithm, the STHSPM algorithm, and the ground feature identification method of this application, in the ground feature identification result, mutual influence among the categories of the SPSAM algorithm is severe, sub-pixel positioning of different analogs is not accurate, obvious burrs occur at the edge of the image, obvious fusion occurs on ground features close to the image, the STHSPM algorithm is affected by the background, which causes inaccuracy of sub-pixel positioning, obvious burrs occur at the edge of the image, obvious fusion occurs on ground features close to the image edge, and the positioning effect of this application between the categories and the background is significantly improved, and burrs and fusion phenomena and classification of ground features close to reality are almost close to classification. When the scaling factors s =2 and s =3, the Accuracy and the Kappa coefficient corresponding to the three methods are shown in the following table, where Kappa is an index for measuring the classification Accuracy, and Accuracy is the Accuracy, so it can be seen that the feature identification method provided by the present application is superior to the other two methods in terms of both the identification Accuracy and the identification Accuracy.
In the embodiment, the spectral characteristic information corresponding to the sub-pixel remote sensing image can be obtained by inputting the sub-pixel remote sensing image into the long-term and short-term memory network, so that rich spectral information of the remote sensing image can be integrated into the process of identifying the ground features, the value of the remote sensing image is fully exerted, the long-term and short-term memory network is used, the spectral information is considered, the mutual relation among wave bands is fully considered, and the obtained spectral characteristic information is more accurate; in addition, homogeneity information is input into the first back propagation neural network, heterogeneity information is input into the second back propagation neural network, corresponding homogeneity characteristics and heterogeneity characteristics can be obtained, and therefore spectral characteristic information of the remote sensing image and homogeneity and heterogeneity of sub-pixels in the remote sensing image are combined, spectral information and spatial structure information can be fully utilized, and accuracy of ground feature identification is improved.
In the scene of acquiring the homogeneity information and the heterogeneity information of each sub-pixel in the initial remote sensing image, in an embodiment, as shown in fig. 5, the step S201 includes:
s401, carrying out spectral decomposition on the initial remote sensing image to obtain a score map of the initial remote sensing image; the score map comprises the proportion of various ground features in the initial remote sensing image.
The spectrum decomposition is a mixed spectrum decomposition technology used for determining the proportion of different surface feature spectrum components or unknown components in the same pixel, the mixture of different surface feature spectrum components can change the depth of a wave band, the position, the width, the area, the absorption degree and the like of the wave band, and the mixed spectrum decomposition technology generally adopts a matrix equation, a neuron network method, a spectrum absorption index technology and the like to calculate the proportion of each component spectrum in a given pixel.
In this embodiment, the initial remote sensing image with low resolution is used for spectral decomposition to obtain a score map of various types of ground objects in the initial remote sensing image, where the score map includes the proportions of the various types of ground objects in the initial remote sensing image.
And S402, acquiring homogeneity information and heterogeneity information of each sub-pixel in the initial remote sensing image according to the initial remote sensing image and the score map.
Optionally, in this embodiment, each sub-pixel in the initial remote sensing image may be respectively calculated according to a spatial correlation theory, a law of universal gravitation, and a score map, so as to obtain homogeneity information and heterogeneity information of each sub-pixel in the initial remote sensing image. According to the theory of spatial correlation, under the premise that the dimension of a spatial variable is larger than that of pixels of the remote sensing image, the more adjacent targets in space have similar attribute values, and in the remote sensing image, the more exact is the spatial autocorrelation among the pixels, namely, sub-pixels which are closer in distance and among different pixels are more likely to belong to the same ground object type than sub-pixels which are relatively farther in distance. Illustratively, assuming spatial autocorrelation exists between sub-pixels and their neighboring eight mixed pixels, the correlation with the farther-distant mixed pixels is negligible.
In addition, in the embodiment, a universal gravitation theory is introduced, and it can be known that any two objects in nature are mutually attracted through the universal gravitation theory, the size of the gravitation is proportional to the product of the masses of the two objects and inversely proportional to the square of the distance between the two objects, and the universal gravitation law accurately describes the mode of mutual attraction between the objects. For example, in this embodiment, each pixel and sub-pixel in the initial remote sensing image can be regarded as particles with different masses, so that the fractional image of the mixed pixel can be represented as the mass of the particles, and the spatial attraction between the sub-pixel in the mixed pixel and the surrounding neighborhood pixels can be quantitatively represented.
Optionally, in this embodiment, the inter-particle homogeneity is described by the spatial attraction between particles of the same feature type, that is, the particles are of the same feature type and are represented by the spatial attraction. The homogeneity of the sub-pel and the neighborhood pel may be expressed as:
wherein p is m And p n Is a neighborhood pixel, Z (p) m ) Is the percentage content of class Z in the central mixed pixel, Z (p) n ) Is the percentage content of class Z in the central mixed pixel, s is the scaling factor, Z (p) n ) Is the percentage content of class Z in the neighborhood pixels, R in Is a sub-pixel and a neighborhood pixel p m Distance of center, Z (w) in ) Is a sub-pixel p n And the homogeneity of the Z-type ground objects is equal to that of the adjacent mixed image elements. In order to qualitatively describe the homogeneity of the sub-pixel and the mixed pixel, the formula (1) is simplified to obtain:
for example, in this embodiment, in the case that the sub-pel has 8 neighborhood pels, the homogeneity property that the sub-pel is a Z feature under the influence of all the neighborhood pels can be expressed as:
wherein, Z (w)
i ) Is a homogeneous property of the sub-pel.
Optionally, in this embodiment, an accidental phenomenon exists in the homogeneity of the sub-pixels, and when the score maps are different, the situation that the homogeneity of different categories of the sub-pixels is the same may still occur, which affects the determination on the category of the sub-pixels. Therefore, heterogeneity is proposed to describe the correlation between neighborhood pixels and sub-pixels in different categories. The heterogeneity of sub-pel and neighborhood pel may be expressed as:
wherein, Z * (w in ) Then it means that the sub-pixel isAnd when the ground objects are Z-type, the correlation properties between other ground object types and the sub-pixels of adjacent mixed pixels.
Exemplarily, in this embodiment, in the case that the sub-pixel has 8 neighborhood pixels, the heterogeneous property that the sub-pixel is a Z ground object under the influence of all neighborhood pixels can be expressed as:
wherein, Z (w)
i ) Is a heterogeneous property of the sub-pixel.
In the embodiment, the initial remote sensing image is subjected to spectral decomposition to obtain a score map of the initial remote sensing image, homogeneity and heterogeneity of the sub-pixels are introduced by utilizing a spatial correlation theory, a universal gravitation law and the score map, the attribute relation between the sub-pixels and the neighborhood pixels is described together, and the influence of the neighborhood pixels on the sub-pixels is quantized, so that the attribute characteristics of the sub-pixels are more obvious, the attribute characteristics of the sub-pixels are better analyzed, and the accuracy of ground object identification is obtained.
In the above scene where the sub-pixel remote sensing image corresponding to the initial remote sensing image is obtained according to the initial remote sensing image, in an embodiment, S202 includes: and processing the initial remote sensing image by using the double cubic interpolation value to obtain the sub-pixel remote sensing image.
Among them, double cubic interpolation is the most commonly used interpolation method in two-dimensional space, which is a method for "interpolating" or increasing the number/density of "pixels" in an image. The graphic data is added by using an interpolation technology, and when the graphic data is output in other forms, the resolution of the image can be increased.
In this embodiment, the original remote sensing image is processed by using bi-cubic interpolation, so that the resolution of the remote sensing image can be increased, and the sub-pixel remote sensing image can be obtained. For example, in this embodiment, taking the initial remote sensing image as the hyperspectral image as an example, the initial low-resolution hyperspectral image may be processed by using bicubic interpolation to obtain the high-resolution hyperspectral image.
In the embodiment, the initial remote sensing image is processed by using the bicubic interpolation, so that better detail quality can be kept, and the sub-pixel remote sensing image with smoother image edge is obtained, so that a ground feature identification result with higher accuracy is obtained.
Embodiments of the present disclosure are described below with reference to a specific feature recognition scenario, as shown in fig. 6, the method includes the following steps:
s1, carrying out spectral decomposition on the initial remote sensing image to obtain a score map of the initial remote sensing image; the score map comprises the proportion of various ground features in the initial remote sensing image.
And S2, respectively calculating each sub-pixel in the initial remote sensing image according to a spatial correlation theory, a universal gravitation law and a score map to obtain each homogeneity information and each heterogeneity information.
S3, processing the initial remote sensing image by using a double cubic interpolation value to obtain a sub-pixel remote sensing image; and the resolution ratio of the sub-pixel remote sensing image is higher than that of the initial remote sensing image.
S4, inputting the homogeneity information into a first back propagation neural network of a preset positioning model to obtain homogeneity characteristics; and inputting the heterogeneity information into a second back propagation neural network of a preset positioning model to obtain heterogeneity characteristics.
And S5, inputting the sub-pixel remote sensing image into a long-short term memory network of a preset positioning model to obtain spectral characteristic information corresponding to the sub-pixel remote sensing image.
S6, fusing the homogeneity characteristic, the heterogeneity characteristic and the spectral characteristic information through a preset characteristic fusion layer of the positioning model to obtain fused characteristics;
and S7, classifying the fused features by utilizing a preset classification layer of the positioning model to obtain a recognition result of the ground objects in the initial remote sensing image.
It should be understood that, although the steps in the flowcharts related to the embodiments as described above are sequentially displayed as indicated by arrows, the steps are not necessarily performed sequentially as indicated by the arrows. The steps are not limited to being performed in the exact order illustrated and, unless explicitly stated herein, may be performed in other orders. Moreover, at least a part of the steps in the flowcharts related to the embodiments described above may include multiple steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, and the execution order of the steps or stages is not necessarily sequential, but may be rotated or alternated with other steps or at least a part of the steps or stages in other steps.
Based on the same inventive concept, the embodiment of the application also provides a feature identification device for realizing the feature identification method. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the method, so the specific limitations in one or more embodiments of the feature recognition device provided below can be referred to the limitations on the feature recognition method in the foregoing, and details are not repeated here.
In one embodiment, as shown in fig. 7, there is provided a ground recognition apparatus including: first acquisition module, second acquisition module and identification module, wherein:
the first acquisition module is used for acquiring homogeneity information and heterogeneity information of each sub-pixel in the initial remote sensing image according to the initial remote sensing image;
the second acquisition module is used for acquiring the sub-pixel remote sensing image corresponding to the initial remote sensing image according to the initial remote sensing image; the resolution ratio of the sub-pixel remote sensing image is higher than that of the initial remote sensing image;
and the identification module is used for obtaining the identification results of various ground objects in the initial remote sensing image according to the sub-pixel remote sensing image, the homogeneity information, the heterogeneity information and the preset positioning model.
The feature recognition apparatus provided in this embodiment may implement the method embodiments, and the implementation principle and the technical effect are similar, which are not described herein again.
In one embodiment, the identification module comprises: a first acquisition unit, a second acquisition unit, a third acquisition unit, and a fourth acquisition unit, wherein:
and the first acquisition unit is used for inputting the homogeneity information into the first back propagation neural network to obtain the homogeneity characteristic.
And the second acquisition unit is used for inputting the heterogeneity information into the second back propagation neural network to obtain the heterogeneity characteristics.
And the third acquisition unit is used for inputting the sub-pixel remote sensing image into the long-term and short-term memory network to obtain the spectral characteristic information corresponding to the sub-pixel remote sensing image.
And the fourth acquisition unit is used for inputting the homogeneity characteristic, the heterogeneity characteristic and the spectral characteristic information into the classification network to obtain an identification result.
The feature recognition device provided in this embodiment may implement the method embodiments, and the implementation principle and technical effects are similar, which are not described herein again.
In an embodiment, the fourth obtaining unit is configured to fuse the homogeneity characteristic, the heterogeneity characteristic, and the spectral characteristic information through the characteristic fusion layer to obtain a fused characteristic; and classifying the fused features by utilizing a classification layer to obtain an identification result.
The feature recognition apparatus provided in this embodiment may implement the method embodiments, and the implementation principle and the technical effect are similar, which are not described herein again.
In one embodiment, the first obtaining module includes: a fifth acquisition unit and a sixth acquisition unit, wherein:
the fifth acquisition unit is used for carrying out spectral decomposition on the initial remote sensing image to obtain a score map of the initial remote sensing image; the score map comprises the proportion of various ground features in the initial remote sensing image.
And the sixth acquisition unit is used for acquiring the homogeneity information and the heterogeneity information of each sub-pixel in the initial remote sensing image according to the initial remote sensing image and the score map.
The feature recognition device provided in this embodiment may implement the method embodiments, and the implementation principle and technical effects are similar, which are not described herein again.
In an embodiment, the sixth obtaining unit is configured to calculate, according to a spatial correlation theory, a law of universal gravitation, and a score map, each sub-pixel in the initial remote sensing image to obtain each homogeneity information and each heterogeneity information.
The feature recognition device provided in this embodiment may implement the method embodiments, and the implementation principle and technical effects are similar, which are not described herein again.
In an embodiment, the second obtaining module includes a seventh obtaining unit, wherein:
and the seventh acquisition unit is used for processing the initial remote sensing image by utilizing double cubic interpolation to obtain the sub-pixel remote sensing image.
The feature recognition apparatus provided in this embodiment may implement the method embodiments, and the implementation principle and the technical effect are similar, which are not described herein again.
The modules in the above feature recognition device may be implemented in whole or in part by software, hardware, and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as shown in fig. 8. The computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The communication interface of the computer device is used for carrying out wired or wireless communication with an external terminal, and the wireless communication can be realized through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. The computer program is executed by a processor to implement a method of feature recognition. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, a key, a track ball or a touch pad arranged on the shell of the computer equipment, an external keyboard, a touch pad or a mouse and the like.
Those skilled in the art will appreciate that the architecture shown in fig. 8 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.
In one embodiment, a computer device is provided, comprising a memory having a computer program stored therein and a processor that when executing the computer program performs the steps of:
acquiring homogeneity information and heterogeneity information of each sub-pixel in the initial remote sensing image according to the initial remote sensing image;
acquiring a sub-pixel remote sensing image corresponding to the initial remote sensing image according to the initial remote sensing image; the resolution ratio of the sub-pixel remote sensing image is higher than that of the initial remote sensing image;
and obtaining the recognition result of various ground objects in the initial remote sensing image according to the sub-pixel remote sensing image, the homogeneity information, the heterogeneity information and the preset positioning model.
In one embodiment, the processor, when executing the computer program, further performs the steps of:
inputting the homogeneity information into a first back propagation neural network to obtain homogeneity characteristics;
inputting the heterogeneity information into a second back propagation neural network to obtain heterogeneity characteristics;
inputting the sub-pixel remote sensing image into a long-term and short-term memory network to obtain spectral characteristic information corresponding to the sub-pixel remote sensing image;
and inputting the homogeneity characteristic, the heterogeneity characteristic and the spectral characteristic information into a classification network to obtain an identification result.
In one embodiment, the processor, when executing the computer program, further performs the steps of:
fusing the homogeneity characteristic, the heterogeneity characteristic and the spectral characteristic information through the characteristic fusion layer to obtain fused characteristics;
and classifying the fused features by utilizing a classification layer to obtain an identification result.
In one embodiment, the processor when executing the computer program further performs the steps of:
carrying out spectral decomposition on the initial remote sensing image to obtain a score map of the initial remote sensing image; the score map comprises the proportion of various ground features in the initial remote sensing image;
and acquiring homogeneity information and heterogeneity information of each sub-pixel in the initial remote sensing image according to the initial remote sensing image and the score map.
In one embodiment, the processor, when executing the computer program, further performs the steps of:
and respectively calculating each sub-pixel in the initial remote sensing image according to a spatial correlation theory, the law of universal gravitation and the score map to obtain each homogeneity information and each heterogeneity information.
In one embodiment, the processor when executing the computer program further performs the steps of:
and processing the initial remote sensing image by using the double cubic interpolation value to obtain the sub-pixel remote sensing image.
In one embodiment, a computer-readable storage medium is provided, having a computer program stored thereon, which when executed by a processor, performs the steps of:
according to the initial remote sensing image, acquiring homogeneity information and heterogeneity information of each sub-pixel in the initial remote sensing image;
acquiring a sub-pixel remote sensing image corresponding to the initial remote sensing image according to the initial remote sensing image; the resolution ratio of the sub-pixel remote sensing image is higher than that of the initial remote sensing image;
and obtaining the recognition result of various ground objects in the initial remote sensing image according to the sub-pixel remote sensing image, the homogeneity information, the heterogeneity information and the preset positioning model.
In one embodiment, the computer program when executed by the processor further performs the steps of:
inputting the homogeneity information into a first back propagation neural network to obtain homogeneity characteristics;
inputting the heterogeneity information into a second back propagation neural network to obtain heterogeneity characteristics;
inputting the sub-pixel remote sensing image into a long-term and short-term memory network to obtain spectral characteristic information corresponding to the sub-pixel remote sensing image;
and inputting the homogeneity characteristic, the heterogeneity characteristic and the spectral characteristic information into a classification network to obtain an identification result.
In one embodiment, the computer program when executed by the processor further performs the steps of:
fusing the homogeneity characteristic, the heterogeneity characteristic and the spectral characteristic information through the characteristic fusion layer to obtain fused characteristics;
and classifying the fused features by utilizing a classification layer to obtain an identification result.
In one embodiment, the computer program when executed by the processor further performs the steps of:
carrying out spectral decomposition on the initial remote sensing image to obtain a score map of the initial remote sensing image; the score map comprises the proportion of various ground features in the initial remote sensing image;
and acquiring homogeneity information and heterogeneity information of each sub-pixel in the initial remote sensing image according to the initial remote sensing image and the score map.
In one embodiment, the computer program when executed by the processor further performs the steps of:
and respectively calculating each sub-pixel in the initial remote sensing image according to a spatial correlation theory, the law of universal gravitation and the score map to obtain each homogeneity information and each heterogeneity information.
In one embodiment, the computer program when executed by the processor further performs the steps of:
and processing the initial remote sensing image by using the double cubic interpolation value to obtain the sub-pixel remote sensing image.
In one embodiment, a computer program product is provided, comprising a computer program which, when executed by a processor, performs the steps of:
according to the initial remote sensing image, acquiring homogeneity information and heterogeneity information of each sub-pixel in the initial remote sensing image;
acquiring a sub-pixel remote sensing image corresponding to the initial remote sensing image according to the initial remote sensing image; the resolution ratio of the sub-pixel remote sensing image is higher than that of the initial remote sensing image;
and obtaining the recognition result of various ground objects in the initial remote sensing image according to the sub-pixel remote sensing image, the homogeneity information, the heterogeneity information and the preset positioning model.
In one embodiment, the computer program when executed by the processor further performs the steps of:
inputting the homogeneity information into a first back propagation neural network to obtain homogeneity characteristics;
inputting the heterogeneity information into a second back propagation neural network to obtain heterogeneity characteristics;
inputting the sub-pixel remote sensing image into a long-term and short-term memory network to obtain spectral characteristic information corresponding to the sub-pixel remote sensing image;
and inputting the homogeneity characteristic, the heterogeneity characteristic and the spectral characteristic information into a classification network to obtain an identification result.
In one embodiment, the computer program when executed by the processor further performs the steps of:
fusing the homogeneity characteristic, the heterogeneity characteristic and the spectral characteristic information through the characteristic fusion layer to obtain fused characteristics;
and classifying the fused features by utilizing a classification layer to obtain an identification result.
In one embodiment, the computer program when executed by the processor further performs the steps of:
carrying out spectral decomposition on the initial remote sensing image to obtain a score map of the initial remote sensing image; the score map comprises the proportion of various ground features in the initial remote sensing image;
and acquiring homogeneity information and heterogeneity information of each sub-pixel in the initial remote sensing image according to the initial remote sensing image and the score map.
In one embodiment, the computer program when executed by the processor further performs the steps of:
and respectively calculating each sub-pixel in the initial remote sensing image according to a spatial correlation theory, the law of universal gravitation and the score map to obtain each homogeneity information and each heterogeneity information.
In one embodiment, the computer program when executed by the processor further performs the steps of:
and processing the initial remote sensing image by using the double cubic interpolation value to obtain the sub-pixel remote sensing image.
It should be noted that, the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, presented data, etc.) referred to in the present application are information and data authorized by the user or sufficiently authorized by each party.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, database, or other medium used in the embodiments provided herein may include at least one of non-volatile and volatile memory. The nonvolatile Memory may include a Read-Only Memory (ROM), a magnetic tape, a floppy disk, a flash Memory, an optical Memory, a high-density embedded nonvolatile Memory, a resistive Random Access Memory (ReRAM), a Magnetic Random Access Memory (MRAM), a Ferroelectric Random Access Memory (FRAM), a Phase Change Memory (PCM), a graphene Memory, and the like. Volatile Memory can include Random Access Memory (RAM), external cache Memory, and the like. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM), among others. The databases referred to in various embodiments provided herein may include at least one of relational and non-relational databases. The non-relational database may include, but is not limited to, a block chain based distributed database, and the like. The processors referred to in the embodiments provided herein may be general purpose processors, central processing units, graphics processors, digital signal processors, programmable logic devices, quantum computing based data processing logic devices, etc., without limitation.
All possible combinations of the technical features in the above embodiments may not be described for the sake of brevity, but should be considered as being within the scope of the present disclosure as long as there is no contradiction between the combinations of the technical features.
The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the present application. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, and these are all within the scope of protection of the present application. Therefore, the protection scope of the present application shall be subject to the appended claims.