CN112434650A

CN112434650A - Multi-spectral image building change detection method and system

Info

Publication number: CN112434650A
Application number: CN202011427254.3A
Authority: CN
Inventors: 张龙; 陈卓; 黄远胜
Original assignee: Individual
Current assignee: Individual
Priority date: 2020-12-09
Filing date: 2020-12-09
Publication date: 2021-03-02

Abstract

The embodiment of the application provides a multispectral image building change detection method and system, reflection characteristic information of a currently acquired spectral object region is combined in a model training process, all spectral condition characteristics are further considered to learn similar access modes of spectral conditions at each time, in addition, noise labels in building change labels of the spectral conditions at each time are removed for model training, the condition that long-term influence is generated on subsequent spectral conditions due to accumulation of some sudden spectral conditions to lead to mistaken learning of the characteristics in the model training process can be avoided, label prediction accuracy is further improved, and the accuracy degree of an acquisition strategy of subsequent image acquisition equipment is further improved.

Description

Multi-spectral image building change detection method and system

Technical Field

The application relates to the technical field of building detection, in particular to a method and a system for detecting changes of a multi-spectral image building.

Background

The spectral reflection situation of building change relates to the collection surface of a building, so the accuracy degree of the collection strategy of subsequent image collection equipment is also related, the relative error between the building change label of each spectral condition and the building change label predicted by the building change label is large, and the inventor of the application finds that the spectral reflection change characteristics of the currently acquired spectral object region are not considered in the traditional spectral reflection scheme, and the long-term influence on the subsequent spectral conditions is possibly caused by the accumulation of some sudden spectral conditions, so that the characteristics are wrongly learned in the model training process, and the prediction accuracy is seriously reduced.

Disclosure of Invention

In view of the above, an object of the present application is to provide a method and a system for detecting a change in a multispectral image building, in which reflection feature information of a currently acquired spectral object region is combined in a model training process, and further, all spectral condition features are considered to learn a similar access mode of each spectral condition, and in addition, a noise tag is removed from a building change tag of each spectral condition to perform model training, so that a situation that long-term influence is generated on subsequent spectral conditions due to accumulation of some sudden spectral conditions, so that the features are wrongly learned in the model training process can be avoided, and therefore, accuracy of tag prediction is improved, and accuracy of an acquisition strategy of subsequent image acquisition equipment is improved.

According to a first aspect of the present application, there is provided a multispectral image building change detection method, applied to a server, the server being in communication with an image acquisition device, the method including:

extracting multispectral image simulation data corresponding to different spectral conditions, wherein the multispectral image simulation data comprise building change characteristic information extracted based on the spectral conditions and building change labels corresponding to the building change characteristic information, the building change characteristic information comprises spectral reflection change characteristics of the spectral conditions, the spectral reflection change characteristics comprise reflection characteristic information of a spectral object area reflected within a preset time period and spectral condition characteristics corresponding to the spectral conditions, and the building change labels are residual label information in virtual change labels after noise labels are removed;

training a convolutional neural network model according to the multispectral image simulation data corresponding to the different spectral conditions to obtain a multispectral image building change detection model;

when the spectral image online data corresponding to the spectral conditions are obtained from each image acquisition device, predicting the building change label of the spectral image online data of each spectral condition according to the multispectral image building change detection model, and updating the acquisition strategy of each image acquisition device according to the building change label of each spectral condition obtained through prediction.

In one possible implementation manner of the first aspect, the step of extracting multispectral image simulation data corresponding to different spectral conditions includes:

acquiring spectral condition configuration information from the spectral conditions, and extracting spectral condition characteristic information corresponding to different spectral condition items in the spectral condition configuration information;

searching reflection characteristic information of a spectrum object region reflected within a preset time period and spectrum condition characteristics corresponding to the spectrum condition from the spectrum condition characteristic information, wherein the reflection characteristic information of the spectrum object region is cached through an LRU stack structure;

and performing virtual rendering processing according to the spectral condition characteristic information corresponding to the different spectral condition items to obtain virtual change labels, and removing noise labels from the virtual change labels to obtain the building change labels of the building change characteristic information.

In a possible implementation manner of the first aspect, the step of performing virtual rendering processing according to the spectral condition feature information corresponding to the different spectral condition items to obtain a virtual change label includes:

inputting the spectral condition characteristic information corresponding to the different spectral condition items into a virtual rendering model of the corresponding spectral condition item, and acquiring virtual rendering characteristic information corresponding to the spectral condition characteristic information;

determining a first virtual rendering unit consisting of each virtual rendering container in the virtual rendering characteristic information and a virtual rendering container associated with the virtual rendering container, and determining global parameters of all virtual rendering container parameters in the first virtual rendering unit;

determining global parameters of all virtual rendering container parameters in the second virtual rendering unit and global parameters of all virtual rendering container parameters in the third virtual rendering unit; wherein the second virtual rendering unit is associated with the first virtual rendering unit and located at the same rendering service of the first virtual rendering unit, the third virtual rendering unit is associated with the first virtual rendering unit and located at a different rendering service of the first virtual rendering unit, and the first virtual rendering unit, the second virtual rendering unit, and the third virtual rendering unit contain the same number of virtual rendering containers;

calculating the correlation parameters of the global parameters of all the virtual rendering container parameters in the second virtual rendering unit and the global parameters of all the virtual rendering container parameters in the first virtual rendering unit, calculating the correlation parameters of the global parameters of all the virtual rendering container parameters in the third virtual rendering unit and the global parameters of all the virtual rendering container parameters in the first virtual rendering unit, and taking the calculated maximum correlation parameters as the target correlation parameters of the virtual rendering container;

determining a plurality of clustered virtual rendering units according to the target associated parameters of each virtual rendering container in the virtual rendering characteristic information, and obtaining the virtual change label according to the virtual rendering container global parameters of the clustered virtual rendering units.

In a possible implementation manner of the first aspect, the training a convolutional neural network model according to the multispectral image simulation data corresponding to the different spectral conditions to obtain a multispectral image building change detection model includes:

inputting the multispectral image simulation data into the convolutional neural network model, predicting spectral reflection change information of each spectral condition in a spectral reflection process through the convolutional neural network model, determining a label spectral reflection range corresponding to a preset test label interval according to the spectral reflection change information of each spectral condition, and acquiring all spectral reflection labels in the label spectral reflection range to obtain a spectral reflection label matching sequence of each spectral condition;

acquiring label classification information associated with each spectral condition according to the spectral reflection label matching sequence of each spectral condition, extracting label classification characteristic information from the label classification information of each spectral condition, and obtaining a logistic regression network corresponding to each label classification characteristic information according to the matching classification of the extracted label classification characteristic information in each spectral condition label classification information, wherein the label classification characteristic information comprises building change labels;

recording label classification characteristic information extracted from the label classification information fed back by each spectrum condition and a logistic regression network of the label classification characteristic information, and constructing a logistic regression analysis result of each spectrum condition;

according to the sequence of the level levels in the logistic regression analysis result from high to low, sequentially matching the label classification characteristic information with each spectral reflection label in the spectral reflection label matching sequence in the set range, and recording the matching result when any label classification characteristic information in the logistic regression analysis result is matched with the spectral reflection label in the spectral reflection label matching sequence under each spectral condition;

and training according to the matching result to obtain the multispectral image building change detection model.

In a possible implementation manner of the first aspect, the step of training the multispectral image building change detection model according to the matching result includes:

calculating an updating difference parameter of the label classification characteristic information according to the difference between the label classification characteristic information matched in the matching result and the theoretical label classification characteristic information, and determining an updating strategy of each label classification characteristic information according to the updating difference parameter;

extracting a plurality of selectable first updating modes and selectable updating contents of each first updating mode from the determined updating strategy of each label classification characteristic information;

screening a plurality of update patterns identical to a preset second update pattern from the plurality of selectable first update patterns as a plurality of third update patterns, wherein the second update pattern is a label update pattern output by a plurality of logistic regression analysis tree nodes in the convolutional neural network model, and the logistic regression analysis tree nodes comprise: the updating content is the updating content corresponding to the marking feature node, and the updating content is the updating content corresponding to the marking feature node;

inputting selectable update contents and a plurality of update contents of the plurality of third update modes into the correlation model of each spectral condition and the label classification characteristic information for calculation to obtain an update result, and fusing mode parameters of the plurality of selectable first update modes of the update strategy with the update result to obtain a model update parameter of each label classification characteristic information;

and updating the convolutional neural network model according to the model updating parameters of the classification characteristic information of each label, and training to obtain the multispectral image building change detection model.

In a possible implementation manner of the first aspect, the step of updating the acquisition strategy of each image acquisition device according to the building change label of each spectrum condition obtained by prediction includes:

determining an update template set of an acquisition strategy for each current image acquisition device according to the building change labels of each spectral condition obtained by prediction;

and updating the acquisition strategy according to the updating mode and the updating content of each updating template in the updating template set.

In a possible implementation manner of the first aspect, the step of updating the acquisition policy according to the update mode and the update content of each update template in the update template set respectively includes:

clustering the plurality of updating templates according to the updating content of each updating template to obtain a plurality of template clusters, wherein each template cluster corresponds to one updating mode;

generating an updating process corresponding to each updating template under the current template clustering aiming at each template clustering, classifying the updating templates with the same updating behavior and updating source in different updating processes into a class aiming at each template clustering, and fusing target updating processes of each updating process in the class of updating templates in the corresponding updating process when the ratio of the number of updating contents in the class of updating templates to the total number of the updating processes under the current template clustering exceeds a first threshold value to obtain a first target updating process;

or, the processes which appear in the belonged updating process only once and have the same updating mode and target updating processes in different updating processes are classified into one class, and when the ratio of the number of the updating contents in the class of updating templates to the total number of the updating processes under the current template clustering exceeds a first threshold value, the target updating processes in the belonged updating process of each process in the class of updating templates are fused to obtain a first target updating process;

or, grouping the update templates which only appear once in the update process and have the same update mode and target update process in different update processes into one class, and fusing the target update process in the update process of each update template in the class of update templates when the ratio of the number of the update templates in the class of update templates to the total number of the update processes under the current template clustering exceeds a first threshold value to obtain a first target update process;

determining a main updating process in the current template clustering according to the first target updating process, and determining other updating templates in the current template clustering as slave updating processes;

and respectively updating the acquisition strategies according to the updating sequence of the main updating process and the auxiliary updating process in the current template cluster.

According to a second aspect of the present application, there is provided a multispectral image building change detection system, applied to a server, the server being in communication with an image acquisition device, the system comprising:

the extraction module is used for extracting multispectral image simulation data corresponding to different spectral conditions, wherein the multispectral image simulation data comprise building change characteristic information extracted based on the spectral conditions and building change labels corresponding to the building change characteristic information, the building change characteristic information comprises spectral reflection change characteristics of the spectral conditions, the spectral reflection change characteristics comprise reflection characteristic information of a spectral object region reflected within a preset time period and spectral condition characteristics corresponding to the spectral conditions, and the building change labels are residual label information in virtual change labels after noise labels are removed;

the training module is used for training a convolutional neural network model according to the multispectral image simulation data corresponding to the different spectral conditions to obtain a multispectral image building change detection model;

and the prediction module is used for predicting the building change label of the spectral image online data under each spectral condition according to the multispectral image building change detection model when acquiring the spectral image online data under the corresponding spectral condition from each image acquisition device, and updating the acquisition strategy of each image acquisition device according to the building change label under each spectral condition obtained by prediction.

According to a third aspect of the present application, there is provided a server comprising a machine-readable storage medium having stored thereon machine-executable instructions and a processor which, when executing the machine-executable instructions, implements the aforementioned multi-spectral image building change detection method.

According to a fourth aspect of the present application, there is provided a computer-readable storage medium having stored therein computer-executable instructions that, when executed, implement the aforementioned multi-spectral image building change detection method.

Based on any one of the above aspects, the reflection characteristic information of the currently acquired spectral object region is combined in the model training process, all spectral condition characteristics are further considered to learn the similar access mode of each spectral condition, and in addition, the noise label in the building change label of each spectral condition is removed to perform model training, so that the situation that the characteristics are wrongly learned in the model training process due to long-term influence on subsequent spectral conditions caused by accumulation of some sudden spectral conditions can be avoided, the label prediction accuracy is improved, and the accuracy of the acquisition strategy of subsequent image acquisition equipment is improved.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are required to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present application and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained from the drawings without inventive effort.

Fig. 1 is a schematic view illustrating an application scenario of a multispectral image building change detection method provided by an embodiment of the present application;

FIG. 2 is a flow chart of a multispectral image building change detection method provided by an embodiment of the present application;

FIG. 3 is a functional block diagram of a multi-spectral image building change detection system provided by an embodiment of the present application;

fig. 4 is a schematic component structural diagram of a server for implementing the multispectral image building change detection method according to an embodiment of the present application.

Detailed Description

In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it should be understood that the drawings in the present application are for illustrative and descriptive purposes only and are not used to limit the scope of protection of the present application. Additionally, it should be understood that the schematic drawings are not necessarily drawn to scale. The flowcharts used in this application illustrate operations implemented according to some of the embodiments of the present application. It should be understood that the operations of the flow diagrams may be performed out of order, and steps without logical context may be performed in reverse order or simultaneously. One skilled in the art, under the guidance of this application, may add one or more other operations to, or remove one or more operations from, the flowchart.

In addition, the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. The components of the embodiments of the present application, generally described and illustrated in the figures herein, can be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the present application, presented in the accompanying drawings, is not intended to limit the scope of the claimed application, but is merely representative of selected embodiments of the application. All other embodiments, which can be derived by a person skilled in the art from the embodiments of the present application without making any creative effort, shall fall within the protection scope of the present application.

Fig. 1 shows a schematic application scenario diagram of an interactive system 10 provided in an embodiment of the present application. In this embodiment, the interactive system 10 may include a server 100 and an image capture device 200 communicatively coupled to the server 100. In other possible embodiments, the interactive system 10 may also include only some of the components shown in fig. 1 or may also include other components.

In some embodiments, the server 100 may be a single server or a group of servers. The set of servers may be centralized or distributed (e.g., server 100 may be a distributed system). In some embodiments, server 100 may be local or remote to image capture device 200. For example, the server 100 may access information stored in the image capture device 200 and a database, or any combination thereof, via a network. As another example, server 100 may be directly connected to at least one of image capture device 200 and a database to access information and/or data stored therein. In some embodiments, the server 100 may be implemented on a cloud platform; by way of example only, the cloud platform may include a private cloud, a public cloud, a hybrid cloud, a community cloud (community cloud), a distributed cloud, an inter-cloud, a multi-cloud, and the like, or any combination thereof.

In some embodiments, the server 100 may include a processor. The processor may process information and/or data related to the service request to perform one or more of the functions described herein. A processor may include one or more processing cores (e.g., a single-core processor (S) or a multi-core processor (S)).

The aforementioned database may store data and/or instructions. In some embodiments, the database may store data assigned to the image acquisition device 200. In some embodiments, the database may store data and/or instructions for the exemplary methods described herein. In some embodiments, the database may include mass storage, removable storage, volatile read-write memory, read-only memory, or the like, or any combination thereof.

In some embodiments, the database may be connected to a network to communicate with one or more components in the interactive system 10 (e.g., server 100, image capture device 200, etc.). One or more components in the interactive system 10 may access data or instructions stored in a database via a network. In some embodiments, the database may be directly connected to one or more components of the interactive system 10 (e.g., the server 100, the image capture device 200, etc.; or, in some embodiments, the database may be part of the server 100.

Fig. 2 is a schematic flow chart of a multispectral image building change detection method provided in an embodiment of the present application, which may be executed by the server shown in fig. 1. It should be understood that in other embodiments, the order of some steps in the multispectral image building change detection method of the present embodiment may be interchanged according to actual needs, or some steps may be omitted or deleted. The steps of the multispectral image building change detection method are described in detail below.

Step S110, extracting multispectral image simulation data corresponding to different spectral conditions, where the multispectral image simulation data includes building change feature information extracted based on the spectral conditions and building change tags corresponding to the building change feature information.

And step S120, training a convolutional neural network model according to the multispectral image simulation data corresponding to different spectral conditions to obtain a multispectral image building change detection model.

Step S130, when the spectral conditions are sent to each image capturing device 200, predicting the building change label of the spectral image online data of each spectral condition according to the multispectral image building change detection model, and updating the capturing policy of each image capturing device 200 according to the building change label of each spectral condition obtained by prediction.

In this embodiment, the building change characteristic information may include a spectral reflectance change characteristic of the spectral condition. For example, the spectral reflectance change characteristics may include reflectance characteristic information of a spectral object region accessed within a preset time period (such as within the last week) and spectral condition characteristics corresponding to the spectral condition, so that similar reflectance patterns of each spectral condition are learned by combining currently acquired reflectance characteristic information of the spectral object region in the model training process and further considering all spectral condition characteristics.

Correspondingly, the building change label is the residual label information of the virtual change label after the noise label is removed.

Based on the above steps, in the embodiment, the reflection feature information of the currently acquired spectral object region is combined in the model training process, and all spectral condition features are further considered to learn the similar access mode of each spectral condition, and in addition, the noise label is removed from the building change label of each spectral condition to perform model training, so that the situation that the features are wrongly learned in the model training process due to long-term influence on subsequent spectral conditions caused by accumulation of some sudden spectral conditions can be avoided, the label prediction accuracy is improved, and the accuracy of the acquisition strategy of subsequent image acquisition equipment is improved.

In a possible implementation manner, for step S110, the step S110 may be implemented by the following sub-steps:

and a substep S111 of obtaining spectral condition configuration information from the spectral conditions and extracting spectral condition feature information corresponding to different spectral condition items in the spectral condition configuration information.

In the substep S112, the reflection characteristic information of the spectral object region reflected within the preset time period and the spectral condition characteristic corresponding to the spectral condition are searched from the spectral condition characteristic information.

And a substep S113, performing virtual rendering processing according to the spectral condition characteristic information corresponding to different spectral condition items to obtain virtual change labels, and removing the noise labels from the virtual change labels to obtain the building change labels of the building change characteristic information.

For example, in order to accurately determine a building change label of building change feature information, the present embodiment may input spectral condition feature information corresponding to different spectral condition items into a virtual rendering model of corresponding spectral condition items, acquire virtual rendering feature information corresponding to the spectral condition feature information, then determine a first virtual rendering unit formed by each virtual rendering container in the virtual rendering feature information and a virtual rendering container associated with the virtual rendering container, and determine global parameters of all virtual rendering container parameters in the first virtual rendering unit.

Meanwhile, the global parameters of all the virtual rendering container parameters in the second virtual rendering unit and the global parameters of all the virtual rendering container parameters in the third virtual rendering unit are further determined. The second virtual rendering unit is associated with the first virtual rendering unit and located in the same rendering service of the first virtual rendering unit, the third virtual rendering unit is associated with the first virtual rendering unit and located in different rendering services of the first virtual rendering unit, and the first virtual rendering unit, the second virtual rendering unit and the third virtual rendering unit contain the same number of virtual rendering containers.

Then, the correlation parameters of the global parameters of all the virtual rendering container parameters in the second virtual rendering unit and the global parameters of all the virtual rendering container parameters in the first virtual rendering unit can be calculated, the correlation parameters of the global parameters of all the virtual rendering container parameters in the third virtual rendering unit and the global parameters of all the virtual rendering container parameters in the first virtual rendering unit can be calculated, and the maximum correlation parameter obtained through calculation is used as the target correlation parameter of the virtual rendering container. Therefore, a plurality of clustered virtual rendering units can be determined according to the target associated parameter of each virtual rendering container in the virtual rendering characteristic information, and the virtual change label can be obtained according to the global parameter of the virtual rendering container of the clustered virtual rendering units.

In a possible implementation manner, for step S120, in the process of training to obtain the multispectral image building change detection model, a convolutional neural network model may be used for model training to improve the logistic regression analysis accuracy of the multispectral image building change detection model. Referring to fig. 4, step S120 can be further implemented by the following sub-steps:

and a substep S121, inputting the multispectral image simulation data into a convolutional neural network model, predicting spectral reflection change information of each spectral condition in a spectral reflection process through the convolutional neural network model, determining a label spectral reflection range corresponding to a preset test label interval according to the spectral reflection change information of each spectral condition, and acquiring all spectral reflection labels in the label spectral reflection range to obtain a spectral reflection label matching sequence of each spectral condition.

And a substep S122, obtaining label classification information associated with each spectrum condition according to the spectrum reflection label matching sequence of each spectrum condition, extracting label classification characteristic information from the label classification information of each spectrum condition, and obtaining a logistic regression network corresponding to each label classification characteristic information according to the matching classification of the extracted label classification characteristic information in the label classification information of each spectrum condition. And the label classification characteristic information comprises the building change label obtained by prediction.

And a substep S123 of recording label classification characteristic information extracted from the label classification information fed back by each spectrum condition and a logistic regression network of the label classification characteristic information and constructing a logistic regression analysis result of each spectrum condition.

And a substep S124 of matching the label classification characteristic information with each spectral reflection label in the spectral reflection label matching sequence in the set range in sequence according to the sequence of the level levels from high to low in the logistic regression analysis result until the matching result is recorded when any label classification characteristic information in the logistic regression analysis result is matched with the spectral reflection label in the spectral reflection label matching sequence under each spectral condition.

And a substep S125 of training to obtain a multispectral image building change detection model according to the matching result.

For example, as a possible example, the present embodiment may calculate an update difference parameter of the label classification feature information according to a difference between the label classification feature information matched in the matching result and theoretical label classification feature information, determine an update policy of each label classification feature information according to the update difference parameter, and then extract a plurality of selectable first update patterns and selectable update contents of each first update pattern from the determined update policy of each label classification feature information.

Then, a plurality of update modes identical to the preset second update mode can be screened out from the plurality of selectable first update modes as a plurality of third update modes.

It should be noted that the second update pattern is a label update pattern output by a plurality of logistic regression analysis tree nodes in the convolutional neural network model, and the logistic regression analysis tree nodes may include: the updating content is the updating content corresponding to the marking feature node, and the updating content is the updating content corresponding to the marking feature node.

On the basis, selectable update contents and a plurality of update contents of a plurality of third update modes can be input into the association model of each spectral condition and the label classification characteristic information for calculation to obtain an update result, and the mode parameters of a plurality of selectable first update modes of the update strategy are multiplied by the update result to obtain a model update parameter of each label classification characteristic information, so that the convolutional neural network model is updated according to the model update parameter of each label classification characteristic information, and the multispectral image building change detection model is obtained through training.

In a possible embodiment, for step S130, in order to improve the accuracy of dynamic planning and the update efficiency of the acquisition strategy, step S130 may be implemented by the following sub-steps:

sub-step S131, determining an updated template set of acquisition strategies for each current image acquisition device 200 according to the building change labels for each spectral condition obtained by prediction.

And a substep S132, updating the acquisition strategy according to the updating mode and the updating content of each updating template in the updating template set.

For example, a plurality of updated templates may be clustered according to the updated content of each updated template to obtain a plurality of template clusters, where each template cluster corresponds to one of the update modes.

And then, generating an updating process corresponding to each updating template under the current template cluster aiming at each template cluster, classifying the updating templates with the same updating behavior and the same updating source in different updating processes into a class aiming at each template cluster, and fusing target updating processes of each updating process in the class of updating templates in the corresponding updating process when the ratio of the number of the updating contents in the class of updating templates to the total number of the updating processes under the current template cluster exceeds a first threshold value to obtain a first target updating process.

Or in another possible example, the processes which occur only once in the belonging update process and have the same update mode and target update process in different update processes may be classified into one class, and when the ratio of the number of the update contents in the class of update templates to the total number of the update processes under the current template cluster exceeds a first threshold, the target update processes in the belonging update process of each process in the class of update templates are fused to obtain a first target update process.

Or in another possible example, the update templates which appear only once in the update process and have the same update mode and the target update process in different update processes may be classified into one class, and when the ratio of the number of update templates in the class of update templates to the total number of update processes in the current template cluster exceeds a first threshold, the target update processes in the update processes of each update template in the class of update templates are fused to obtain a first target update process.

Therefore, the master update process in the current template cluster can be determined according to the first target update process determined by any one of the above possible examples, and other update templates in the current template cluster are determined as slave update processes, so that the acquisition strategies can be updated according to the update sequence of the master update process and the slave update process in the current template cluster.

Based on the same inventive concept, please refer to fig. 3, which shows a schematic diagram of functional modules of the multispectral image building change detection system 110 provided in the embodiment of the present application, and the embodiment may divide the multispectral image building change detection system 110 into the functional modules according to the above method embodiment. For example, the functional blocks may be divided for the respective functions, or two or more functions may be integrated into one processing block. The integrated module can be realized in a hardware mode, and can also be realized in a software functional module mode. It should be noted that, in the embodiment of the present application, the division of the module is schematic, and is only one logic function division, and there may be another division manner in actual implementation. For example, the multispectral image building change detection system 110 shown in fig. 3 is only a schematic device in the case of dividing each functional module according to each function. The multispectral image building change detection system 110 may include an extraction module 111, a training module 112, and a prediction module 113, and the functions of the functional modules of the multispectral image building change detection system 110 are described in detail below.

The extraction module 111 is configured to extract multispectral image simulation data corresponding to different spectral conditions, where the multispectral image simulation data includes building change feature information extracted based on the spectral conditions and a building change label corresponding to each piece of building change feature information, where the building change feature information includes a spectral reflection change feature of the spectral conditions, the spectral reflection change feature includes reflection feature information of a spectral object region reflected within a preset time period and a spectral condition feature corresponding to the spectral conditions, and the building change label is remaining label information of the virtual change label from which the noise label has been removed. It is understood that the extracting module 111 can be used to execute the step S110, and for the detailed implementation of the extracting module 111, reference can be made to the contents related to the step S110.

And the training module 112 is configured to train the convolutional neural network model according to the multispectral image simulation data corresponding to different spectral conditions, so as to obtain a multispectral image building change detection model. It is understood that the training module 112 can be used to perform the step S120, and the detailed implementation of the training module 112 can refer to the content related to the step S120.

The prediction module 113 is configured to, when the spectral conditions are sent to each image capturing device 200, predict a building change label of the spectral image online data of each spectral condition according to the multispectral image building change detection model, and update the capturing policy of each image capturing device 200 according to the building change label of each spectral condition obtained through prediction. It is understood that the prediction module 113 may be configured to perform the step S130, and for the detailed implementation of the prediction module 113, reference may be made to the content related to the step S130.

In one possible embodiment, the extraction module 111 is configured to extract multispectral image simulation data corresponding to different spectral conditions by:

acquiring spectral condition configuration information from spectral conditions, and extracting spectral condition characteristic information corresponding to different spectral condition items in the spectral condition configuration information;

searching the reflection characteristic information of the spectral object region reflected in the preset time period and the spectral condition characteristic corresponding to the spectral condition from the spectral condition characteristic information;

and performing virtual rendering processing according to the spectral condition characteristic information corresponding to different spectral condition items to obtain virtual change labels, and removing noise labels from the virtual change labels to obtain the building change labels of the building change characteristic information.

In a possible implementation, the extraction module 111 is configured to perform a virtual rendering process to obtain a virtual change tag by:

inputting the spectral condition characteristic information corresponding to different spectral condition items into the virtual rendering model of the corresponding spectral condition item, and acquiring the virtual rendering characteristic information corresponding to the spectral condition characteristic information;

determining global parameters of all virtual rendering container parameters in the second virtual rendering unit and global parameters of all virtual rendering container parameters in the third virtual rendering unit; the second virtual rendering unit is associated with the first virtual rendering unit and is positioned in the same rendering service of the first virtual rendering unit, the third virtual rendering unit is associated with the first virtual rendering unit and is positioned in different rendering services of the first virtual rendering unit, and the first virtual rendering unit, the second virtual rendering unit and the third virtual rendering unit contain the same number of virtual rendering containers;

calculating the correlation parameters of the global parameters of all the virtual rendering container parameters in the second virtual rendering unit and the global parameters of all the virtual rendering container parameters in the first virtual rendering unit, calculating the correlation parameters of the global parameters of all the virtual rendering container parameters in the third virtual rendering unit and the global parameters of all the virtual rendering container parameters in the first virtual rendering unit, and taking the maximum correlation parameter obtained by calculation as the target correlation parameter of the virtual rendering container;

determining a plurality of clustering virtual rendering units according to the target associated parameters of each virtual rendering container in the virtual rendering characteristic information, and obtaining a virtual change label according to the virtual rendering container global parameters of the clustering virtual rendering units.

In one possible implementation, the training module 112 is configured to train the convolutional neural network model to obtain a multi-spectral image building change detection model by:

inputting multispectral image simulation data into a convolutional neural network model, predicting spectral reflection change information of each spectral condition in a spectral reflection process through the convolutional neural network model, determining a label spectral reflection range corresponding to a preset test label interval according to the spectral reflection change information of each spectral condition, and acquiring all spectral reflection labels in the label spectral reflection range to obtain a spectral reflection label matching sequence of each spectral condition;

and training according to the matching result to obtain a multi-spectral image building change detection model.

In one possible implementation, the training module 112 is configured to train the multispectral image building change detection model by:

screening a plurality of update modes which are the same as a preset second update mode from a plurality of selectable first update modes to serve as a plurality of third update modes, wherein the second update mode is a mark update mode output by a plurality of logistic regression analysis tree nodes in the convolutional neural network model, and the logistic regression analysis tree nodes comprise: the updating content is the updating content corresponding to the marking feature node, and the updating content is the updating content corresponding to the marking feature node;

inputting selectable update contents and a plurality of update contents of a plurality of third update modes into the correlation model of each spectral condition and the label classification characteristic information for calculation to obtain an update result, and multiplying mode parameters of a plurality of selectable first update modes of the update strategy by the update result to obtain a model update parameter of each label classification characteristic information;

and updating the convolutional neural network model according to the model update parameters of each label classification characteristic information, and training to obtain a multispectral image building change detection model.

In one possible implementation, the prediction module 113 is configured to update the acquisition strategy of each image acquisition device 200 by:

determining an updated template set of acquisition strategies for each current image acquisition device 200 from the predicted building change labels for each spectral condition;

and respectively updating the acquisition strategy according to the updating mode and the updating content of each updating template in the updating template set.

In one possible embodiment, the prediction module 113 is configured to update the acquisition strategies by:

or, grouping the updating templates which only appear once in the updating process and have the same updating mode and target updating process in different updating processes into one class, and fusing the target updating processes of each updating template in the class of updating templates in the updating process when the ratio of the number of the updating templates in the class of updating templates to the total number of the updating processes under the current template clustering exceeds a first threshold value to obtain a first target updating process;

and respectively updating the acquisition strategies according to the updating sequence of the main updating process and the auxiliary updating process in the current template clustering.

Based on the same inventive concept, referring to fig. 4, a schematic block diagram of a server 100 for performing the above-mentioned multispectral image building change detection method according to an embodiment of the present application is shown, where the server 100 may include a multispectral image building change detection group device 110, a machine-readable storage medium 120, and a processor 130.

In this embodiment, the machine-readable storage medium 120 and the processor 130 are both located in the server 100 and are separately located. However, it should be understood that the machine-readable storage medium 120 may be separate from the server 100 and may be accessed by the processor 130 through a bus interface. Alternatively, the machine-readable storage medium 120 may be integrated into the processor 130, e.g., may be a cache and/or general purpose registers.

The processor 130 is a control center of the server 100, connects various parts of the entire server 100 using various interfaces and lines, performs various functions of the server 100 and processes data by running or executing software programs and/or modules stored in the machine-readable storage medium 120 and calling data stored in the machine-readable storage medium 120, thereby performing overall monitoring of the server 100. Alternatively, processor 130 may include one or more processing cores; for example, the processor 130 may integrate an application processor, which primarily handles operating systems, user interfaces, applications, etc., and a modem processor, which primarily handles wireless communications. It will be appreciated that the modem processor described above may not be integrated into the processor.

The processor 130 may be a general-purpose Central Processing Unit (CPU), a microprocessor, an Application-Specific Integrated Circuit (ASIC), or one or more Integrated circuits for controlling the execution of the procedures of the multispectral image building change detection method provided in the above-described method embodiments.

The machine-readable storage medium 120 may be, but is not limited to, a ROM or other type of static storage device that can store static information and instructions, a RAM or other type of dynamic storage device that can store information and instructions, an Electrically Erasable programmable Read-Only MEMory (EEPROM), a compact disc Read-Only MEMory (CD-ROM) or other optical disk storage, optical disk storage (including compact disc, laser disk, optical disk, digital versatile disk, blu-ray disk, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer. The machine-readable storage medium 120 may be self-contained and coupled to the processor 130 via a communication bus. The machine-readable storage medium 120 may also be integrated with the processor. The machine-readable storage medium 120 is used for storing machine-executable instructions for performing aspects of the present application. The processor 130 is configured to execute machine-executable instructions stored in the machine-readable storage medium 120 to implement the multi-spectral image building change detection method provided by the foregoing method embodiments.

The multispectral image building change detection system 110 may include software functional modules (e.g., the extraction module 111, the training module 112, and the prediction module 113 shown in fig. 3) stored in the machine-readable storage medium 120, when executed by the processor 130 in the multispectral image building change detection system 110, to implement the multispectral image building change detection method provided by the foregoing method embodiments.

Since the server 100 provided in the embodiment of the present application is another implementation form of the method embodiment executed by the server 100, and the server 100 may be configured to execute the method for detecting a change in a building with a multispectral image provided in the method embodiment, the method embodiment may be selected as the technical effect that can be obtained by the server 100, and details are not described here again.

Further, the present application also provides a readable storage medium containing computer executable instructions, which when executed, can be used to implement the multispectral image building change detection method provided by the above method embodiments.

Of course, the storage medium provided in the embodiments of the present application contains computer-executable instructions, and the computer-executable instructions are not limited to the above method operations, and may also perform related operations in the multispectral image building change detection method provided in any embodiments of the present application.

Embodiments of the present application are described with reference to flowchart illustrations and/or block diagrams of methods, apparatus and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

While the present application has been described in connection with various embodiments, other variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed application, from a review of the drawings, the disclosure, and the appended claims. In the claims, the word "comprising" does not exclude other elements or steps, and the word "a" or "an" does not exclude a plurality. A single processor or other unit may fulfill the functions of several items recited in the claims. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.

The above description is only for various embodiments of the present application, but the scope of the present application is not limited thereto, and any person skilled in the art can easily conceive of changes or substitutions within the technical scope of the present application, and all such changes or substitutions are included in the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. A multispectral image building change detection method is applied to a server, the server is in communication connection with an image acquisition device, and the method comprises the following steps:

2. The method for detecting building changes according to multispectral images as claimed in claim 1, wherein said step of extracting multispectral image simulation data corresponding to different spectral conditions comprises:

searching the reflection characteristic information of the spectral object region reflected within a preset time period and the spectral condition characteristic corresponding to the spectral condition from the spectral condition characteristic information;

3. The method for detecting building changes according to multispectral images as defined in claim 2, wherein the step of performing virtual rendering processing on the spectral condition feature information corresponding to the different spectral condition items to obtain virtual change tags includes:

4. The method according to claim 1, wherein the step of training the convolutional neural network model according to the multispectral image simulation data corresponding to the different spectral conditions to obtain the multispectral image building change detection model comprises:

training according to the matching result to obtain the multispectral image building change detection model;

the step of training to obtain the multispectral image building change detection model according to the matching result comprises the following steps:

5. The multi-spectral image building change detection method according to claim 1, wherein the step of updating the collection strategy of each image collection device according to the building change label of each spectral condition obtained by prediction comprises:

updating the acquisition strategy according to the updating mode and the updating content of each updating template in the updating template set;

the step of updating the acquisition strategy according to the updating mode and the updating content of each updating template in the updating template set comprises the following steps:

6. A multi-spectral image building change detection system for use with a server communicatively coupled to an image capture device, the system comprising:

7. The system according to claim 6, wherein the means for extracting the multispectral image modeling data corresponding to different spectral conditions comprises:

8. The system according to claim 7, wherein the means for performing virtual rendering processing according to the spectral condition feature information corresponding to the different spectral condition items to obtain virtual change labels comprises:

9. The system according to claim 6, wherein the means for training the convolutional neural network model according to the multispectral image simulation data corresponding to the different spectral conditions to obtain the multispectral image building change detection model comprises:

the method for training and obtaining the multispectral image building change detection model according to the matching result comprises the following steps:

10. The multi-spectral image building change detection system according to claim 6, wherein the manner of updating the acquisition strategy of each image acquisition device according to the building change label of each spectral condition obtained by prediction comprises:

the method for updating the acquisition strategy according to the updating mode and the updating content of each updating template in the updating template set comprises the following steps: