ENVI platform-based target change detection system and control method thereof

Abstract

The invention relates to the field of image processing, and discloses a target change detection system based on an ENVI platform and a control method thereof. The system comprises an image processing module realized based on an IDL compilation function, a target screening module realized based on a Python compilation function, and a display module. Wherein the image processing module is used for obtaining a change pattern spot slice based on a preprocessed image obtained by a plurality of time phase images; the target screening module is used for carrying out change target judgment on the change pattern spot slices through a preset deep learning network to obtain a target change detection result containing a change target object; and the display module is used for highlighting the target change detection result based on a preset display strategy. The problem that a platform needs to be switched when traditional image processing and deep learning method processing are carried out on the multi-source remote sensing image can be avoided, so that a deep learning algorithm can be conveniently and effectively introduced into a target change detection task, use is easy, expandability is high, and operation and visualization of operators are facilitated.

Description

Technical Field

[0001] The present disclosure relates to the field of image processing, and in particular to an ENVI platform-based target change detection system and a control method thereof. Background Art

[0002] Currently, some studies have proposed solutions to the problem of target change detection at the algorithm level, but they are generally not applied to mature commercial software systems for implementation. Figure 1 In the scheme shown, if each link is implemented using algorithm programming, the effective calling of the CPU and GPU can be ensured, but it involves the need to call a large number of general image processing algorithms, and it is difficult to take into account the processing requirements of different types of data. It is not as convenient to use and has less visualization effects than general image processing software. If general image processing software is used to implement preprocessing and change detection, and algorithm programming is used to implement target detection, it involves switching between multiple software platforms and computing devices (CPU or GPU), making it difficult to ensure the automatic implementation of the entire system process.

[0003] Among the common remote sensing image processing software, ENVI (The Environment for Visualizing Images) is a commercial remote sensing software developed using the interactive data language IDL, which integrates image data input / output, image calibration, image enhancement, correction, orthorectification, mosaic, data fusion and various transformations, information extraction, image classification, knowledge-based decision tree classification, integration with GIS, DEM and terrain information extraction, radar data processing, 3D display analysis and other functions. It also supports users to expand new functions by themselves, that is, to conduct secondary research and development of ENVI using the underlying IDL language. Summary of the invention

[0004] The purpose of the present invention is to at least provide a target change detection system based on the ENVI platform and a control method thereof. The target change detection system based on the ENVI platform disclosed by the present invention provides a system that can support target change detection through mixed programming of IDL and Python. Based on the current mainstream general image processing software architecture ENVI, it integrates cutting-edge deep learning technology. By building menus, calling system functions, calling IDL compilation functions, and calling Python compilation functions, a software system integrating multiple functions such as multi-phase image reading, multi-phase image preprocessing, multi-feature extraction, change detection, and change target discrimination is designed. In actual applications, the manual switching of different software platforms, different compilation languages, and different computing units is avoided, so that operators can simply and directly complete the target change detection task.

[0005] At present, although the general remote sensing image processing software ENVI has many basic remote sensing image processing algorithms, it does not integrate the relevant algorithms and functions of deep learning. In view of this, this application adopts the IDL+Python hybrid programming method to achieve the organic combination of common remote sensing image processing algorithms and deep learning algorithms, and discloses a target change detection system, which has both mature and efficient traditional remote sensing image processing functions, and can perform deep learning-based processing and achieve GPU acceleration.

[0006] The system disclosed in this application is based on the general remote sensing image processing software ENVI5.3, and is completed through secondary development using two development languages, IDL and Python. At the same time, it innovatively integrates deep learning algorithms to achieve the successful application of common deep learning frameworks (such as Caffe and Theano). The system avoids the problem of switching platforms when performing traditional image processing and deep learning methods on multi-source remote sensing images, thereby facilitating the effective introduction of deep learning algorithms into target change detection tasks, and possesses the main functions of each link of remote sensing image target change detection. In addition, the system is fully functional, easy to use, and highly scalable, which is convenient for operators to operate and visualize.

[0007] To solve the above technical problems, at least one embodiment of the present application provides a target change detection system based on the ENVI platform, the system comprising:

[0008] Image processing module implemented based on IDL compiled functions, target screening module and display module implemented based on Python compiled functions; among them:

[0009] The image processing module is used to obtain a change pattern slice based on the pre-processed image obtained from the multiple phase images;

[0010] The target screening module is used to perform change target discrimination on the change pattern slices through a preset deep learning network to obtain a target change detection result containing a change target object;

[0011] The display module is used to highlight the target change detection result based on a preset display strategy.

[0012] At least one embodiment of the present application further provides a control method for the target change detection system based on the ENVI platform as described above, the method comprising:

[0013] Controlling the image processing module to perform preset image processing on the received multiple phase images to obtain change pattern slices;

[0014] The control target screening module performs change target discrimination on the change pattern slices through a preset deep learning network to obtain a target change detection result containing a change target object;

[0015] The target change detection result is highlighted by the display module based on a preset display strategy.

[0016] At least one embodiment of the present application further provides an electronic device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method described above.

[0017] At least one embodiment of the present application further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the method described above are implemented.

[0018] At least one embodiment of the present application further provides a computer program product, including a computer program, which implements the steps of the control method described above when executed by a processor.

[0019] The target change detection system based on the ENVI platform provided in the embodiment of the present application is completed by secondary development using two development languages, IDL and Python, based on the general remote sensing image processing software ENVI5.3. At the same time, it innovatively integrates deep learning algorithms to achieve the successful application of commonly used deep learning frameworks (such as Caffe and Theano). The system avoids the problem of switching platforms when traditional image processing and deep learning methods are currently used for multi-source remote sensing images, thereby facilitating the effective introduction of deep learning algorithms into target change detection tasks, and possesses the main functions of each link of remote sensing image target change detection. In addition, the system is fully functional, easy to use, and highly scalable, which is convenient for operators to operate and visualize.

[0020] In some optional embodiments, the system further comprises:

[0021] The basic functional module is used to perform morphological processing on multiple phase images to obtain the pre-processed image, and perform multi-feature extraction, band superposition and calculation operations on the pre-processed image to obtain a multi-feature difference vector; wherein the multi-features include: spatial features and texture features.

[0022] In some optional embodiments, the image processing module includes:

[0023] IR-MAD conversion unit, used to obtain a change intensity map based on two phase images;

[0024] An automatic sample unit, used for obtaining a first sample with a change and a second sample without a change from the change intensity map based on a preset threshold and according to a histogram statistics method;

[0025] An SVM classification unit, configured to perform binary classification on the multi-feature difference vector provided by the basic function module based on the first sample and the second sample to obtain a classification result;

[0026] A post-classification processing unit, used for performing filtering optimization processing on the classification result to ensure that the identified changing target object is continuous and obtain an optimized detection result;

[0027] The spot extraction unit is used to obtain the local image containing the changed target object from the optimized detection result, and use the local image containing the changed target object as the spot slice.

[0028] In some optional embodiments, the target screening module includes:

[0029] A model training unit, used to train the preset deep learning network based on a preset sample training set to obtain a trained deep learning network;

[0030] A target initial screening unit, used to discriminate the change pattern slices according to the trained deep learning network to obtain an initial screening result;

[0031] A target fine screening unit is used to merge the image patch slices that meet the preset conditions in the initial screening results to obtain processed screening results, and to discriminate the processed screening results according to the trained deep learning network to obtain fine screening results;

[0032] The result output unit is used to mark the refined screening results based on preset requirements to obtain target change detection results.

[0033] In some optional embodiments, the image processing module further includes a spot merging unit, wherein:

[0034] The spot merging unit is used to merge multiple spot slices that meet preset conditions to obtain merged spot slices.

[0035] In some optional embodiments, the target fine screening unit is further used to issue a spot merging instruction to control the spot merging unit in the image processing module to perform a spot merging operation and feed back the merged spot slices.

[0036] In some optional embodiments, highlighting the target change detection result based on a preset display strategy includes:

[0037] A highlighted display area of ​​a preset style is set around the change target object.

[0038] In some optional embodiments, the annotation information of the target change detection result includes:

[0039] The preset attribute information and / or real-time status information of the target object is changed.

[0040] In some optional embodiments, highlighting the target change detection result based on a preset display strategy includes:

[0041] Displaying preset attribute information of the change target object at a preset position of the change target object; and / or,

[0042] The real-time status information of the changing target object is displayed at the preset position of the changing target object. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] One or more embodiments are exemplarily described by pictures in the corresponding drawings, and these exemplified descriptions do not constitute limitations on the embodiments. Elements with the same reference numerals in the drawings represent similar elements, and unless otherwise stated, the figures in the drawings do not constitute proportional limitations.

[0044] Figure 1 A schematic diagram of a target change detection process and implementation approach;

[0045] Figure 2 A schematic diagram of the structure of a target change detection system provided by an embodiment of the present disclosure;

[0046] Figure 3 A schematic diagram of a target change detection process and implementation approach provided in an embodiment of the present disclosure;

[0047] Figure 4 A schematic diagram of the structure of another target change detection system provided by an embodiment of the present disclosure;

[0048] Figure 5 A variation intensity diagram using an aircraft as an example provided in an embodiment of the present disclosure;

[0049] Figure 6 A classification result diagram using an airplane as an example provided in an embodiment of the present disclosure;

[0050] Figure 7 A schematic diagram of a display effect provided by an embodiment of the present disclosure;

[0051] Figure 8 A flow chart of a control method provided by an embodiment of the present disclosure. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical scheme and advantages of the embodiments of the present invention clearer, the embodiments of the present invention will be described in detail below with reference to the accompanying drawings. However, it can be understood by those skilled in the art that in the embodiments of the present invention, many technical details are provided to enable readers to better understand the present invention. However, even without these technical details and various changes and modifications based on the following embodiments, the technical scheme claimed in the present invention can be implemented.

[0053] Considering that the general remote sensing image processing software ENVI has many basic remote sensing image processing algorithms, it does not integrate the relevant algorithms and functions of deep learning. In view of this, this application adopts the IDL+Python hybrid programming method to achieve the organic combination of common remote sensing image processing algorithms and deep learning algorithms, and discloses a target change detection system, which has both mature and efficient traditional remote sensing image processing functions, and can perform deep learning-based processing and achieve GPU acceleration.

[0054] The system disclosed in this application is based on the general remote sensing image processing software ENVI5.3, and is completed through secondary development using two development languages, IDL and Python. At the same time, it innovatively integrates deep learning algorithms to achieve the successful application of common deep learning frameworks (such as Caffe and Theano). The system avoids the problem of switching platforms when performing traditional image processing and deep learning methods on multi-source remote sensing images, thereby facilitating the effective introduction of deep learning algorithms into target change detection tasks, and possesses the main functions of each link of remote sensing image target change detection. In addition, the system is fully functional, easy to use, and highly scalable, which is convenient for operators to operate and visualize.

[0055] Embodiment 1:

[0056] An embodiment of the present invention relates to a target change detection system based on the ENVI platform.

[0057] The implementation details of the target change detection system based on the ENVI platform of this embodiment are specifically described below. The following content is only the implementation details provided for the convenience of understanding and is not necessary for the implementation of this solution.

[0058] like Figure 2 As shown, the target change detection system based on the ENVI platform provided in this embodiment includes:

[0059] Image processing module implemented based on IDL compiled functions, target screening module and display module implemented based on Python compiled functions; among them:

[0060] The image processing module is used to obtain a change pattern slice based on the pre-processed image obtained from the multiple phase images;

[0061] The target screening module is used to perform change target discrimination on the change pattern slices through a preset deep learning network to obtain a target change detection result containing a change target object;

[0062] The display module is used to highlight the target change detection result based on a preset display strategy.

[0063] Specifically, the target change detection system disclosed in this embodiment is implemented based on the ENVI (eg, ENVI5.3) platform.

[0064] It is understandable that data can be exchanged between the modules in the system; as an example, data can be exchanged through interface calls and / or function module calls.

[0065] Optionally, refer to Figure 3 In this embodiment, taking the aircraft target as an example, the target change detection algorithm flow and its implementation method used in this system are given. The algorithm implementation process involves multi-temporal image reading, multi-temporal image preprocessing and multi-feature extraction realized through the original functions of the ENVI platform, change detection realized by the image processing module, and change target discrimination realized by the target screening module.

[0066] Embodiment 2:

[0067] Based on the above embodiments, the implementation manner of the present invention further explains and illustrates the target change detection system based on the ENVI platform.

[0068] The target change detection system disclosed in this embodiment is implemented based on the ENVI (for example, ENVI5.3) platform. The overall architecture of this system can be referred to Figure 4 .

[0069] In some embodiments, the system further comprises:

[0070] The basic functional module is used to perform morphological processing on multiple phase images to obtain the pre-processed image, and perform multi-feature extraction, band superposition and calculation operations on the pre-processed image to obtain a multi-feature difference vector; wherein the multi-features include: spatial features and texture features.

[0071] Specifically, the system retains some of the original functions of the ENVI system, including: opening images, geometric registration, radiation correction, spatial feature extraction, texture feature extraction, band overlay, band operation, SVM classification, classification post-processing, etc. The specific function names and function index names are shown in Table 1. Among them, geometric registration uses the method of manually selecting control points (image to image) in the ENVI software, and the function name is register0; radiation correction uses the dark pixel method, and the function name is dark subtract; classification post-processing uses the clustering method, and the function name is clump; other functions are completely corresponding to the function index names in the original ENVI system. For details, please refer to the following Table 1:

[0072] Table 1 ENVI system function call table

[0073] Function Name Function Index Name Open Image open envi file Geometric Registration register0 Radiation correction dark subtract Spatial feature extraction convolution tool Texture feature extraction texture co-occurrence Band overlay Layer stacking Band Operation band math SVM Classification svm tool Post-classification processing clump

[0074] In this embodiment, a menu is built under the ENVI system framework, which is easy to operate and understand, and realizes the reasonable calling of the original ENVI system function, IDL compilation function, and Python compilation function. At the same time, it also has good scalability, and users can further improve the system at any time according to their actual needs.

[0075] In this embodiment, the original functional units under the ENVI system framework are retained, and the main retained functional units include: image opening unit, geometric registration unit, radiation correction unit, spatial feature extraction unit, texture feature extraction unit, band overlay unit and band operation unit; specifically:

[0076] The function of "Open Image Unit" is mainly to realize the input of multi-temporal images. It supports ENVI system image formats and almost all common remote sensing image formats. The image display and interactive interface are the original operation interfaces of the ENVI system. It is reasonably designed and can automatically establish image pyramids with fast response speed.

[0077] The function of the "geometric registration unit" is mainly to realize geometric correction in multi-temporal image preprocessing. By manually selecting correction points, two images that correspond one to one in geometric position are obtained.

[0078] The function of the "radiation correction unit" is mainly to realize the radiation correction in the multi-temporal image preprocessing. Through the dark pixel method, two images with basically the same radiation values ​​are obtained.

[0079] The function of the "spatial feature extraction unit" is mainly to implement the convolution of the Sobel operator on the image. Different convolution window sizes can be set as needed to obtain the edge features of the image.

[0080] The function of the "texture feature extraction unit" is mainly to realize the extraction of features such as mean, inverse moment, entropy, angular second moment, etc. calculated by the gray-level co-occurrence matrix. The window size and calculation direction can be set as needed to obtain different texture features.

[0081] The function of the "band overlay unit" is mainly to combine the original image spectral features, extracted spatial features, and extracted texture features to form a multi-feature set of images and complete multi-feature extraction.

[0082] The function of the "band operation unit" is mainly to normalize the multi-feature set, subtract the multi-feature sets of two phases, and construct a multi-feature difference vector, so as to describe the changes between the two phase images from multiple feature angles.

[0083] In some embodiments, the image processing module includes:

[0084] IR-MAD conversion unit, used to obtain a change intensity map based on two phase images;

[0085] An automatic sample unit, used for obtaining a first sample with a change and a second sample without a change from the change intensity map based on a preset threshold and according to a histogram statistical method; wherein the preset threshold can be set according to actual needs;

[0086] An SVM classification unit, configured to perform binary classification on the multi-feature difference vector provided by the basic function module based on the first sample and the second sample to obtain a classification result;

[0087] A post-classification processing unit, used for performing filtering optimization processing on the classification result to ensure that the identified changing target object is continuous and obtain an optimized detection result;

[0088] The spot extraction unit is used to obtain the local image containing the changed target object from the optimized detection result, and use the local image containing the changed target object as the spot slice.

[0089] Optionally, the IDL compilation environment of the ENVI system is used to implement programming of functions such as IR-MAD transformation, automatic sampling, spot extraction, and result output, which are not available in the original ENVI system.

[0090] Among them, IR-MAD transformation is a widely used method in multispectral image change detection. The input data is the image data of two phases, and the output result is the change intensity map, which increases the separability of the changed area and the unchanged area; the automatic sample uses histogram analysis and statistical methods. The input is the change intensity map after IR-MAD transformation, and the output result is a certain number of changed samples and unchanged samples in the ASCII encoded .txt file format; the spot extraction obtains the change spot slices one by one based on the morphological processing of the two-class change detection results of SVM, and stores them as .jpg format pictures; the result output refers to the storage of the final change detection results, including image map and animated image formats. The image map refers to the .jpg format image that annotates the specific new and disappeared target positions and quantities, and the animated image refers to the .gif format file that loops the original image and the annotated image.

[0091] As an example, specifically:

[0092] The function of the "IR-MAD transformation unit" is to obtain the result that the changed area and the unchanged area are easier to distinguish through iteration, laying the foundation for the automatic extraction of subsequent changed samples; among them, taking the aircraft target as an example, the change intensity map obtained by the IR-MAD transformation unit can be referred to Figure 5 ;

[0093] The function of "Automatic Sample Unit" is to automatically obtain the threshold of the changed sample and the unchanged sample through the histogram statistics method, and compare it with the preset threshold (to obtain the comparison result), without manual intervention, to provide the required sample data for the subsequent SVM-based change detection;

[0094] The function of the "SVM classification unit" is to perform SVM-based binary classification on multi-feature difference vectors with the support of a certain number of changed and unchanged samples to obtain the change detection results of the entire image; among them, taking the aircraft target as an example, the classification result diagram obtained by the SVM classification unit can be referred to Figure 6 ;

[0095] The function of the "post-classification processing unit" is mainly to post-process the results of SVM classification (for example, further perform some filtering optimization and other processing), filter some fragmented classification results, merge some adjacent change areas, and ensure that the objects of subsequent target recognition are continuous.

[0096] The function of the "pattern extraction unit" is mainly to provide change pattern slices for subsequent target recognition.

[0097] In some embodiments, the target screening module comprises:

[0098] A model training unit, used to train the preset deep learning network based on a preset sample training set to obtain a trained deep learning network;

[0099] A target initial screening unit, used to discriminate the change pattern slices according to the trained deep learning network to obtain an initial screening result;

[0100] A target fine screening unit is used to merge the image patch slices that meet the preset conditions in the initial screening results to obtain processed screening results, and to discriminate the processed screening results according to the trained deep learning network to obtain fine screening results;

[0101] The result output unit is used to mark the refined screening results based on preset requirements to obtain target change detection results.

[0102] The preset requirements can be set according to actual requirements. Taking an aircraft target as an example, the preset requirements can include the attribute information of the aircraft and the real-time status information of the aircraft.

[0103] Optionally, use Python language to implement model training and target recognition functions involving deep learning, and use the command "caffe.set_mode_gpu()" to call GPU operation to improve the computational efficiency of the algorithm. By referencing the command "import python" in the IDL program, ensure that IDL and Python languages ​​can be compiled at the same time. The main functions implemented include model training, target initial screening and target fine screening.

[0104] Among them, the model training called the caffe framework, used the NIN network structure and the weights pre-trained by cifar10 as the initial weights for training; the initial target screening and the fine target screening were also carried out under the caffe framework. The difference between the two is that the thresholds used to judge whether the pattern belongs to the target class of interest are different. The input data are all changing patterns, that is, pictures in .jpg format, and the output results are the picture numbers belonging to the target class of interest, so as to achieve the purpose of target recognition of changing patterns.

[0105] As an example, specifically:

[0106] The function of the "model training unit" is to input a certain number of target training samples, call GPU calculation, and obtain the target deep learning network function parameters through learning;

[0107] The function of the "target initial screening unit" is to use the network parameters obtained by "model training" to call GPU calculation to identify the input change spots, that is, to determine whether the spots are the target of interest. By giving a relatively rough threshold, the pictures that obviously do not belong to the target of interest are filtered out;

[0108] The function of the "target fine screening unit" is to merge the spots with a certain overlap after the "target initial screening", set high and low thresholds according to the size of the spots, call GPU calculation, complete more accurate spot recognition and change analysis, and obtain the final disappeared targets and newly added targets;

[0109] The function of the "result output unit" is mainly to generate labeled target change detection results according to the actual needs of the user, and the display method is intuitive and clear.

[0110] In some embodiments, the image processing module further includes a spot merging unit, wherein:

[0111] The spot merging unit is used to merge multiple spot slices that meet preset conditions to obtain merged spot slices; that is, the post-classification processing unit can be called.

[0112] Taking an airplane target as an example, the preset conditions may include: what should have been a complete image patch slice containing the airplane is split into two or more image patch slices each containing a certain part of the airplane.

[0113] In some embodiments, the target fine screening unit is further used to issue a spot merging instruction to control the spot merging unit in the image processing module to perform a spot merging operation and feed back the merged spot slices.

[0114] In some embodiments, the highlighting of the target change detection result based on a preset display strategy includes:

[0115] A highlighted display area of ​​a preset style is set around the change target object.

[0116] Among them, the preset style can be set as: setting a box that is highlighted in a specific color; the preset style can also be set according to actual needs.

[0117] Furthermore, when displaying, the boxes corresponding to the disappeared target and the newly added target can be highlighted in different colors respectively.

[0118] The labeling information of the target change detection result includes:

[0119] In some embodiments, the change includes preset attribute information and / or real-time status information of the target object.

[0120] In some embodiments, the highlighting of the target change detection result based on a preset display strategy includes:

[0121] Displaying preset attribute information of the change target object at a preset position of the change target object; and / or,

[0122] The real-time status information of the changing target object is displayed at the preset position of the changing target object.

[0123] Among them, taking the aircraft target as an example, the preset attribute information may include: size (length, width and height), aircraft model, flight number, etc.; the real-time status information may include: altitude, heading, speed, etc.; the preset direction can be set to above, below, left or right, etc. In addition, the preset direction, preset attribute information and real-time status information can all be set according to actual needs.

[0124] For a specific display method, please refer to Figure 7 .

[0125] Therefore, this system adopts an integrated architecture design, effectively integrating the two development languages ​​of IDL and Python, and can realize the full process processing of multi-temporal image target change detection. This system is a secondary development based on the powerful general remote sensing image processing software ENVI. It has strong applicability and can process a variety of satellite data. This system makes up for the shortcomings of traditional remote sensing image processing software that lacks deep learning capabilities and GPU high-performance computing. This system has strong scalability and can be further developed according to user needs to adapt to the change detection of multiple targets and output different change detection results customized according to user needs.

[0126] In summary, the target change detection system disclosed in this embodiment is based on the general remote sensing image processing software ENVI5.3, and is completed through secondary development using two development languages, IDL and Python. At the same time, it innovatively integrates deep learning algorithms to achieve the successful application of common deep learning frameworks (such as Caffe and Theano). The system avoids the problem of switching platforms when performing traditional image processing and deep learning methods on multi-source remote sensing images, thereby facilitating the effective introduction of deep learning algorithms into target change detection tasks, and possesses the main functions of each link of remote sensing image target change detection. In addition, the system is fully functional, easy to use, and highly scalable, which is convenient for operators to operate and visualize.

[0127] Embodiment three:

[0128] On the basis of the above embodiments, this embodiment provides a control method of the system as described in the above embodiments.

[0129] The control method provided in this embodiment can be applied to electronic devices with communication, computing and data storage capabilities. Figure 8 As shown, the control method provided in this embodiment includes the following steps:

[0130] Step 810, controlling the image processing module to perform preset image processing on the received multiple phase images to obtain change spot slices;

[0131] It can be understood that the preset image processing in step 810 may include: performing morphological processing, band superposition and calculation operations through the basic function module, and performing related processing operations through the IR-MAD transformation unit, the automatic sample unit, the SVM classification unit, the post-classification processing unit, and the spot extraction unit respectively;

[0132] Step 820, controlling the target screening module to perform change target discrimination on the change pattern slices through a preset deep learning network to obtain a target change detection result containing a change target object;

[0133] It can be understood that in step 820, the model training unit, the target preliminary screening unit, the target fine screening unit and the result output unit perform relevant processing operations respectively;

[0134] In some possible situations, some functional units in the image processing module may be interspersed and called in step 820 to perform corresponding operations;

[0135] Step 830: highlighting the target change detection result through the display module based on a preset display strategy.

[0136] Specifically, the target change detection result obtained in step 820 is sent to the display module, and the target change detection result can be highlighted based on a preset display strategy.

[0137] Although the present application provides the method operation steps as shown in the above-mentioned embodiments or drawings, more or fewer operation steps may be included in the method based on routine or no creative labor. In the steps where there is no necessary causal relationship logically, the execution order of these steps is not limited to the execution order provided in the embodiments of the present application. When the method is executed in an actual device or terminal product, it can be connected in sequence or in parallel according to the method shown in the embodiments or drawings.

[0138] Embodiment 4:

[0139] Another embodiment of the present application relates to an electronic device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the control method in the above-mentioned embodiments.

[0140] Among them, the memory and the processor are connected in a bus manner, and the bus may include any number of interconnected buses and bridges, and the bus connects various circuits of one or more processors and memories together. The bus can also connect various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. The bus interface provides an interface between the bus and the transceiver. The transceiver can be one element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices on a transmission medium. The data processed by the processor is transmitted on a wireless medium via an antenna, and further, the antenna also receives data and transmits the data to the processor.

[0141] The processor is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory can be used to store data used by the processor when performing operations.

[0142] Embodiment five:

[0143] Another embodiment of the present application relates to a computer-readable storage medium storing a computer program, which implements the above method embodiment when executed by a processor.

[0144] That is, those skilled in the art can understand that all or part of the steps in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a program, and the program is stored in a storage medium, including a number of instructions to enable a device (which can be a single-chip microcomputer, chip, etc.) or a processor to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as: ROM), random access memory (Random Access Memory, referred to as: RAM), disk or optical disk and other media that can store program codes.

[0145] In some embodiments of the present application, a computer program product is also provided, including a computer program, which implements the steps of the method described in the above embodiment when executed by a processor.

[0146] Those skilled in the art will appreciate that the above embodiments are specific embodiments for implementing the present application, and in actual applications, various changes may be made thereto in form and detail without departing from the spirit and scope of the present application.

Claims

1. A target change detection system based on ENVI platform, characterized in that: The system comprises: an image processing module implemented based on an IDL compilation function, a target screening module and a display module implemented based on a Python compilation function; wherein: The image processing module is used to obtain a change pattern slice based on the pre-processed image obtained from the multiple phase images; The target screening module is used to perform change target discrimination on the change pattern slices through a preset deep learning network to obtain a target change detection result containing a change target object; The display module is used to highlight the target change detection result based on a preset display strategy.

2. The system according to claim 1, characterized in that Also includes: The basic functional module is used to perform morphological processing on multiple phase images to obtain the pre-processed image, and perform multi-feature extraction, band superposition and calculation operations on the pre-processed image to obtain a multi-feature difference vector; wherein the multi-features include: spatial features and texture features.

3. The system according to claim 2, characterized in that The image processing module comprises: IR-MAD conversion unit, used to obtain a change intensity map based on two phase images; An automatic sample unit, used for obtaining a first sample with a change and a second sample without a change from the change intensity map based on a preset threshold and according to a histogram statistics method; An SVM classification unit, configured to perform binary classification on the multi-feature difference vector provided by the basic function module based on the first sample and the second sample to obtain a classification result; A post-classification processing unit, used for performing filtering optimization processing on the classification result to ensure that the identified changing target object is continuous and obtain an optimized detection result; The spot extraction unit is used to obtain the local image containing the changed target object from the optimized detection result, and use the local image containing the changed target object as the spot slice.

4. The system according to claim 1, characterized in that The target screening module comprises: A model training unit, used to train the preset deep learning network based on a preset sample training set to obtain a trained deep learning network; A target initial screening unit, used to discriminate the change pattern slices according to the trained deep learning network to obtain an initial screening result; A target fine screening unit is used to merge the image patch slices that meet the preset conditions in the initial screening results to obtain processed screening results, and to discriminate the processed screening results according to the trained deep learning network to obtain fine screening results; The result output unit is used to mark the refined screening results based on preset requirements to obtain target change detection results.

5. The system according to claim 1, characterized in that The image processing module also includes a spot merging unit, wherein: The spot merging unit is used to merge multiple spot slices that meet preset conditions to obtain merged spot slices.

6. The system according to claim 4, characterized in that The target fine screening unit is further used to issue a spot merging instruction to control the spot merging unit in the image processing module to perform a spot merging operation and feed back the merged spot slices.

7. The system according to claim 1, characterized in that The highlighting of the target change detection result based on a preset display strategy includes: A highlighted display area of ​​a preset style is set around the change target object.

8. The system according to claim 4, characterized in that The labeling information of the target change detection result includes: The preset attribute information and / or real-time status information of the target object is changed.

9. The system according to claim 8, characterized in that The highlighting of the target change detection result based on a preset display strategy includes: Displaying preset attribute information of the change target object at a preset position of the change target object; and / or, The real-time status information of the changing target object is displayed at the preset position of the changing target object.

10. A control method for a target change detection system based on the ENVI platform as claimed in any one of claims 1 to 9, characterized in that: The method comprises: Controlling the image processing module to perform preset image processing on the received multiple phase images to obtain change pattern slices; The control target screening module performs change target discrimination on the change pattern slices through a preset deep learning network to obtain a target change detection result containing a change target object; The target change detection result is highlighted by the display module based on a preset display strategy.

Images