A scene change detection and damage assessment method based on infrared polarization imaging
By using infrared polarization imaging technology, combined with the PCANet network and SVM classification and detection module, the problem of insufficient utilization of infrared polarization information in existing technologies is solved, and high-precision scene change detection and damage assessment are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIJING INST OF ENVIRONMENTAL FEATURES
- Filing Date
- 2023-11-09
- Publication Date
- 2026-07-21
AI Technical Summary
Existing technologies cannot fully utilize infrared polarization information for scene change detection, resulting in insufficient detection accuracy.
An infrared polarization imaging-based method is adopted. By acquiring infrared intensity, infrared polarization degree and polarization angle images of the target scene, feature registration and differential processing are performed. The PCANet network and SVM classification and detection module are used to extract and filter image blocks. Combined with region filling and noise filtering, the final change detection image result is obtained.
It improves the accuracy of scene change detection, provides a new technical approach for damage performance assessment, and realizes high-precision detection and damage assessment of scene changes.
Smart Images
Figure CN117474889B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of image detection technology, and in particular to a method for scene change detection and damage assessment based on infrared polarization imaging. Background Technology
[0002] Damage assessment based on optical image detection hinges on high-precision extraction of change information, and the results of the extracted changes play a decisive role in the accuracy of damage detection.
[0003] Current optical imaging systems are highly sensitive to the natural environment and easily affected by weather conditions such as rain, snow, and fog, making single-mode optical image information insufficient for accurate analysis of scene damage. Compared with traditional optical imaging techniques, infrared polarization imaging technology utilizes the polarization characteristics of the target's radiated light waves. This significantly enhances the difference between camouflaged or dimly lit targets and the background, filters out interference, improves signal contrast, highlights target contour information, and achieves high-precision extraction of scene change features. It has potential advantages in the field of damage scene change detection. Currently, existing technologies cannot fully utilize the target's polarization information to detect scene changes, thus affecting detection accuracy. Summary of the Invention
[0004] To address the problem that existing technologies struggle to fully utilize target polarization information for scene change detection, this invention provides a scene change detection and damage assessment method based on infrared polarization imaging. This method fully leverages the infrared polarization characteristics of the scene, improves scene change detection accuracy, and offers a new technical approach for damage performance assessment.
[0005] In a first aspect, embodiments of the present invention provide a scene change detection method based on infrared polarization imaging, comprising:
[0006] Acquire infrared intensity, infrared polarization degree, and polarization angle images of the target scene before and after the change;
[0007] Feature registration was performed on the infrared intensity, infrared polarization degree, and polarization angle images before and after the change, respectively, to obtain the registered infrared intensity, infrared polarization degree, and polarization angle image group;
[0008] Intra-group difference processing was performed on the registered infrared intensity, infrared polarization degree and polarization angle image groups to obtain three kinds of two-phase difference images corresponding to infrared intensity, infrared polarization degree and polarization angle.
[0009] Based on the obtained two-phase difference images, one or more candidate change point image blocks are obtained through image block extraction and filtering.
[0010] The extracted candidate change point image patches are input into the trained detection model to obtain change detection results; the detection model is built based on the PCANet network and the SVM classification and detection module.
[0011] Based on the change detection results, the final change detection image of the target scene is obtained through region filling and noise filtering.
[0012] Optionally, acquiring the infrared intensity, infrared polarization degree, and polarization angle images of the target scene before and after the change includes:
[0013] Using a polarization detection system, the radiance of the target scene before and after the change in different polarization directions is measured, and the infrared intensity, infrared polarization degree, and polarization angle images of the target scene before and after the change are calculated respectively.
[0014] Optionally, the feature registration of the infrared intensity, infrared polarization degree, and polarization angle images before and after the change includes:
[0015] The MSER-SURF feature fusion registration method was used to perform feature registration on the infrared intensity, infrared polarization degree, and polarization angle images before and after the change.
[0016] Optionally, based on the obtained two-phase difference images, one or more candidate change point image patches are obtained through image patch extraction and filtering, including:
[0017] Based on the preset image block size k1×k1, the obtained two-phase difference images are cropped using a sliding window with a window size of k1×k1 and a step size of k1; if the cropped image size is less than k1×k1, zero-padding is performed to obtain an image block of k1×k1; k1 is a positive integer;
[0018] The obtained image block is binarized according to the preset threshold; the binarization process includes setting the pixels in the image block that exceed the corresponding threshold to 1 and the rest to 0.
[0019] Output the image patch with the non-zero matrix as the candidate change point image patch.
[0020] Optionally, the detection model includes two PCA layers, one output layer, and one SVM classification and detection module.
[0021] The first PCA layer is used to perform PCA convolution operation on the input candidate change point image patch and output a feature map.
[0022] The second PCA layer is used to perform PCA convolution operation on the feature map output by the first PCA layer again, and output a new feature map.
[0023] The output layer is used to perform layer normalization on the feature maps output by the first PCA layer and the second PCA layer respectively, and then merge them in the L2 dimension to obtain the final output feature map; where L2 represents the number of convolution kernels in the second PCA layer performing PCA convolution operation;
[0024] The SVM classification and detection module is used to perform classification and detection based on the final output feature map to obtain change detection results.
[0025] Optionally, the step of obtaining the final change-detected image of the target scene based on the change detection results through region filling and noise filtering includes:
[0026] Based on the change detection results, the changed regions are identified in the target scene image;
[0027] The closed operation is used to fill the determined change areas in the target scene image;
[0028] Median filtering is used to remove noise from the identified change areas in the target scene image, resulting in the final change detection image of the target scene.
[0029] Secondly, embodiments of the present invention also provide a damage assessment method based on infrared polarization imaging, comprising:
[0030] Using any of the scene change detection methods described above, the final change detection image result of the target scene is obtained;
[0031] Based on the final change detection image results of the target scene, the degree of damage is assessed.
[0032] Thirdly, embodiments of the present invention also provide a scene change detection device based on infrared polarization imaging, comprising:
[0033] The acquisition module is used to acquire infrared intensity, infrared polarization degree, and polarization angle images of the target scene before and after the change.
[0034] The registration module is used to perform feature registration on the infrared intensity, infrared polarization degree, and polarization angle images before and after the change, respectively, to obtain the registered infrared intensity, infrared polarization degree, and polarization angle image group.
[0035] The difference module is used to perform intra-group difference processing on the registered infrared intensity, infrared polarization degree and polarization angle image groups respectively to obtain three kinds of two-phase difference images corresponding to infrared intensity, infrared polarization degree and polarization angle.
[0036] The extraction module is used to extract and filter image patches based on the obtained two-phase difference images to obtain one or more candidate change point image patches.
[0037] The detection module is used to input the extracted candidate change point image patches into the trained detection model to obtain change detection results; the detection model is built based on the PCANet network and the SVM classification and detection module.
[0038] The post-processing module is used to obtain the final change detection image result of the target scene based on the change detection results, through region filling and noise filtering.
[0039] Fourthly, embodiments of the present invention also provide an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, it implements the method described in any embodiment of this specification.
[0040] Fifthly, embodiments of the present invention also provide a computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, causes the computer to perform the methods described in any embodiment of this specification.
[0041] This invention provides a method, apparatus, electronic device, and storage medium for scene change detection and damage assessment based on infrared polarization imaging. It can make full use of the infrared polarization intensity, degree of polarization, and polarization angle information of the target scene before and after the change, improve the accuracy of scene change detection, and provide a new technical approach for damage performance assessment. Attached Figure Description
[0042] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0043] Figure 1 This is a flowchart illustrating the steps of a scene change detection method based on infrared polarization imaging according to an embodiment of the present invention.
[0044] Figure 2 This is a schematic flowchart of a scene change detection method based on infrared polarization imaging provided in an embodiment of the present invention;
[0045] Figure 3 This is a hardware architecture diagram of an electronic device provided in an embodiment of the present invention;
[0046] Figure 4 This is a structural diagram of a scene change detection device based on infrared polarization imaging, provided in an embodiment of the present invention. Detailed Implementation
[0047] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.
[0048] Current optical imaging systems are highly sensitive to the natural environment and are easily affected by weather conditions such as rain, snow, and fog, making single-mode optical image information insufficient for accurate analysis of scene damage. Infrared polarization imaging technology utilizes the polarization characteristics of the target's radiated light waves to significantly enhance the difference between camouflaged or dimly lit targets and the background, filter out interference, improve signal contrast, highlight target contour information, and achieve high-precision extraction of scene change features. This has potential application advantages in the field of damage scene change detection.
[0049] Machine learning-based optical image detection methods, represented by deep learning, have attracted increasing attention. They employ a layered structure similar to neural networks, consisting of multiple layers such as convolutional layers, pooling layers, and fully connected layers, with different layers performing different functions. In the field of image change detection, deep learning technology is increasingly demonstrating its advantages in accuracy and robustness. It uses deep neural networks to learn feature representations suitable for change detection from images, thereby improving change detection performance.
[0050] However, while directly applying these deep learning change detection models to infrared polarization imaging can detect changes in the target scene, it fails to fully utilize the polarization information of different targets, thus affecting the detection accuracy.
[0051] In view of this, given the difficulty of fully utilizing the polarization information of the target to detect scene changes using existing technologies, this invention provides a scene change detection and damage assessment method based on infrared polarization imaging. This method can fully utilize the infrared polarization characteristics of the scene, improve the accuracy of scene change detection, and provide a new technical approach for damage performance assessment.
[0052] The following describes the specific implementation of the above concept.
[0053] Please refer to Figure 1 This invention provides a scene change detection method based on infrared polarization imaging, the method comprising:
[0054] Step 100: Obtain infrared intensity, infrared polarization degree, and polarization angle images of the target scene before and after the change;
[0055] Step 102: Perform feature registration on the infrared intensity, infrared polarization degree, and polarization angle images before and after the change, respectively, to obtain the registered infrared intensity, infrared polarization degree, and polarization angle image group.
[0056] This step 102 involves performing feature registration on the infrared intensity images before and after the change to obtain an infrared intensity image group, performing feature registration on the infrared polarization degree images before and after the change to obtain an infrared polarization degree image group, and performing feature registration on the polarization angle images before and after the change to obtain a polarization angle image group.
[0057] Step 104: Perform intra-group difference processing on the registered infrared intensity, infrared polarization degree, and polarization angle image groups respectively to obtain three kinds of two-phase difference images corresponding to infrared intensity, infrared polarization degree, and polarization angle.
[0058] Step 106: Based on the obtained two-phase difference images, one or more candidate change point image blocks are obtained through image block extraction and filtering.
[0059] Step 108: Input the extracted candidate change point image patch into the trained detection model to obtain the change detection result; the detection model is constructed based on the PCANet network and the SVM classification and detection module.
[0060] Step 110: Based on the change detection results, the final change detection image of the target scene is obtained through region filling and noise filtering. The final change detection image of the target scene can display the actual areas in the target scene that are identified as having changed in image form.
[0061] In this embodiment of the invention, infrared intensity, infrared polarization degree, and polarization angle images of the target scene before and after the change are used as detection input data. Combined with a deep learning detection model, automated target scene change detection is achieved, thereby quickly identifying the changed areas in the scene. This fully utilizes infrared polarization intensity, polarization degree, and polarization angle information to improve the accuracy of scene change detection and provides a new technical approach for assessing the damage effectiveness of scene changes caused by natural disasters, human-caused damage, and other factors.
[0062] The following description Figure 1 The execution method for each step is shown.
[0063] Optionally, for step 100, images of the infrared intensity, infrared polarization degree, and polarization angle of the target scene before and after the change are obtained, including:
[0064] Using a polarization detection system, the radiance of the target scene before and after the change in different polarization directions is measured, and the infrared intensity, infrared polarization degree, and polarization angle images of the target scene before and after the change are calculated respectively.
[0065] Among them, the radiance of the target scene before and after the change in different polarization directions can be measured by measuring three different polarization directions of 0°, 60° and 120° before and after the change, or by measuring four different polarization directions of 0°, 45°, 90° and 135° before and after the change. No further limitation is given here.
[0066] Using the above embodiments, the radiance in different polarization directions can be measured using a polarization detection system, and the infrared intensity, infrared polarization degree, and polarization angle images of the target scene before and after the change can be determined. In other embodiments, the infrared intensity, infrared polarization degree, and polarization angle images before and after the change can also be directly read.
[0067] Optionally, for step 102, feature registration is performed on the infrared intensity, infrared polarization degree, and polarization angle images before and after the change, including:
[0068] The MSER-SURF feature fusion registration method was used to perform feature registration on the infrared intensity, infrared polarization degree, and polarization angle images before and after the change.
[0069] Sensor shifts may occur when measuring data before and after changes. Feature registration can eliminate image differences introduced by external factors. The above embodiment uses the MSER-SURF feature fusion registration method to achieve feature registration, which has the advantages of fast processing speed and high processing accuracy, and is more suitable for infrared image feature registration.
[0070] For step 104, the registered infrared intensity, infrared polarization degree, and polarization angle image groups should be subjected to intra-group difference processing to maintain consistency, that is, both should be either the result after the change minus the result before the change, or both should be the result before the change minus the result after the change.
[0071] Optionally, for step 106, based on the obtained two-phase difference images, one or more candidate change point image patches are obtained through image patch extraction and filtering, including:
[0072] Based on the preset image block size k1×k1, the obtained two-phase differential images are cropped using a sliding window with a window size of k1×k1 and a step size of k1. The cropped image block has no missing or overlapping areas relative to the original two-phase differential images. If the cropped image size is less than k1×k1, zero-padding is performed to obtain a k1×k1 image block; k1 is a positive integer.
[0073] The obtained image block is binarized according to the preset threshold; the binarization process includes setting the pixels in the image block that exceed the corresponding threshold to 1 and the rest to 0.
[0074] Output the image patch with the non-zero matrix as the candidate change point image patch.
[0075] The above embodiment cropps the three types of two-phase difference images into k1×k1 image blocks, and then performs binarization processing. Taking the two-phase difference image corresponding to infrared intensity as an example, the preset corresponding threshold is a grayscale threshold. In the cropped image block, if the grayscale of a pixel is greater than the preset grayscale threshold, the pixel is set to 1; otherwise, it is set to 0, thus achieving binarization. Binarization processing can filter out pixels that have not changed or have only slightly changed in the temporal difference image. After binarization processing, an image block with an all-zero matrix can be considered to have no change information. The above processing is beneficial for redundancy removal, reducing computational load, and improving change detection efficiency.
[0076] Optionally, for step 108, the detection model includes two PCA layers, one output layer, and one SVM classification and detection module;
[0077] The first PCA layer is used to perform PCA convolution operation on the input candidate change point image patch and output a feature map.
[0078] The second PCA layer is used to perform PCA convolution operation on the feature map output by the first PCA layer again, and output a new feature map.
[0079] The output layer is used to perform layer normalization on the feature maps output by the first PCA layer and the second PCA layer respectively, and then merge them in the L2 dimension to obtain the final output feature map; L2 represents the number of convolutional kernels in the second PCA layer, which can usually be determined based on experience.
[0080] The SVM classification and detection module is used to perform classification and detection based on the final output feature map to obtain change detection results.
[0081] Principal Component Analysis (PCANet) network models are characterized by their simple structure and ease of training, and they can easily produce good results with a small number of samples. However, the output layer of PCANet takes too long to encode features, making it difficult to meet real-time requirements. Furthermore, change detection does not rely on deep features, and directly calling the model does not easily achieve ideal detection results. To address this issue, this invention proposes a detection model based on the Fast-PCANet network and an SVM classification and detection module, such as... Figure 2 As shown, a Fast-PCANet network is used, taking candidate change point image patches as input, and performing convolution operations on the input data using L1 PCA convolution filters of size k1×k1 to obtain N. block ×L1 feature maps of size k1×k1 are generated, which yields the feature map of PCA convolutional layer 1, where L1 represents the number of convolution kernels used in the first PCA layer for PCA convolution operations, and N... blockThis represents the number of input image patches; then, the feature map of PCA convolutional layer 1 is used as input, and L2 PCA convolutional filters of size L1×k2×k2 are used to perform convolution operations on the feature map of PCA convolutional layer 1 to obtain N. block The feature maps of size k1×k1 are obtained by generating L1×L2 feature maps, which represents the feature map of PCA convolutional layer 2. L2 represents the number of convolutional kernels in the second PCA layer. The feature maps of PCA convolutional layer 1 and PCA convolutional layer 2 are then merged in the L2 dimension after layer normalization to obtain the final output feature map. Finally, the final output feature map is classified and detected by support vector machine (SVM).
[0082] In the above embodiments, the present invention adjusts the output layer to obtain the Fast-PCANet network, which is designed for the characteristics of infrared polarization imaging change detection. Compared with the complex and time-consuming output layer form of the traditional PCANet network, the processing speed is significantly improved, which can effectively reduce the scene change detection time. Furthermore, it can fully integrate multi-dimensional infrared polarization information to improve the scene change detection accuracy and provide technical support for realizing real-time scene change detection.
[0083] Optionally, for step 110, based on the change detection results, the final change detection image result of the target scene is obtained through region filling and noise filtering, including:
[0084] Based on the change detection results, the changed regions are identified in the target scene image;
[0085] The closed operation is used to fill the determined change areas in the target scene image;
[0086] Median filtering is used to remove noise from the identified change areas in the target scene image, resulting in the final change detection image of the target scene.
[0087] The above embodiments, based on change detection results, process the final change detection image of the target scene for display through region filling and noise filtering, making the display results more intuitive. Furthermore, the final change detection image of the target scene can also be displayed in color to differentiate between missed and false detection areas.
[0088] This invention also provides a damage assessment method based on infrared polarization imaging, the method comprising:
[0089] Using the scene change detection method described in any of the above embodiments, the final change detection image result of the target scene is obtained;
[0090] Based on the final change detection image results of the target scene, the degree of damage is assessed.
[0091] Using the above embodiments, the degree of damage can be assessed based on the final change detection image results of the target scene. For example, the damage ratio can be assessed based on the area ratio of the changed region in the target scene, and the damage area can be assessed based on the specific location of the changed region in the target scene.
[0092] like Figure 3 , Figure 4 As shown, this embodiment of the invention provides a scene change detection device based on infrared polarization imaging. The device embodiment can be implemented through software, hardware, or a combination of both. From a hardware perspective, as... Figure 3 The diagram shown is a hardware architecture diagram of an electronic device for detecting scene changes based on infrared polarization imaging, provided in an embodiment of the present invention. Except for... Figure 3 In addition to the processor, memory, network interface, and non-volatile memory shown, the electronic device in the embodiment may also include other hardware, such as a forwarding chip responsible for processing packets. Taking software implementation as an example, such as... Figure 4 As shown, a device in a logical sense is formed by the CPU of its host electronic device reading the corresponding computer program from non-volatile memory into memory and running it. This embodiment provides a scene change detection device based on infrared polarization imaging, comprising:
[0093] The acquisition module 401 is used to acquire infrared intensity, infrared polarization degree and polarization angle images of the target scene before and after the change;
[0094] The registration module 402 is used to perform feature registration on the infrared intensity, infrared polarization degree, and polarization angle images before and after the change, respectively, to obtain a group of registered infrared intensity, infrared polarization degree, and polarization angle images.
[0095] The differential module 403 is used to perform intra-group differential processing on the registered infrared intensity, infrared polarization degree and polarization angle image groups respectively to obtain three kinds of two-phase differential images corresponding to infrared intensity, infrared polarization degree and polarization angle.
[0096] The extraction module 404 is used to extract and filter image blocks based on the obtained two-time-phase difference images to obtain one or more candidate change point image blocks.
[0097] Detection module 405 is used to input the extracted candidate change point image patch into the trained detection model to obtain change detection results; the detection model is built based on PCANet network and SVM;
[0098] The post-processing module 406 is used to obtain the final change detection image result of the target scene by region filling and noise filtering based on the change detection result.
[0099] In this embodiment of the invention, the acquisition module 401 can be used to execute step 100 in the above method embodiment, the registration module 402 can be used to execute step 102 in the above method embodiment, the difference module 403 can be used to execute step 104 in the above method embodiment, the extraction module 404 can be used to execute step 106 in the above method embodiment, the detection module 405 can be used to execute step 108 in the above method embodiment, and the post-processing module 406 can be used to execute step 110 in the above method embodiment.
[0100] It is understood that the structures illustrated in the embodiments of the present invention do not constitute a specific limitation on a scene change detection device based on infrared polarization imaging. In other embodiments of the present invention, a scene change detection device based on infrared polarization imaging may include more or fewer components than illustrated, or combine some components, or split some components, or arrange different components. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.
[0101] The information interaction and execution process between the modules in the above-mentioned device are based on the same concept as the method embodiment of the present invention, and the specific details can be found in the description of the method embodiment of the present invention, and will not be repeated here.
[0102] This invention also provides an electronic device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements a scene change detection or damage assessment method based on infrared polarization imaging according to any embodiment of this invention.
[0103] This invention also provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program causes the processor to perform a scene change detection or damage assessment method based on infrared polarization imaging according to any embodiment of this invention.
[0104] Specifically, a system or apparatus equipped with a storage medium may be provided, on which software program code implementing the functions of any of the embodiments described above is stored, and the computer (or CPU or MPU) of the system or apparatus may read and execute the program code stored in the storage medium.
[0105] In this case, the program code read from the storage medium can itself implement the function of any of the above embodiments, and therefore the program code and the storage medium storing the program code constitute part of the present invention.
[0106] Examples of storage media used to provide program code include floppy disks, hard disks, magneto-optical disks, optical disks (such as CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD+RW), magnetic tapes, non-volatile memory cards, and ROMs. Alternatively, program code can be downloaded from a server computer via a communication network.
[0107] Furthermore, it should be clear that not only can the program code read by the computer be executed, but also the operating system or other components operating on the computer can be instructed based on the program code to perform some or all of the actual operations, thereby realizing the function of any of the embodiments described above.
[0108] Furthermore, it is understood that the program code read from the storage medium is written to the memory set in the expansion board inserted into the computer or to the memory set in the expansion module connected to the computer. Then, based on the instructions of the program code, the CPU or other components installed on the expansion board or expansion module execute some and all of the actual operations, thereby realizing the function of any of the above embodiments.
[0109] In summary, this invention provides a method, device, electronic device, and storage medium for scene change detection and damage assessment based on infrared polarization imaging. By combining multi-dimensional infrared polarization information with a fast detection deep learning network model, it fully utilizes the infrared polarization characteristics of the scene and adjusts the deep learning network model for change detection, thereby reducing scene change detection time and improving scene change detection accuracy, providing a new technical approach for real-time damage performance assessment.
[0110] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0111] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media that can store program code, such as ROM, RAM, magnetic disk, or optical disk.
[0112] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A scene change detection method based on infrared polarization imaging, characterized in that, include: Acquire infrared intensity, infrared polarization degree, and polarization angle images of the target scene before and after the change; Feature registration was performed on the infrared intensity, infrared polarization degree, and polarization angle images before and after the change, respectively, to obtain the registered infrared intensity, infrared polarization degree, and polarization angle image group; Intra-group difference processing was performed on the registered infrared intensity, infrared polarization degree and polarization angle image groups to obtain three kinds of two-phase difference images corresponding to infrared intensity, infrared polarization degree and polarization angle. Based on the obtained two-phase difference images, one or more candidate change point image blocks are obtained through image block extraction and filtering. The extracted candidate change point image patches are input into the trained detection model to obtain change detection results; The detection model is built based on the PCANet network and the SVM classification and detection module; Based on the change detection results, the final change detection image of the target scene is obtained through region filling and noise filtering.
2. The scene change detection method according to claim 1, characterized in that, The acquisition of infrared intensity, infrared polarization degree, and polarization angle images of the target scene before and after the change includes: Using a polarization detection system, the radiance of the target scene before and after the change in different polarization directions is measured, and the infrared intensity, infrared polarization degree, and polarization angle images of the target scene before and after the change are calculated respectively.
3. The scene change detection method according to claim 1, characterized in that, The feature registration of the infrared intensity, infrared polarization degree, and polarization angle images before and after the change includes: The MSER-SURF feature fusion registration method was used to perform feature registration on the infrared intensity, infrared polarization degree, and polarization angle images before and after the change.
4. The scene change detection method according to claim 1, characterized in that, Based on the obtained two-phase difference images, one or more candidate change point image patches are obtained through image patch extraction and filtering, including: Based on the preset image block size k1×k1, the obtained two-phase difference images are cropped using a sliding window with a window size of k1×k1 and a step size of k1; if the cropped image size is less than k1×k1, zero-padding is performed to obtain an image block of k1×k1; k1 is a positive integer; The obtained image block is binarized according to the preset threshold; the binarization process includes setting the pixels in the image block that exceed the corresponding threshold to 1 and the rest to 0. Output the image patch with the non-zero matrix as the candidate change point image patch.
5. The scene change detection method according to claim 4, characterized in that, The detection model includes two PCA layers, one output layer, and one SVM classification and detection module. The first PCA layer is used to perform PCA convolution operation on the input candidate change point image patch and output a feature map. The second PCA layer is used to perform PCA convolution operation on the feature map output by the first PCA layer again, and output a new feature map. The output layer is used to perform layer normalization on the feature maps output by the first PCA layer and the second PCA layer respectively, and then merge them in the L2 dimension to obtain the final output feature map; where L2 represents the number of convolution kernels in the second PCA layer performing PCA convolution operation; The SVM classification and detection module is used to perform classification and detection based on the final output feature map to obtain change detection results.
6. The scene change detection method according to claim 1, characterized in that, The process of obtaining the final change-detected image of the target scene based on the change detection results, through region filling and noise filtering, includes: Based on the change detection results, the changed regions are identified in the target scene image; The closed operation is used to fill the determined change areas in the target scene image; Median filtering is used to remove noise from the identified change areas in the target scene image, resulting in the final change detection image of the target scene.
7. A damage assessment method based on infrared polarization imaging, characterized in that, include: Using the scene change detection method as described in any one of claims 1-6, the final change detection image result of the target scene is obtained; Based on the final change detection image results of the target scene, the degree of damage is assessed.
8. A scene change detection device based on infrared polarization imaging, characterized in that, include: The acquisition module is used to acquire infrared intensity, infrared polarization degree, and polarization angle images of the target scene before and after the change. The registration module is used to perform feature registration on the infrared intensity, infrared polarization degree, and polarization angle images before and after the change, respectively, to obtain the registered infrared intensity, infrared polarization degree, and polarization angle image group. The difference module is used to perform intra-group difference processing on the registered infrared intensity, infrared polarization degree and polarization angle image groups respectively to obtain three kinds of two-phase difference images corresponding to infrared intensity, infrared polarization degree and polarization angle. The extraction module is used to extract and filter image patches based on the obtained two-phase difference images to obtain one or more candidate change point image patches. The detection module is used to input the extracted candidate change point image patches into the trained detection model to obtain change detection results; The detection model is built based on the PCANet network and the SVM classification and detection module; The post-processing module is used to obtain the final change detection image result of the target scene based on the change detection results through region filling and noise filtering.
9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the method as described in any one of claims 1-7.
10. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed in the computer, it causes the computer to perform the method of any one of claims 1-7.