A hyperspectral image enhancement method and device based on two-dimensional stochastic resonance

Abstract

The application provides a hyperspectral image enhancement method and equipment based on two-dimensional random resonance, wherein the method comprises the following steps: step 1, obtaining a hyperspectral image data set; step 2, normalizing the values of all sample points of the hyperspectral image data; step 3, inputting the data of each wave band of the normalized hyperspectral image into a two-dimensional random resonance model 2D DSR for processing, processing the signal of the pixel point by comprehensively utilizing the information of adjacent pixel points, and obtaining enhanced data of each pixel point; and step 4, normalizing the enhanced hyperspectral image data to obtain a final hyperspectral image enhancement result data set. The application popularizes the use of DSR to two-dimensional data processing, can fully utilize the spatial position information, and the output of each pixel point is updated four times, the information in the four directions of up, down, left and right is comprehensively utilized, the signal is effectively enhanced, and the information expression capability is improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of hyperspectral image processing, and particularly relates to a hyperspectral image enhancement method and equipment based on two-dimensional stochastic resonance. BACKGROUND

[0002] Hyperspectral image (HSI) has a wide range of applications in urban planning, forest condition evaluation, biomedicine and many other fields due to its rich spatial and spectral information. Therefore, the detection and classification of hyperspectral images have become an important task, and improving the image quality and the performance of the classification network to improve the classification accuracy of the hyperspectral image has become an important research content in the field.

[0003] The current method for obtaining information-rich high-quality hyperspectral images can be roughly divided into two kinds. One method is to use high-performance image acquisition equipment on the hardware, improve the image quality through parameter setting and long exposure, but the method of improving the hardware setting has high equipment cost. The other method is to enhance the image by constructing intelligent algorithm, which can be roughly divided into: distribution mapping, model optimization and deep learning based methods. The distribution mapping based method improves the image brightness and clarity by improving the pixel distribution of the image, but often ignores the intrinsic relationship within the distribution. The existing traditional algorithm based on model optimization improves the signal-to-noise ratio by suppressing noise, which often produces problems of insufficient exposure, color unsaturation and obvious noise artifacts. The existing one-dimensional stochastic resonance theory for hyperspectral image enhancement using noise cannot effectively utilize the position information between each pixel point in the image. The deep learning based method can effectively enhance the image information, but the current hyperspectral image training dataset containing shadow area for information enhancement is insufficient, and a lot of time is needed for data preparation work. SUMMARY

[0004] To solve the above problems, the first aspect of the application provides a hyperspectral image shadow enhancement method, which comprises the following steps:

[0005] Step 1, obtaining a hyperspectral image dataset;

[0006] Step 2, normalizing the values of all sample points of the hyperspectral image data;

[0007] Step 3, inputting the data of each band of the normalized hyperspectral image into a two-dimensional stochastic resonance model 2DDSR for processing, comprehensively processing the signals of the pixel points by using the information of adjacent pixel points, and obtaining the enhanced data of each pixel point; the formula of the two-dimensional stochastic resonance model is as follows:

[0008]

[0009] wherein i, j represent the horizontal and vertical coordinates of the input signal respectively, i = (1, 2, 3,..., h), j = (1, 2, 3,..., w), h and w are the length and width of the input two-dimensional data respectively; u i,j,k (n) represents the output of the signal at (i, j) in the kth band after n iterations in the system; f i,j,k represents the hyperspectral image signal input with noise at (i, j) in the kth band; t x , t y respectively represent the step size in the horizontal and vertical directions; a and b are system parameters;

[0010] Step 4, normalize the enhanced hyperspectral image data to obtain the final hyperspectral image enhancement result data set.

[0011] In a possible design, the normalization of the values of all sample points in the hyperspectral image data in step 2 is specifically: subtracting the minimum value of all sample points from all sample points in the data, and then dividing by the difference between the maximum value and the minimum value of all sample points, and the formula is as follows:

[0012]

[0013] wherein H is the original hyperspectral image data set, and I is the normalized data.

[0014] In a possible design, the specific processing process of step 3 is: using a two-dimensional stochastic resonance model to process the data of each band of the normalized hyperspectral image, sampling with a 2*2 pixel window, and the sliding mode is from the second row and the second column to the last column, and then to the next row to repeat the operation until the last row, and processing each pixel by two-dimensional stochastic resonance; using the four-way parallel formula in the two-dimensional stochastic resonance model to calculate u(i, j) in each window, in each window, the calculation of the direction corresponding to the formula in the equation group is repeated, and after scanning all the pixel enhancement processing, the output of each pixel point is updated four times, and the information of the four directions above, below, left and right is integrated to obtain the hyperspectral image enhancement processing data.

[0015] In a possible design, the calculation of the direction corresponding to the formula in the equation group in each window is specifically: the image signal f i,j-1 at (i, j) on the left (i, j-1) is used to calculate the first output u i,j-1 (n+1) of (i, j) through the n th iteration of the output u i,j (n) of (i, j) and the image input signal f i,j at (i, j) of the point; and the image input signal f i,j, the system output at (i, j-1) is updated again to obtain u i,j-1 (n+1) of the (i, j) point after the update i,j (n) and the image input signal f i,j of the point, to obtain the output u i-1,j (n+1) of the (i-1, j) point above (i, j) i-1,j after the update of the image signal f i-1,j (n) of the point and the output u i,j (n+1) of the (i, j) point after the second update of the point; when the current window position is moved right by one unit, the output u i,j+1 (n) of the (i, j+1) point and the image input signal f i,j+1 of the point, the system output at (i, j) is updated again to obtain u i,j (n+1) of the (i+1, j) point; when the current window position is moved down by one unit, the output u i+1,j (n) of the (i+1, j) point and the image input signal f i+1,j of the point, to obtain the fourth updated output u i,j (n+1) of the (i, j) point above (i+1, j); finally, the system output of each pixel point in the 2th to h+1th row and the 2th to w+1th column is obtained, and the output of each pixel point is updated four times, and the information in the four directions above, below, left and right is integrated.

[0016] In a possible design, for a hyperspectral image dataset with large noise influence and weak signal strength in a shadow part, before step 3, the normalized hyperspectral image data is extracted to obtain a two-dimensional data group of the shadow region of the hyperspectral image in the data, and the two-dimensional data group of the extracted shadow region is filled with data; meanwhile, finally, the enhanced shadow region and the non-shadow region are fused to obtain the hyperspectral image dataset after enhancement of the shadow region.

[0017] The second aspect of the present application further provides a hyperspectral image enhancement device based on two-dimensional stochastic resonance, the device comprising at least one processor and at least one memory, the processor and the memory being coupled; the memory stores a computer execution program; when the processor executes the computer execution program stored in the memory, the processor can execute the hyperspectral image enhancement method as described in the first aspect.

[0018] The third aspect of the present application further provides a computer readable storage medium, the computer readable storage medium stores a computer program or instructions, and when the program or instructions are executed by a processor, the processor can execute the hyperspectral image enhancement method as described in the first aspect.

[0019] Compared with the prior art, the application mainly provides a hyperspectral image enhancement method based on two-dimensional random resonance, and the application of DSR is popularized to two-dimensional data processing.

[0020] As a three-dimensional tensor, each pixel in the hyperspectral image has certain correlation with its neighborhood pixels. The ordinary one-dimensional DSR samples the image from one direction, and cannot effectively utilize the position information between the pixels in the image. The two-dimensional random resonance-based hyperspectral image enhancement method in the application can fully utilize the spatial position information through four-way scanning, the output of each pixel point is updated four times, the information in the four directions above, below, left and right is integrated, and the effect is more obvious through iterative enhancement. The normalization processing restores the image data features and makes the enhanced result data more balanced, visually improves the brightness of the image, effectively enhances the signal and improves the information expression capability. BRIEF DESCRIPTION OF DRAWINGS

[0021] Figure 1 It is a flowchart of the hyperspectral image enhancement and classification in embodiment 1.

[0022] Figure 2 It is a color image diagram and a binary image diagram of the shadow area mask of the HYDICE data set used for verifying the application in embodiment 1.

[0023] Figure 3 It is a signal processing diagram of two-dimensional random resonance on each pixel point in the two-dimensional data of the single waveband of the hyperspectral image in embodiment 1.

[0024] Figure 4 It is a comparison diagram of the hyperspectral image original data, 1D DSR spectral dimension enhancement and 2D DSR enhanced data in the first waveband gray scale diagram in embodiment 1.

[0025] Figure 5 It is a comparison diagram of the spectral curves of the road and grass in the shadow area before and after enhancement and the spectral curves of the road and grass in the non-shadow area in embodiment 1.

[0026] Figure 6 It is a flowchart of the ECA-3D CNN algorithm for classifying the hyperspectral image in embodiment 1.

[0027] Figure 7 It is a comparison diagram of the classification results of the original data, 1D DSR enhanced data and 2D DSR enhanced data by 3D CNN and the classification result of the 2D DSR enhanced data by ECA-3D CNN in embodiment 1.

[0028] Figure 8 It is a simple block diagram of the structure of the hyperspectral image shadow enhancement device in embodiment 2. DETAILED DESCRIPTION

[0029] The application will be further described in connection with specific embodiments.

[0030] Embodiment 1

[0031] The data set used in this embodiment has a large shadow noise impact and weak signal strength, so the data in the shadow part is extracted for enhancement. The method in the application can be used for enhancing all weak signal images containing noise.

[0032] The specific operation process is as shown in the flowchart. Figure 1 The flowchart is shown as follows:

[0033] Step 1: The hyperspectral image data affected by noise and having a shadow area is taken as an example, and the shadow area of the data is enhanced.

[0034] Data set introduction

[0035] The hyperspectral image data is three-dimensional, that is, a two-dimensional data set. The data set is collected by a hyperspectral digital image collection experiment (HYDICE) in August 1995, the threshold of each pixel point is [0, 1], the size is: 316*216*148 (h, w, b), the spatial resolution is 0.75 m, the spectral resolution is 10 nm, and there are 8 labels, which are: grass, tree, road, shadowed road, shadowed grass, target 1, target 2 and target 3. The label data is shown in Table 1.

[0036] Table 1 Label data

[0037] Serial number Sample number Category 1 33184 Grass 2 10850 Tree 3 3376 Road 4 1686 Road under shadow 5 323 Grass under shadow 6 537 Target 1 7 514 Target 2 8 4135 Target 3

[0038] Step 2: The obtained hyperspectral image data is normalized. Specifically, all sample points in the data are subtracted by the minimum value of all sample points and divided by the difference between the maximum value and the minimum value of all sample points.

[0039]

[0040] Wherein, H is the original hyperspectral image data set, and I is the normalized data.

[0041] Reason: Dynamic stochastic resonance requires small parameters.

[0042] Step 3: The two-dimensional data set of the shadow area of the hyperspectral image in the normalized hyperspectral image data is extracted.

[0043] Specifically, according to the labels 4 and 5 of the shadow area in the data set, a shadow mask is set, and the schematic diagram of the three-channel color image of the original hyperspectral image data and the binary image of the shadow area mask are shown as follows: Figure 2The normalized hyperspectral image data is multiplied by the mask;

[0044] F = I x mask

[0045] Wherein, F is the extracted shadow area, mask is the shadow mask with pixel point values only 0 and 1, and the figure is a binary image display.

[0046] Step four: data padding is performed on the extracted shadow area two-dimensional data set.

[0047] Specifically, the size of the extracted shadow area two-dimensional data set is expanded to (h+2, w+2, b) with 0.

[0048] Reason: when using two-dimensional stochastic resonance for enhancement, the spatial information of the adjacent pixel points in the four directions above, below, left and right needs to be used, and the edges of the two-dimensional image data lack 1-2 direction pixel points and corresponding spatial information, in order to avoid separate calculation of the output of the edge pixel points, the edges of the two-dimensional data are filled.

[0049] Step five: based on the two-dimensional data set of the filled hyperspectral shadow area obtained in step four, for each pixel point in each two-dimensional data, the information of the adjacent pixel points is used to process the signal of the pixel point by using a two-dimensional stochastic resonance image processing system, and the enhancement data of each pixel point is obtained.

[0050] The two-dimensional stochastic resonance image processing model is as follows:

[0051]

[0052] Wherein, i = 1, 2,..., w, j = 1, 2,..., h, represent the horizontal and vertical coordinates of the input signal; u i,j,k (n) is the output of the signal of the kth band at (i, j) after n iterations in the system; f i,j,k represents the input of the hyperspectral image signal with noise in the kth band at (i, j); t x y represents the step size in the horizontal and vertical directions; a, b are system parameters.

[0053] Specifically, first, set the initial value of the parameter t x = t y ​=a=b=0.01, n=1. For each two-dimensional image data in the filled two-dimensional data set, starting from the first column of the first row, sampling is performed with a window unit of 2*2 pixels and a step of 1 pixel. The sampling window is slid along the x-axis to the (w+2)th column, and then the sampling is repeated from the first to the (w+2)th column until the last row. The u(i,j) in each window is calculated using the four-way parallel formula in two-dimensional stochastic resonance.

[0054] Combining the system of equations and Figure 3 In each window, the calculations for the directions corresponding to the formulas in the system of equations are repeated. That is, the image signal f at position (i,j-1) to the left of (i,j) is calculated. i,j-1 and the output u of the nth iteration at that point. i,j-1 (n) Calculate the first output u of point (i,j). i,j (n+1); that is, the output u of point (i,j) i,j (n) and the image input signal f at that point i,j The system output at (i,j-1) is updated to obtain u. i,j-1 (n+1); that is, the output u of point (i,j) obtained from the update. i,j (n) and the image input signal f at that point i,j The output u of the point (i-1,j) above (i,j) is obtained. i-1,j (n+1); that is, the image signal f at position (i-1,j) above the updated (i,j). i-1,j and the output u of the nth iteration at that point. i-1,j (n) Output u again for point (i,j) i,j The output of the second update is obtained by updating (n+1). Finally, the system output of each pixel in rows 2 to h+1 and columns 2 to w+1 is obtained. The output of each pixel has been updated four times, combining information from the four directions of up, down, left, and right.

[0055] Step Six: Normalize the data of the shadow regions in the enhanced hyperspectral image, using the same normalization method as in Step Two. The specific two-dimensional data set of the enhanced shadow regions is as follows:

[0056]

[0057] Where U represents the shadow region data of all bands of the hyperspectral image after 2D DSR processing.

[0058] Reason for normalization: To unify the scale range of the processed data and the original data.

[0059] Step 7: Fuse the enhanced shadow regions with the non-shadow regions to obtain a hyperspectral image dataset with enhanced shadow regions.

[0060] Specifically, a non-shading area mask is obtained by subtracting the shading area mask from a matrix with all values being 1 and the same size as the shading mask, the non-shading area mask is multiplied by the normalized original data set to obtain a non-shading area, and then the non-shading area is added to the enhanced shading area to obtain the hyperspectral image data set after enhancement of the shading area.

[0061] I' = I x (1-mask) + f 2D-DSR (F)

[0062] I' is the final hyperspectral image enhancement result data set after enhancement of the shading area.

[0063] Step eight: the principal component analysis (PCA) is used to reduce the dimension of the hyperspectral image data set after enhancement of the shading area, and 10 principal components most capable of representing the characteristics of the original data are extracted.

[0064] Specifically, the average value of each feature in the enhanced hyperspectral image data set is obtained, and the corresponding average value is subtracted for all samples; the covariance matrix is solved, and the eigenvalues and eigenvectors of the matrix are calculated; the eigenvalues and their corresponding eigenvectors are rearranged from large to small, and the l eigenvectors with the largest eigenvalues are selected; the sample points after subtracting the average value are projected onto the l eigenvectors to obtain the reduced data.

[0065] The reason for this step: the hyperspectral image has a high dimension, and the information of different bands is often repeated and redundant, so extracting principal components in this step is beneficial to feature extraction in the subsequent steps.

[0066] Step nine: a window size is set for the reduced data to be cut off for processing as input of the classification network, and the window size is m*m, and m=11 in the example of the present application.

[0067] Specifically, the reduced data is first expanded to (h+m-1, w+m-1, l) in the spatial dimension with 0, and then the image data is cut into data groups with a size of m*m*l with a step of m.

[0068] Step ten: the 3D CNN hyperspectral image classification network model is used to classify the enhanced data, and the classification result is obtained.

[0069] Specifically, the divided training set and test set after cutting data, the proportion of training set and test set in the example of the application is 2:8; the network model of 3D CNN hyperspectral image classification is built, the model structure and parameter setting are as shown in table two and table three; the network model is trained by using the training set, the 3D CNN network model with trained parameters is obtained, the model is used for classifying the cut data, the classification result and overall classification accuracy (OA) are obtained, and the OA is calculated as follows.

[0070]

[0071] Wherein, N is the number of categories, M i,j Indicates the number of samples belonging to category i in the test sample is misclassified as category j.

[0072] Table two 3D CNN network structure

[0073] Activation function Convolution kernel size Number of neurons Dropout 3D Conv ReLU 3*3*3 8 - Max pooling - 2*2*2 - - Flatten - - - - Dense ReLU - 256 0.4 Dense ReLU - 128 0.4

[0074] Table three 3D CNN parameter setting

[0075]

[0076] Step eleven: the OA obtained in step ten is used as the evaluation index of image signal enhancement effect, the parameters t x ,t y ,a,b,n (the better the feature expression ability of image information, the higher the classification accuracy, so the higher classification accuracy can indicate that the corresponding parameter combination in the two-dimensional random resonance system is better), and the optimal enhancement result after parameter optimization is obtained.

[0077] Specifically, under the initial value setting of the parameters in step five, n is manually adjusted, and the steps after step five are repeated, the classification results of data enhancement under different iteration numbers are obtained, the optimal iteration number 5 is obtained by comparing OA, and then the parameters t x ,t y ,a,b are adjusted, the step is 0.01, the intermediate steps are repeated, the OA of the enhancement results under different parameter conditions is compared, and finally the optimal enhancement result is obtained when the parameter combination is t x =t y =a=b=0.01, n=5.

[0078] Figure 4 It is a gray scale comparison of the original image data, 1D DSR enhanced data and 2D DSR enhanced data in the first waveband;

[0079] It can be seen from the image that the signals of the shadow area of the hyperspectral image are effectively enhanced under the action of 2D DSR, and the effect is obviously better than that of 1D DSR.

[0080] Figure 5 The spectral curves of the road and grass in the shadow area of the hyperspectral image before and after enhancement are compared with the spectral curves of the road and grass in the non-shadow area.

[0081] Through Figure 5 The spectral curve analysis of the pixels of the road in (a) shows that the spectral information intensity is obviously improved, the information expression ability is better, and the signal intensity is effectively improved compared with before 2D DSR; Figure 5 In (b), the spectral information of the grass in the shadow area is obviously changed under the action of 2D DSR, and the change trend approaches the change trend of the spectral curve of the pixels of the same category in the non-shadow area, indicating that the spectral information of the hyperspectral image is effectively improved.

[0082] Step twelve: optimize the structure of the classification network model, classify the enhanced data after parameter optimization in step eleven, and obtain better classification results of the optimal enhanced data after model optimization.

[0083] Specifically, the 3D CNN established in step ten is combined with efficient channel attention (ECA) to optimize the network model. In the attention module, the global average pooling extracts channel attention information, the one-dimensional convolution layer summarizes the cross-channel information, obtains accurate attention, and uses the activation function Sigmoid to obtain the weight, as Figure 6 The classification results are compared with the classification results of the 3D CNN without combining ECA, and the classification indicators are overall classification accuracy OA, average classification accuracy AA and Kappa coefficient. The calculation formulas of the classification indicators are as follows:

[0084]

[0085]

[0086] Wherein, P is the total number of test samples.

[0087] Table four: comparison of classification results

[0088]

[0089] Table four is the classification result accuracy of the original data, 1D DSR enhanced data, 2D DSR enhanced data under 3D CNN, and the classification result of the improved 3D CNN combined with ECA. By comparing various indicators, it can be found that, similarly, 3D CNN is used for classification, the OA of the 2D DSR enhanced result data is about 1% higher than that of the original data, which shows that 2D DSR has obvious effect on signal enhancement; compared with the 1D DSR enhanced result, the overall classification accuracy of 2D DSR is higher, about 0.32%, which shows that 2D DSR has more advantages in improving image signal strength and feature expression ability, and can better utilize spatial information to enhance the signal. Similarly, the classification of 2D DSR enhanced result data is carried out, and compared with 3D CNN, the OA of ECA-3D CNN is improved by about 0.2%, and the improved network has better feature extraction ability, and the network classification performance is improved.

[0090] Figure 7 Table four is the classification result accuracy of the original data, 1D DSR enhanced data, 2D DSR enhanced data under 3D CNN, and the classification result of the improved 3D CNN combined with ECA. By comparing various indicators, it can be found that, similarly, 3D CNN is used for classification, the OA of the 2D DSR enhanced result data is about 1% higher than that of the original data, which shows that 2D DSR has obvious effect on signal enhancement; compared with the 1D DSR enhanced result, the overall classification accuracy of 2D DSR is higher, about 0.32%, which shows that 2D DSR has more advantages in improving image signal strength and feature expression ability, and can better utilize spatial information to enhance the signal. Similarly, the classification of 2D DSR enhanced result data is carried out, and compared with 3D CNN, the OA of ECA-3D CNN is improved by about 0.2%, and the improved network has better feature extraction ability, and the network classification performance is improved.

[0091] Example 2:

[0092] As Figure 8As shown, the application also provides a hyperspectral image enhancement device based on two-dimensional stochastic resonance, which comprises at least one processor and at least one memory, and further comprises a communication interface and an internal bus; the memory stores a computer execution program; when the processor executes the computer execution program stored in the memory, the processor can execute the hyperspectral image enhancement method as described in Embodiment 1. The internal bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component (PCI) bus, an.Xtended Industry Standard Architecture (EISA) bus, or the like. The bus can be divided into an address bus, a data bus, a control bus, and the like. For the convenience of representation, the bus in the drawings of the present application does not limit to only one bus or one type of bus. The memory can include a high-speed RAM memory, and can also include a non-volatile storage NVM, for example, at least one disk memory, and can also be a U disk, a mobile hard disk, a read-only memory, a magnetic disk or an optical disk, and the like.

[0093] The device can be provided as a terminal, a server or other forms of devices.

[0094] Figure 8 is a block diagram of a device shown by way of example. The device can include one or more of the following components: a processing component, a memory, a power supply component, a multimedia component, an audio component, an input / output (I / O) interface, a sensor component, and a communication component. The processing component usually controls overall operations of the electronic device, such as operations associated with display, telephone calling, data communication, camera operation and recording operation. The processing component can include one or more processors to execute instructions to complete all or part of the steps of the above-mentioned methods. In addition, the processing component can include one or more modules to facilitate the interaction between the processing component and other components. For example, the processing component can include a multimedia module to facilitate the interaction between the multimedia component and the processing component.

[0095] The memory is configured to store various types of data to support the operation of the electronic device. Examples of these data include instructions for any application or method operating on the electronic device, contact data, phonebook data, messages, pictures, videos, and the like. The memory can be realized by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0096] The power component provides power to various components of the electronic device. The power component can include a power management system, one or more power sources, and other components associated with generating, managing, and distributing power for the electronic device. The multimedia component includes a screen to provide an output interface between the electronic device and a user. In some embodiments, the screen can include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen can be implemented as a touch screen to receive input signals from a user. The touch panel includes one or more touch sensors to sense touch, swiping, and gestures on the touch panel. The touch sensors can not only sense a boundary of a touch or swiping action, but also detect duration and pressure associated with the touch or swiping action. In some embodiments, the multimedia component includes a front camera and / or a rear camera. When the electronic device is in an operating mode, such as a shooting mode or a video mode, the front camera and / or the rear camera can receive external multimedia data. Each of the front and rear cameras can be a fixed optical lens system or have a focal length and optical zoom capability.

[0097] The audio component is configured to output and / or input audio signals. For example, the audio component includes a microphone (MIC) to receive an external audio signal when the electronic device is in an operating mode, such as a call mode, a recording mode, and a voice recognition mode. The received audio signal can be further stored in the memory or transmitted via the communication component. In some embodiments, the audio component further includes a speaker to output audio signals.

[0098] The sensor component includes one or more sensors to provide various state assessments for the electronic device. For example, the sensor component can detect an open / closed state of the electronic device, relative positioning of components, such as a display and a keypad of the electronic device, a change in position of the electronic device or a component of the electronic device, presence or absence of user contact with the electronic device, the electronic device orientation or acceleration / deceleration, and a temperature change of the electronic device. The sensor component can include a proximity sensor configured to detect presence of a nearby object without any physical touch. The sensor component can further include a light sensor, such as a CMOS or CCD image sensor, for use in an imaging application. In some embodiments, the sensor component can further include an acceleration sensor, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

[0099] The communication component is configured to facilitate wired or wireless communication between the electronic device and other devices. The electronic device can access a wireless network based on a communication standard, such as WiFi, 2G, or 3G, or a combination thereof. In an example embodiment, the communication component receives a broadcast signal or broadcast related information from an external broadcast management system via a broadcast channel. In an example embodiment, the communication component also includes a Near Field Communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on Radio Frequency Identification (RFID) technology, Infrared Data Association (IrDA) technology, Ultra-WideBand (UWB) technology, Bluetooth (BT) technology, and other technologies.

[0100] In an example embodiment, the electronic device can be implemented with one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, micro-controllers, microprocessors, or other electronic elements to perform the above-described methods.

[0101] Embodiment 3:

[0102] The present application also provides a computer readable storage medium, which stores a computer program or instructions, and the program or instructions, when executed by a processor, can cause the processor to perform the hyperspectral image enhancement method according to Embodiment 1.

[0103] Specifically, a system, device or apparatus equipped with a readable storage medium can be provided, and the readable storage medium stores software program codes for implementing the functions of any of the above-described embodiments, and the computer or processor of the system, device or apparatus reads and executes the instructions stored in the readable storage medium. In this case, the program codes read from the readable medium can implement the functions of any of the above-described embodiments, and thus the machine readable codes and the readable storage medium storing the machine readable codes constitute a part of the present application.

[0104] The storage medium described above can be implemented by any type of volatile or non-volatile storage devices or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk (such as CD-ROM, CD-R, CD-RW, DVD-20ROM, DVD-RAM, DVD-RW, DVD-RW), magnetic tape, etc. The storage medium can be any available medium that can be accessed by a general or special purpose computer.

[0105] It is to be understood that the above-mentioned processor can be a central processing unit (CPU), but can also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in combination with the application can be directly embodied as hardware processor execution, or executed by a combination of hardware and software modules in the processor.

[0106] It is to be understood that the storage medium is coupled to the processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be an integral part of the processor. The processor and the storage medium can be located in an application specific integrated circuit (ASIC). Of course, the processor and the storage medium can also exist as discrete components in the terminal or server.

[0107] The computer readable program instructions described herein can be downloaded from a computer readable storage medium to each of the various computing / processing devices, or to external computers or external storage devices through a network, for example, the Internet, a local area network, a wide area network and / or a wireless network. The network can include copper transmission cables, optical transmission fibers, wireless transmission, routers, firewalls, switches, gateway computers and / or edge servers. The network adapter card or network interface in each computing / processing device receives computer readable program instructions from the network and forwards the computer readable program instructions to the computer readable storage medium in each computing / processing device for storage.

[0108] Computer readable program instructions for carrying out operations of the present disclosure can be assembly instructions, instruction-set-architecture (ISA) instructions, machine instructions, machine dependent instructions, microcode, firmware instructions, state-setting data, or either source code or object code written in any combination of one or more programming languages, including an object oriented programming language such as Smalltalk, C++ or the like, and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The computer readable program instructions can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider). In some embodiments, electronic circuitry including, for example, programmable logic circuitry, field-programmable gate arrays (FPGA), or programmable logic arrays (PLA) can execute the computer readable program instructions by utilizing state information of the computer readable program instructions to personalize the electronic circuitry, in order to perform aspects of the present disclosure.

[0109] The specific implementation described above is only a preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0110] Although the specific implementation of the present application is described above, it is not intended to limit the protection scope of the present application. Those skilled in the art should understand that various modifications or changes made on the basis of the technical solutions of the present application without creative labor are still within the protection scope of the present application.

Claims

1. A hyperspectral image enhancement method based on two-dimensional random resonance, characterized in that, Includes the following steps: Step 1: Obtain the hyperspectral image dataset; Step 2: Normalize the values ​​of all sample points in the hyperspectral image data; Step 3: Input the data of each band of the normalized hyperspectral image into a two-dimensional stochastic resonance model (2D DSR) for processing. Combine the information of adjacent pixels to process the signal of each pixel to obtain the enhanced data of each pixel; the formula of the two-dimensional stochastic resonance model is as follows: Where i and j represent the horizontal and vertical coordinates of the input signal, respectively, i = (1, 2, 3, ..., h), j = (1, 2, 3, ..., w), and h and w are the length and width of the input two-dimensional data, respectively; u i,j,k (n) represents the output of the signal in band k located at (i,j) after n iterations in the system; f i,j,k t represents the input of a noisy hyperspectral image signal located at (i,j) in the k-th band; x t y These represent the stride length in the horizontal and vertical directions, respectively; a and b are system parameters. Step 4: Normalize the enhanced hyperspectral image data to obtain the final hyperspectral image enhancement result dataset.

2. The hyperspectral image enhancement method based on two-dimensional stochastic resonance as described in claim 1, characterized in that, In step 2, normalizing the values ​​of all sample points in the hyperspectral image data specifically involves subtracting the minimum value from all sample points in the data, and then dividing by the difference between the maximum and minimum values ​​of all sample points, as shown in the following formula: Where H is the original hyperspectral image dataset and I is the normalized data.

3. The hyperspectral image enhancement method based on two-dimensional stochastic resonance as described in claim 1, characterized in that, The specific processing procedure of step 3 is as follows: the data of each band of the normalized hyperspectral image is sampled by sliding in a window of 2*2 pixels using a two-dimensional stochastic resonance model. The sliding method is to slide from the second row and second column to the last column, and then move to the next row and repeat the operation until the last row. The data is processed pixel by pixel through two-dimensional stochastic resonance. u(i,j) in each window is calculated using the four-way parallel formula in the two-dimensional stochastic resonance model. In each window, the calculation of the direction corresponding to the formula in the equation system is repeated. After scanning and enhancing all pixels, the output of each pixel is updated four times, integrating the information in the four directions of up, down, left and right, to obtain the hyperspectral image enhancement processing data.

4. The hyperspectral image enhancement method based on two-dimensional stochastic resonance as described in claim 3, characterized in that, In each window, the calculation of the direction corresponding to the formula in the repeated equation system is specifically as follows: the image signal f at (i,j-1) to the left of (i,j) is used. i,j-1 and the output u of the nth iteration at that point. i,j-1 (n) Calculate the first output u of point (i,j). i,j (n+1); the output u from point (i,j) i,j (n) and the image input signal f at that point i,j The system output at (i,j-1) is updated to obtain u. i,j-1 (n+1); the output u of point (i,j) obtained from the update. i,j (n) and the image input signal f at that point i,j The output u of the point (i-1,j) above (i,j) is obtained. i-1,j (n+1); the image signal f at position (i-1,j) above the updated (i,j) i-1,j and the output u of the nth iteration at that point i-1,j (n) Output u again for point (i,j) i,j (n+1) is used to update the output of the second updated point; When calculating by shifting the current window position one unit to the right, the output u at point (i,j+1) is used. i,j+1 (n) and the image input signal f at that point i,j+1 The system output at (i,j) is updated a third time to obtain u. i,j (n+1); When calculating by moving down one unit from the current window position, the output u of point (i+1,j) is used. i+1,j (n) and the image input signal f at that point i+1,j The fourth update output u of the point (i,j) above (i+1,j) is obtained. i,j (n+1); Finally, the system output of each pixel in rows 2 to h+1 and columns 2 to w+1 is obtained. The output of each pixel is updated four times, combining information from the four directions of up, down, left, and right.

5. The hyperspectral image enhancement method based on two-dimensional stochastic resonance as described in claim 1, characterized in that: For hyperspectral image datasets where shadow areas are subject to significant noise and have weak signal strength, before step 3, a two-dimensional data set of the shadow area in the normalized hyperspectral image data is extracted, and the extracted shadow area two-dimensional data set is filled with data. Finally, the enhanced shadow area is fused with the non-shadow area to obtain a hyperspectral image dataset with enhanced shadow area.

6. A hyperspectral image enhancement device based on two-dimensional random resonance, characterized in that: The device includes at least one processor and at least one memory, the processor and the memory being coupled together; the memory stores a computer-executable program; when the processor executes the computer-executable program stored in the memory, the processor can execute the hyperspectral image enhancement method as described in any one of claims 1 to 5.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program or instructions that, when executed by a processor, cause the processor to perform the hyperspectral image enhancement method as described in any one of claims 1 to 5.

Images