A hyperspectral image shadow enhancement method, device and storage medium

Abstract

The application provides a hyperspectral image shadow enhancement method, equipment and a storage medium, wherein the method comprises the following steps: acquiring a hyperspectral image dataset; normalizing the hyperspectral dataset and extracting a shadow part by using a mask; continuously sampling and unfolding the shadow part image in a row scanning manner in a spatial dimension into a one-dimensional signal, sequentially passing through a bistable nonlinear system for dynamic stochastic resonance; re-composing a square matrix from the sequence output by the system in a row scanning manner and then normalizing; continuously sampling and unfolding the obtained shadow part data in a column scanning manner into a one-dimensional sequence, and inputting the one-dimensional sequence into the bistable nonlinear system again for dynamic stochastic resonance; re-composing a square matrix from the output sequence in a column scanning manner and then normalizing, and fusing the normalized square matrix with a non-shadow area of the original hyperspectral image. The application utilizes the scanning and sampling modes in two different directions of rows and columns, fully utilizes the correlation of the pixels in the spatial dimension, retains more image information, and achieves better image enhancement effect.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of hyperspectral image processing, and particularly relates to a hyperspectral image shadow enhancement method, equipment and storage medium. BACKGROUND

[0002] Hyperspectral image (HSI) contains rich scene information, which can not only extract information of each position in two-dimensional space, but also extract multiple features of ground objects from different spectral bands. With the increasing maturity of hyperspectral imaging technology, HSI plays an important role in fields such as space exploration, weather forecasting, food safety, agriculture, environmental monitoring, medical diagnosis and industrial quality inspection. However, the strong noise and weak signal nature of the shadow area of HSI makes the data processing extremely difficult. The shadow area of HSI formed by light obstruction, cloud layer, ground undulation or ground obstruction has low reflected light energy and weak spectral signal, and the noise is stronger than that in the non-shadow area, so that the pixels and targets are difficult to distinguish and detect. At present, image enhancement and pixel classification are mostly carried out by removing noise, but this will lose a large amount of useful information in the shadow area. The enhancement of the information in the shadow area is a necessary preparation work for subsequent hyperspectral image classification, which can greatly improve the efficiency of human-computer interaction and the performance of computer vision, such as detection, classification and tracking.

[0003] Stochastic resonance (SR) theory is currently widely used in the amplification of weak signals in a strong noise environment, that is, when the three of bistable nonlinear system, noise and weak signal are coordinated, the noise energy can be converted into signal energy, which lays a foundation for the extraction of shadow information in HSI.

[0004] LIU et al. proposed a gray image recovery enhancement method based on adaptive bistable array stochastic resonance in a low peak signal-to-noise ratio environment. Chouhan et al. proposed a spatial domain analysis method based on dynamic stochastic resonance (DSR) to enhance low-contrast images. These methods have good enhancement effect on uniform low-contrast images with fixed parameters, but the scope of application is small, and there are certain limitations for enhancing HSI. SUMMARY

[0005] In view of the above problems, the present application proposes a D-DSR algorithm based on a bistable nonlinear system, and the implementation process consists of two consecutive DSRs. Compared with DSR, D-DSR retains the correlation between adjacent pixels in HSI to a greater extent, so that the image after stochastic resonance retains more original information, and the shadow area enhancement effect is more significant.

[0006] The first aspect of the present application provides a hyperspectral image shadow enhancement method, comprising the following steps:

[0007] Step 1, obtaining a hyperspectral image data set;

[0008] Step 2, normalizing the gray scale of the image pixels in the hyperspectral data set to the range of [0, 1], and extracting the shadow part using a mask;

[0009] Step 3, continuously sampling and unfolding the extracted shadow part image in the spatial dimension in a row scanning sampling manner into a one-dimensional signal, making it sequentially pass through a bistable nonlinear system for dynamic stochastic resonance, and outputting a group of sequences after stochastic resonance;

[0010] Step 4, re-combining the system output sequence into a square matrix in a row scanning manner and then normalizing it;

[0011] Step 5, continuously sampling and unfolding the shadow part data obtained in step 4 in a column scanning sampling manner into a one-dimensional sequence, and inputting it again into the bistable nonlinear system for the second dynamic stochastic resonance, to obtain a group of sequences;

[0012] Step 6, re-combining the sequence output by the second dynamic stochastic resonance into a square matrix in a column scanning manner and then normalizing it, and finally fusing it with the non-shadow area of the original hyperspectral image to obtain a hyperspectral image enhanced by double dynamic stochastic resonance.

[0013] In a possible design, after the step 1 of obtaining the hyperspectral image data set, the hyperspectral image data can be pre-processed; the pre-processing is to calculate the F-norm value of each band in the hyperspectral data set, the size of which represents the amount of information contained in each band image, and then artificially remove the bands with less information, and then re-combine the remaining bands in order to form a hyperspectral data set.

[0014] In a possible design, the specific process of the step 2 is as follows: first, the gray scale of the image pixels ranges from 0 to 255, and the gray value of each pixel is divided by 255 to make the gray scale of all the image pixels in the range of [0, 1]; then, a shadow extraction mask made of a ground truth image is used to extract the shadow area in the hyperspectral image, and only the shadow area is processed, and after the mask is extracted, three-dimensional hyperspectral image shadow data is obtained from the original hyperspectral image data.

[0015] In a possible design, the row scanning sampling manner in the step 3 is as follows: starting from the first pixel of each image in the spatial dimension, sampling in the order from left to right, when reaching the rightmost pixel of the first row, continuing to sample from the rightmost pixel of the second row, and then sampling in the order from right to left until reaching the first pixel of the second row; then, repeating the above steps until the last pixel is sampled.

[0016] In a possible design, the sampling manner of the column scanning in step 5 is specifically as follows: starting from the first pixel of each image in the spatial dimension, sampling in the order from top to bottom, when reaching the pixel in the last row of the first column, continuing sampling from the last row of the second column, and sampling in the order from bottom to top until reaching the first row of the second column; the above steps are repeated to sample the last pixel.

[0017] The second aspect of the present application further provides a hyperspectral image shadow enhancement device, 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 shadow enhancement method according to the first aspect.

[0018] The third aspect of the present application further provides a computer readable storage medium, the computer readable storage medium storing a computer program or instructions, the program or instructions being executed by the processor to make the processor execute the hyperspectral image shadow enhancement method according to the first aspect.

[0019] Compared with the prior art, the present application provides a hyperspectral image shadow enhancement method, the implementation process of which consists of two consecutive DSRs, and D-DSR retains the correlation between adjacent pixels in HSI to a greater extent than DSR, so that the image after random resonance retains more original information, and the shadow area enhancement effect is more significant.

[0020] As a three-dimensional tensor, each pixel in the hyperspectral image has a certain correlation with its neighborhood pixels, and DSR samples the image from one direction, without fully utilizing the correlation between the spatial pixels; the present application proposes a D-DSR algorithm based on a bistable nonlinear system, the implementation process of which consists of two consecutive DSRs, and the correlation between adjacent pixels in the hyperspectral image is retained to a greater extent by using the scanning sampling manner in the row and column directions, more original image information is retained, and better shadow enhancement effect can be achieved, so that the subsequent classification of the hyperspectral image is better, and the classification accuracy is higher. BRIEF DESCRIPTION OF DRAWINGS

[0021] Figure 1 The flowchart of the hyperspectral image shadow enhancement method of the present application.

[0022] Figure 2 The F-norm values of each waveband in the HYDICE image data set.

[0023] Figure 3 The scene image containing the shadow area selected from the HYDICE image data set and the shadow area mask image thereof.

[0024] Figure 4 Flow chart of the algorithm D-DSR of the present application.

[0025] Figure 5 Image of the original hyperspectral image of HYDICE in band 1 and the images after DSR enhancement and D-DSR enhancement in the spatial dimension, respectively.

[0026] Figure 6 HybridSN classification result comparison chart of the data when the training set is 20%.

[0027] Figure 7 3D-CNN classification result comparison chart of the data when the training set is 20%.

[0028] Figure 8 Structure simple block diagram of the hyperspectral image shadow enhancement device. DETAILED DESCRIPTION

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

[0030] Embodiment 1:

[0031] The present application proposes a D-DSR (double dynamic stochastic resonance) algorithm based on a bistable nonlinear system, and the implementation process consists of two consecutive DSRs (dynamic stochastic resonance). Compared with DSR, D-DSR retains the correlation between adjacent pixels in HSI to a greater extent, so that the image after stochastic resonance retains more original information, and the shadow area enhancement effect is more significant.

[0032] In this embodiment, a real hyperspectral data set HYDICE is selected to do comparative experiments of different denoising algorithms, and the results show that the D-DSR algorithm can extract more shadow information than other algorithms, and can better improve the classification accuracy of the shadow area of HSI.

[0033] (I) Image data set acquisition

[0034] In this embodiment, the region containing shadow information in the data set HYDICE is used to verify the denoising ability of D-DSR. The data set used in this embodiment is collected by the hyperspectral digital image collection experiment (HYDICE) in August 1995. The HYDICE has 148 bands, a total of 316 rows and 216 columns, and can be expressed as X∈R 316×216×148The spatial resolution is 0.75 m and the spectral resolution is 10 nm. HYDICE has 7 land cover classes: grass, tree, road, shadow (including a section of road) and 3 different targets. In order to verify the classification effect of the shadow area, this paper divides the shadow area into two categories: shadow and road under shadow, and the 8 different categories of ground objects are represented by 8 different colors after classification, as shown in Table 1.

[0035] Table 1 HYDICE sample categories and quantities

[0036]

[0037]

[0038] (ii) Preprocessing of hyperspectral image data

[0039] Affected by factors such as noise and water absorption in the atmosphere, some bands in the HYDICE data set contain less information. In order to improve the efficiency of image processing, these bands need to be removed. In this embodiment, the Frobenius norm (F-norm) is used to represent the size of the energy of each band image in the hyperspectral image, so as to compare the energy contained in each band image. The F-norm is represented by the square root of the sum of squares of all elements of a matrix. The F-norm of each band image in the HSI spectral dimension can be solved by the following formula:

[0040]

[0041] In the formula, i1 represents the number of rows of the image, i2 represents the number of columns of the image, and I3 represents the number of spectral bands.

[0042] The F-norm values of the 148 bands of HYDICE are calculated as shown in Figure 2 From the figure, it can be seen that the information contained in the first 24 bands is less, and the F-norm value is less than 20, so the first 24 bands of the data set are artificially removed, and the remaining 124 bands are recombined into a 316x216x124 hyperspectral image in order; At the same time, the hyperspectral data is preprocessed by dimensionality reduction.

[0043] At the same time, in order to verify the effect of D-DSR on enhancing HSI, a scene containing a shadow area is selected from HYDICE for comparison, as shown in Figure 3 (a), and the shadow area mask is shown in Figure 3 (b).

[0044] (III) Enhancing the shadow in the hyperspectral image by D-DSR

[0045] The specific process is as shown in Figure 4

[0046] Step one: normalize the gray scale of the image pixels in the hyperspectral data set to the range of [0, 1], and extract the shadow part using a mask;

[0047] Normalization: the gray value range of a general two-dimensional gray image is between 0 and 255, but according to the adiabatic approximation theory or linear response theory of stochastic resonance, only a small parameter can be substituted, that is, the amplitude, frequency and noise intensity of the signal are much smaller than 1. Therefore, the HSI gray value needs to be linearly normalized to the range of [0, 1] before stochastic resonance.

[0048] The gray scale range of the image pixels is [0, 255], and the gray value of each pixel is divided by 255, so that the gray scale range of all the pixels of the image is in the range of [0, 1].

[0049] Mask: use the shadow extraction mask made of the ground truth image to extract the shadow area in the HSI, and only further process the shadow area. After extracting the mask, three-dimensional HSI shadow data can be obtained from the original HSI data.

[0050] Step two: continuously sample and expand the extracted shadow part image in the spatial dimension according to the row scanning sampling mode into a one-dimensional signal, so that it sequentially passes through the dynamic stochastic resonance of the bistable nonlinear system, and outputs a group of sequences after stochastic resonance;

[0051] Row scanning sampling: starting from the first pixel of each image in the spatial dimension, sample in the order from left to right, when reaching the rightmost pixel of the first row, continue sampling from the rightmost pixel of the second row, and then sample in the order from right to left until reaching the first column of the second row. Repeat the above steps until the last pixel is sampled.

[0052] Step three: reorganize the sequence output by the system into a square matrix according to the row scanning mode and then normalize it;

[0053] At this time, the pixel histogram of the gray scale image converted from the square matrix will be concentrated in a narrow range, which increases the brightness of the image but weakens the contrast, resulting in insufficient clarity of details. Normalize the square matrix so that the gray scale range is within [0, 1], so that the distribution of the overall pixel gray value of the image becomes balanced.

[0054] Step four: continuously sample and expand the shadow part data obtained in step three into a one-dimensional sequence according to the column scanning mode, and input it again into the bistable nonlinear system for the second dynamic stochastic resonance to obtain a group of sequences.

[0055] ​Column scanning sampling: starting from the first pixel of each image in the spatial dimension, sampling in the order from top to bottom, when reaching the pixel of the last row of the first column, continue sampling from the last row of the second column, and then in the order from bottom to top until reaching the first row of the second column. Repeat the above steps until the last pixel is sampled.

[0056] As a three-dimensional tensor, each pixel in the hyperspectral image has a certain correlation with its neighborhood pixels. DSR samples the image from one direction, without fully utilizing the correlation between spatial pixels. The present application proposes a D-DSR algorithm based on a bistable nonlinear system, which is composed of two consecutive DSRs. The scanning sampling mode in the row and column directions retains the correlation between adjacent pixels in the hyperspectral image to a greater extent, retains more original image information, and achieves better shadow enhancement effect, so that the subsequent classification of the hyperspectral image is better and the classification accuracy is higher.

[0057] Step five: reconstitute the sequence obtained by the second dynamic stochastic resonance output into a square matrix in the column scanning mode and then normalize it (at this time, the pixel histogram of the gray image converted from the square matrix will be concentrated in a narrow range, the image brightness is increased but the contrast is weakened, and after normalization, the distribution is uniform and the details can be clearly seen), and finally fuse it with the non-shadow area of the original hyperspectral image to obtain the D-DSR enhanced hyperspectral image.

[0058] (Four) Comparison of experimental results

[0059] 1. Enhancement results:

[0060] The shadow parts of the original HSI and the images enhanced by the DSR algorithm and the D-DSR algorithm respectively are extracted using a mask, and the mean and standard deviation of the shadow area and the overall accuracy (overall accuracy, OA) and Kappa (using 3DCNN classification, training sample is 20% of the total sample) are compared, as shown in Table 2. The mean refers to the average value of the image pixels, denoted by u, which reflects the average brightness of the image. The larger the average brightness, the better the image quality. The standard deviation refers to the dispersion degree of the image pixel gray value relative to the mean value, denoted by STD. If the standard deviation is larger, it means that the gray levels in the image are more dispersed, and the image quality is better. The mean and standard deviation calculation formulas are formula (2) and formula (3) respectively:

[0061]

[0062]

[0063] where F is the image to be evaluated, with size M x N, M is the length of the image, N is the width of the image, (i, j) is the position of the pixel in space.

[0064] The original hyperspectral image of HYDICE in band 1 is shown in Fig. (a) of Figure 5 , and the DSR enhanced and D-DSR enhanced images in spatial dimension are shown in Fig. (b) and (c) of Figure 5 , respectively. Figure 5

[0065] Table 2 Mean and standard deviation of original hyperspectral image and image after stochastic resonance

[0066] Raw data DSR (row scan) D-DSR (column scan) Mean 839.1554 1288.7000 1448.8000 Standard deviation 2978.0000 4632.1000 5219.4000 OA 96.5433 97.0300 97.3664 Kappa 94.0223 94.8845 95.4496

[0067] 2. Classification results:

[0068] For the D-DSR enhanced data set, the parameters of the 3D-CNN model are optimized according to the OA. The optimization process is as follows: first, input the D-DSR enhanced HYDICE data, and set the parameters to default or empirical values. According to the single variable principle, adjust the epoch (iteration number), batch size, learning rate, optimizer and loss function respectively, keep other parameters unchanged when adjusting one parameter, and select the optimal parameter according to the OA in the classification results.

[0069] In order to verify the image enhancement effect of D-DSR, the DSR denoising algorithm is used as a comparison method, and 3D-CNN and HybridSN are used for classification on the HSI enhanced by the two algorithms respectively, and the overall accuracy (OA) and Kappa coefficient are compared. OA refers to the ratio of the number of correctly classified class pixels to the total number of classes, and Kappa coefficient represents the proportion of error reduction of classification compared with completely random classification. The calculation formulas of OA and Kappa coefficient are as follows:

[0070]

[0071]

[0072] where N, C are the total number of samples and the total number of classes in HSI, x ii is the number of correctly classified test samples, x i+ is the number of real samples of each class, and x +i is the number of predicted samples of each class.

[0073] Table 3 OA of classification of original hyperspectral image and enhanced image​

[0074]

[0075]

[0076] Figure 6 The figure is a comparison of the classification results of the HybridSN of the data when the training set is 20%. In the figure (a), the classification result of the original data is shown; in the figures (b) and (c), the classification results of the data enhanced by DSR and D-DSR are shown, respectively.

[0077] Figure 7 The figure is a comparison of the classification results of the 3D-CNN of the data when the training set is 20%. In the figure (a), the classification result of the original data is shown; in the figures (b) and (c), the classification results of the data enhanced by DSR and D-DSR are shown, respectively.

[0078] Therefore, after the hyperspectral image is enhanced by the method of the application, the classification effect is better and the classification accuracy is higher.

[0079] Embodiment 2

[0080] As shown in Figure 8 The application also provides a hyperspectral image shadow enhancement device, 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 shadow enhancement method as described in Embodiment 1. The internal bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component (PCI) bus or an.Xtended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the convenience of representation, the bus in the drawings of the application does not mean only one bus or one type of bus. The memory can comprise a high-speed RAM memory, and can also comprise 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, etc.

[0081] The device can be provided as a terminal, a server or other forms of devices.

[0082] Figure 8is a block diagram of one device illustratively shown. 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 typically controls overall operations of the electronic device, such as operations associated with display, phone calls, data communications, camera operations, and recording operations. The processing component can include one or more processors to execute instructions to complete the steps of the methods described above, in whole or in part. Moreover, the processing component can include one or more modules to facilitate 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.

[0083] The memory is configured to store various types of data to support operations of the electronic device. Examples of such 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 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 memory, flash memory, magnetic disk or optical disk.

[0084] The power supply component supplies power for the various components of the electronic device. The power supply component can include a power management system, one or more power supplies, 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 related to the touch or swiping action. In some embodiments, the multimedia component includes a front-facing camera and / or a rear-facing camera. The front-facing camera and / or the rear-facing camera can receive external multimedia data when the electronic device is in an operation mode, such as a shooting mode or a video mode. Each of the front-facing camera and the rear-facing camera can be a fixed optical lens system or have a focal length and optical zoom capability.

[0085] The audio component is configured to output and / or input audio signals. For example, the audio component includes a microphone (MIC) that receives an external audio signal when the electronic device is in an operation 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 for outputting audio signals.

[0086] The sensor component includes one or more sensors configured to provide various state assessments of the electronic device. For example, the sensor component can detect an open / closed position of the electronic device, relative positioning of components, such as a display and a keypad of the electronic device, a change of position of the electronic device or a component of the electronic device, presence or absence of user contact with the electronic device, the orientation or acceleration / deceleration of the electronic device, 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, configured to capture images 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.

[0087] 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 broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel. In an example embodiment, the communication component further 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) techniques, infrared data association (IrDA) techniques, ultra-wideband (UWB) techniques, Bluetooth (BT) techniques, and other techniques.

[0088] 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.

[0089] Embodiment 3:

[0090] The application further provides a computer readable storage medium, wherein a computer program or instructions are stored in the computer readable storage medium, and the program or instructions can make a processor execute the hyperspectral image shadow enhancement method in embodiment 1 when executed by the processor.

[0091] Specifically, a system, device or apparatus equipped with a readable storage medium on which software program codes for realizing the functions of any one of the above embodiments are stored can be provided, and a computer or processor of the system, device or apparatus is caused to read and execute the instructions stored in the readable storage medium. In this case, the program codes read from the readable medium can realize the functions of any one of the above embodiments by itself, and thus the machine readable codes and the readable storage medium storing the machine readable codes constitute a part of the application.

[0092] The storage medium can be realized 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-purpose or special-purpose computer.

[0093] It should be understood that the above processor can be a central processing unit (English: Central Processing Unit, for short: CPU), and can also be other general-purpose processors, digital signal processors (English: Digital Signal Processor, for short: DSP), application specific integrated circuits (English: Application Specific Integrated Circuit, for short: ASIC) and the like. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor or the like. 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.

[0094] It should be understood that the storage media can be coupled to the processor such that the processor can read information from, and write information to, the storage media. Of course, the storage media can be a part of the processor. The processor and the storage media can be located in an Application Specific Integrated Circuit (ASIC) or other chip. Of course, the processor and the storage media can be located in discrete components as well.

[0095] The computer readable program instructions described herein can be downloaded to respective computing / processing devices from a computer readable storage medium or to an external computer or external storage device via a network, for example, the Internet, a local area network, a wide area network and / or a wireless network. The network can comprise copper transmission cables, optical transmission fibers, wireless transmission, routers, firewalls, switches, gateway computers and / or edge servers. A 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 for storage in a computer readable storage medium within the respective computing / processing device.

[0096] Computer readable program instructions for carrying out operations of the present disclosure can be assembler 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 computing / processing device, partly on the user's computing / processing device, as a stand-alone software package, partly on the user's computing / processing device and partly on a remote computing / processing device or entirely on the remote computing / processing device or server. In the latter scenario, the remote computing / processing device can be connected to the user's computing / processing device 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.

[0097] The above merely describes preferred embodiments of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, and the like made by those of ordinary skill in the art based on the technical solutions of the present application shall fall within the protection scope of the present application.

[0098] Although the specific embodiments of the present application are described above, the present application is not limited to the above, and 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 method for enhancing shadows in hyperspectral images, characterized in that, Includes the following steps: Step 1: Obtain the hyperspectral image dataset; Step 2: Normalize the gray levels of the image pixels in the hyperspectral dataset to the range of [0, 1], and extract the shadow parts using a mask; First, the gray level range of the image pixels is [0, 255]. The gray value of each pixel is divided by 255 so that the gray level range of all pixels in the image is within the range of [0, 1]. Then, the shadow extraction mask made from the ground truth image is used to extract the shadow area in the hyperspectral image. Only the shadow area is processed. After the mask is extracted, the three-dimensional hyperspectral image shadow data is obtained from the original hyperspectral image data. Step 3: In the spatial dimension, the extracted shadow area image is continuously sampled and expanded into a one-dimensional signal using a row-scan sampling method. This signal is then passed sequentially through a bistable nonlinear system for dynamic random resonance, outputting a set of randomly resonated sequences. The row-scan sampling method is as follows: starting from the first pixel of each image in the spatial dimension, sampling is performed from left to right. When the rightmost pixel of the first row is reached, sampling continues from the rightmost column of the second row, and then from right to left until the first column of the second row is reached. The above steps are repeated until the last pixel is sampled. Step 4: Reassemble the system output sequence into a square matrix by scanning rows and then normalize it. Step 5: The shaded data obtained in Step 4 is continuously sampled and expanded into a one-dimensional sequence by column scanning, and then input into the bistable nonlinear system for the second dynamic random resonance to obtain a set of sequences. Step 6: The sequence obtained from the second dynamic random resonance output is reassembled into a square matrix according to the column scanning method and then normalized. Finally, it is fused with the non-shaded area of ​​the original hyperspectral image to obtain the hyperspectral image enhanced by double dynamic random resonance.

2. The hyperspectral image shadow enhancement method as described in claim 1, characterized in that: After obtaining the hyperspectral image dataset in step 1, the hyperspectral image data is preprocessed. The preprocessing involves calculating the F-norm value of each band in the hyperspectral dataset, the magnitude of which indicates the amount of information contained in each band image. Then, bands with less information are manually removed, and the remaining bands are recombined in order to form a hyperspectral dataset.

3. The hyperspectral image shadow enhancement method as described in claim 1, characterized in that, The column scanning sampling method in step 5 is as follows: starting from the first pixel of each image in the spatial dimension, sampling is performed in a top-down order. When the last pixel of the first column is reached, sampling continues from the last row of the second column, and then in a bottom-up order until the first row of the second column is reached. The above steps are repeated until the last pixel is sampled.

4. A hyperspectral image shadow enhancement device, 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 performs the hyperspectral image shadow enhancement method as described in any one of claims 1 to 3.

5. 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 shadow enhancement method as described in any one of claims 1 to 3.

Images