Multispectral remote sensing enteromorpha red tide detection method and system

Abstract

The invention discloses a multispectral remote sensing enteromorpha red tide detection method and system, and the method comprises the steps: carrying out the pseudo-exposure of an input multispectral image # imgabs0 # of a marine water color remote sensing satellite, generating a pseudo-exposure multispectral image sequence # imgabs1 #, decomposing the pseudo-exposure multispectral image into structural blocks, decomposing three components of signal intensity, average intensity and signal structure for each structure block, and fusing and reconstructing an enhanced multispectral remote sensing image # imgabs2; and performing fog intensity and fog abundance estimation on the enhanced multispectral remote sensing image # imgabs4 # by using a dark target subtraction method based on the reference image # imgabs3 #, and removing interference of thin clouds in the multispectral remote sensing image to obtain a multispectral image # imgabs5 # after clouds interference is removed for enteromorpha red tide detection. The method aims at long-time-sequence and large-scale enteromorpha red tide detection, the information extraction precision of the enteromorpha red tide is improved, and the efficiency of enteromorpha red tide detection is improved.

Description

Technical Field

[0001] The invention relates to the technical field of image processing, and in particular to a multi-spectral remote sensing enteromorpha red tide detection method and system. Background Art

[0002] Due to abnormal or drastic changes in the natural environment of the ocean, the Enteromorpha red tide disaster is particularly serious, endangering the local ecosystem or the safety of people's property. In the past decade, my country's Enteromorpha red tide has frequently erupted, accumulating in the Yellow Sea area during the summer every year and gradually drifting. Mapping the main distribution areas of the Enteromorpha red tide is an important demand for disaster prevention and mitigation. Therefore, long-term and large-scale monitoring of the Enteromorpha red tide can provide an effective scientific basis for the response and disposal of local governments. With the development of remote sensing images, the means of detecting Enteromorpha red tide have gradually become richer, which is more convenient and faster than the previous monitoring stations and monitoring ships. Compared with optical remote sensing satellites, multispectral remote sensing records information from several spectral bands and covers more spectral information of ground objects; at the same time, sensors such as the Korean Ocean Color Remote Sensing Satellite and the US MODIS satellite have a wide detection coverage area and a short revisit period, so multispectral satellites meet the long-term and large-scale monitoring needs of Enteromorpha red tide. However, there are two main problems with the information of Enteromorpha red tide in remote sensing images: first, due to the transmittance of ocean water, the reflection signal is weak, and the hydrological information such as Enteromorpha red tide is weak; second, remote sensing images are often interfered by clouds and fog, resulting in poor contrast and low visibility of the collected Enteromorpha red tide images, which seriously interferes with the identification and detection of Enteromorpha red tide. In terms of remote sensing image processing, although the existing technology records some image enhancement or declouding methods, including those based on image fusion, image transformation, deep learning and other technologies to improve the visualization effect and information extraction accuracy of remote sensing images, these methods face many limitations and are difficult to meet the long-term and large-scale Enteromorpha red tide detection applications. Summary of the invention

[0003] The technical problem to be solved by the present invention is as follows: In view of the above-mentioned problems in the prior art, a multi-spectral remote sensing enteromorpha red tide detection method and system are provided. The present invention aims to improve the information extraction accuracy of enteromorpha red tide and the efficiency of enteromorpha red tide detection for long time series and large scale.

[0004] In order to solve the above technical problems, the technical solution adopted by the present invention is: A multi-spectral remote sensing enteromorpha red tide detection method and system, comprising the following steps: S1, the multispectral image of ocean color remote sensing satellite is input Normalization is performed, and then pseudo-exposure image processing is performed to generate pseudo-exposure multispectral image sequences with different exposure levels. , and there is ,in ~ is a pseudo-exposure multispectral image sequence No. 1~ Pseudo-exposure multispectral images; S2, pseudo-exposure multispectral image sequence The pseudo-exposure multispectral image in the image is decomposed into structural blocks, and the signal intensity, average intensity and signal structure components of each structural block are decomposed and fused to reconstruct the enhanced multispectral remote sensing image. ; S3, based on reference images with the same viewpoint, scene, and satellite, without cloud or fog interference As a reference object, dark target subtraction is used to enhance the multispectral remote sensing image. The fog intensity and fog abundance are estimated and the interference of thin clouds and fog in multispectral remote sensing images is removed to obtain multispectral images after removing the interference of clouds and fog. ; S4, remove the multispectral image after cloud and fog interference The input image is classified using a pre-trained classifier to obtain the distribution map of the Enteromorpha red tide.

[0005] Optionally, the normalized function expression in step S1 is: , In the above formula, is the normalized multispectral image, For multispectral images Matrix data obtained after matrix transformation The Line List, Representing Matrix Data Middle The minimum value of the column, Representing Matrix Data Middle The maximum value of the column.

[0006] Optionally, in step S1, pseudo-exposure image processing is performed to generate pseudo-exposure multispectral image sequences with different exposure levels. When, before The generating function expression of a pseudo-exposure multispectral image is: , In the above formula, For the Pseudo-exposure multispectral images, is the logarithmic coefficient of variation, is the normalized multispectral image, The pseudo-exposure multispectral image is based on the normalized multispectral image It is obtained by using adaptive histogram equalization.

[0007] Optionally, step S2 includes: S2.1, pseudo-exposure multispectral image sequence The pseudo-exposure multispectral image in is decomposed into structural blocks, and the signal intensity is decomposed for each structural block , average intensity , signal structure Three components; S2.2: According to the following formula, the signal intensity, average intensity and signal structure of each pseudo-exposure multispectral image are fused to generate the fused signal intensity. , signal structure and average intensity ; , , , In the above formula, Indicates taking the maximum value, and is an intermediate variable, and there are: , , in, and is a constant, is the standard deviation of the building blocks, It means summation, is the inverse tangent function; Solve the intermediate variables according to the following formula : , in, is the signal strength; S2.3, intermediate variables of all pseudo-exposure multispectral images The stitching is restored to the original image size and represented as a feature , intermediate variable The stitching is restored to the original image size and represented as a feature , intermediate variable The stitching is restored to the original image size and represented as a feature , the enhanced multispectral remote sensing image is reconstructed according to the following formula ; , In the above formula, It means summation, The kernel size is The mean filter, Represents element-by-element multiplication operation, feature This is the result obtained by aggregating the original structural blocks and then performing mean filtering; S2.4, will enhance multispectral remote sensing images Multi-scale decomposition and fusion are performed to enhance the spatial detail information of multispectral images, thereby obtaining the final enhanced multispectral remote sensing image. .

[0008] Optionally, step S2.1 includes: based on a given structural block size parameter , the pseudo-exposure multispectral image sequence The pseudo-exposure multispectral image in is decomposed into The structural blocks of For multispectral images The number of bands; for each structural block, the signal strength, average strength, and signal structure are decomposed according to the following formula: , , , In the above formula, , and Respectively Pseudo-exposure multispectral image The signal strength, average strength and signal structure are three components. For the Pseudo-exposure multispectral image A collection of structural blocks, is equal to The intermediate variable express norm, It means to find the mean of the structure block.

[0009] Optionally, step S2.4 includes: S2.4.1, Initialization iteration number is 1, feature No. Layer Features Value as feature ,feature No. Layer Features The original feature , the feature No. Layer Features Value as feature , the feature No. Layer Features Value as feature ; S2.4.2, calculate the Image detail layer after layer decomposition and image base layer : , , In the above formula, It means summation, The kernel size is The mean filter, Represents element-by-element multiplication operation, feature This is the result obtained by aggregating the original structural blocks and then performing mean filtering; S2.4.3, the number of iterations Add 1 to determine the number of iterations Equal to the preset decomposition level threshold Is it true? If not, then the feature No. Layer Features ,feature No. Layer Features ,feature No. Layer Features ,feature No. Layer Features Downsample by a specified multiple as new features No. Layer Features ,feature No. Layer Features ,feature No. Layer Features ,feature No. Layer Features , jump to step S2.4.2; otherwise jump to step S2.4.4; where the decomposition level threshold The calculation function expression is: , In the above formula, Indicates rounding down. To obtain the minimum value, and Multispectral images The width and height of S2.4.4, the final enhanced multispectral remote sensing image is obtained according to the following formula: : + , , In the above formula, For the The base layer of the image after layer decomposition, For upsampling, For the Image detail layer after layer decomposition, For the The base layer of the image after layer decomposition, For the Image detail layer after layer decomposition, For multispectral images The original size of the base layer image, For multispectral images The original size of the image detail layer, .

[0010] Optionally, step S3 includes: S3.1, based on reference images with the same viewpoint, scene, satellite, and no cloud or fog interference As a reference object, dark target subtraction is used to enhance the multispectral remote sensing image. Estimation of fog intensity and fog abundance to obtain enhanced multispectral remote sensing images Fog intensity map and fog abundance map : , , In the above formula, Represents the reference image The bands that do not contain information on the red tide of Enteromorpha and Represents the reference image The number of bands that do not contain information on Enteromorpha red tides, Represents multispectral remote sensing images The bands that do not contain information on the red tide of Enteromorpha and Represents multispectral remote sensing images The number of bands that do not contain information on Enteromorpha red tide; To enhance multispectral remote sensing images Water body normalization index The pixel mean of the cloud area with a specified percentage before, The fog intensity map Water body normalization index The pixel mean of the cloud area with a specified percentage before, To enhance multispectral remote sensing images Water body normalization index The average value of the pixels in the cloud-free area before the specified percentage, The fog intensity map Water body normalization index The average value of pixels in the cloud-free area of ​​the specified percentage; S3.2, the multispectral image after removing the cloud and fog interference is calculated according to the following formula : ， In the above formula, To enhance multispectral remote sensing images, is the fog abundance map, This is the fog intensity map.

[0011] In addition, the present invention also provides a multi-spectral remote sensing Enteromorpha red tide detection system, comprising a microprocessor and a memory connected to each other, wherein the microprocessor is programmed or configured to execute the multi-spectral remote sensing Enteromorpha red tide detection method.

[0012] In addition, the present invention also provides a computer-readable storage medium, in which a computer program or instruction is stored. The computer program or instruction is programmed or configured to execute the multi-spectral remote sensing Enteromorpha red tide detection method through a processor.

[0013] In addition, the present invention also provides a computer program product, including a computer program or instructions, which are programmed or configured to execute the multi-spectral remote sensing Enteromorpha red tide detection method through a processor.

[0014] Compared with the prior art, the present invention has the following advantages: the present invention comprises the input multi-spectral image of ocean water color remote sensing satellite Perform pseudo exposure to generate pseudo exposure multispectral image sequence The pseudo-exposure multispectral image is decomposed into structural blocks, and each structural block is decomposed into three components: signal intensity, average intensity, and signal structure, and fused to reconstruct the enhanced multispectral remote sensing image. ; Based on the reference image Using dark object subtraction to enhance multispectral remote sensing images The fog intensity and fog abundance are estimated and the interference of thin clouds and fog in multispectral remote sensing images is removed to obtain multispectral images after removing the interference of clouds and fog. For detection of Enteromorpha red tide. The present invention is aimed at long-term, large-scale Enteromorpha red tide detection, and solves the problem of complex ocean background environment, which is often interfered by clouds and fog, as well as weak image reflection signals and small spectral differences between phenomena. It can quickly and efficiently restore the Enteromorpha red tide information in the image, enhance the category differences between Enteromorpha red tide, clouds and fog, land, and seawater, so that it can be applied to large-scale, long-term Enteromorpha red tide detection through traditional classification methods, effectively enhance the contrast between Enteromorpha red tide and other landforms, and can detect it from large-scale remote sensing images, with the advantages of high information extraction accuracy of Enteromorpha red tide and high detection efficiency of Enteromorpha red tide. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 Schematic diagram of the basic flow of the method of the embodiment of the present invention.

[0016] Figure 2 is a multispectral image in an embodiment of the present invention Schematic diagram (false color diagram composed of bands 7, 6 and 3).

[0017] Figure 3 is the pseudo-exposure multispectral image sequence in the embodiment of the present invention Schematic diagram, where (a) to (f) are six pseudo-exposure multispectral images (false color images composed of the 7th, 6th, and 3rd bands).

[0018] Figure 4 The enhanced multispectral remote sensing image obtained in the embodiment of the present invention is Schematic diagram (false color map composed of bands 7, 6, and 3).

[0019] Figure 5 The multispectral image obtained in the embodiment of the present invention after removing the interference of clouds and fog Schematic diagram (false color map composed of bands 7, 6, and 3).

[0020] Figure 6 The distribution map of Enteromorpha red tide obtained in the embodiment of the present invention, wherein (a) is the result of classification using support vector machine before image enhancement, and (b) is the result of classification using support vector machine after image enhancement. DETAILED DESCRIPTION

[0021] like Figure 1 As shown, the multi-spectral remote sensing enteromorpha red tide detection method of this embodiment includes the following steps: S1, the multispectral image of ocean color remote sensing satellite is input Normalization is performed, and then pseudo-exposure image processing is performed to generate pseudo-exposure multispectral image sequences with different exposure levels. , and there is ,in ~ is a pseudo-exposure multispectral image sequence No. 1~ Pseudo-exposure multispectral images; S2, the pseudo-exposure multispectral image sequence The pseudo-exposure multispectral image in the image is decomposed into structural blocks, and the signal intensity, average intensity and signal structure components of each structural block are decomposed and fused to reconstruct the enhanced multispectral remote sensing image. ; S3, based on reference images with the same viewpoint, scene, and satellite, without cloud or fog interference As a reference object, dark target subtraction is used to enhance the multispectral remote sensing image. The fog intensity and fog abundance are estimated and the interference of thin clouds and fog in multispectral remote sensing images is removed to obtain multispectral images after removing the interference of clouds and fog. ; S4, remove the multispectral image after cloud and fog interference The input image is classified using a pre-trained classifier to obtain the distribution map of the Enteromorpha red tide.

[0022] The multispectral image of the ocean color remote sensing satellite input in step S1 of this embodiment like Figure 2 As shown, the multispectral image For low-resolution, large-scale multispectral images of ocean color remote sensing satellites, the images are normalized and then pseudo-exposure image processing is performed to generate multispectral image sequences with different exposure levels. In step S1, a normalization operation is performed to scale the data to the same range. The normalization operation is performed on the original multispectral image. , perform image matrix transformation operation to make its matrix length , width is the number of bands , and scale the data to In the range of , the normalized function expression in step S1 of this embodiment is: , In the above formula, is the normalized multispectral image, For multispectral images Matrix data obtained after matrix transformation The Line List, Representing Matrix Data Middle The minimum value of the column, Representing Matrix Data Middle After the data is normalized, it is restored to its original width, height, and number of bands.

[0023] In order to obtain an image with a high dynamic range, in step S1 of this embodiment, pseudo-exposure image processing is performed to generate a pseudo-exposure multispectral image sequence with different exposure levels. In the step S1 of this embodiment, a pseudo-exposure image processing is performed to generate a pseudo-exposure multispectral image sequence with different exposure degrees. When, before The generating function expression of a pseudo-exposure multispectral image is: , In the above formula, For the Pseudo-exposure multispectral images, is the logarithmic coefficient of variation, is the normalized multispectral image, The pseudo-exposure multispectral image is based on the normalized multispectral image Adaptive histogram equalization is used to obtain the ,therefore Using the logarithmic transformation above, Adaptive histogram equalization is used, as shown in Figure 3 shown.

[0024] In this embodiment, step S2 includes: S2.1, pseudo-exposure multispectral image sequence The pseudo-exposure multispectral image in is decomposed into structural blocks, and the signal intensity is decomposed for each structural block , average intensity , signal structure Three components; S2.2: According to the following formula, the signal intensity, average intensity and signal structure of each pseudo-exposure multispectral image are fused to generate the fused signal intensity. , signal structure and average intensity ; , , , In the above formula, Indicates taking the maximum value, and is an intermediate variable, and there are: , , in, and is a constant, is the standard deviation of the building blocks, It means summation, is the inverse tangent function; Solve the intermediate variables according to the following formula : , in, is the signal strength; S2.3, intermediate variables of all pseudo-exposure multispectral images The stitching is restored to the original image size and represented as a feature , intermediate variable The stitching is restored to the original image size and represented as a feature , intermediate variable The stitching is restored to the original image size and represented as a feature , the enhanced multispectral remote sensing image is reconstructed according to the following formula ; , In the above formula, It means summation, The kernel size is The mean filter, Represents element-by-element multiplication operation, feature This is the result obtained by aggregating the original structural blocks and then performing mean filtering; S2.4, will enhance multispectral remote sensing images Multi-scale decomposition and fusion are performed to enhance the spatial detail information of multispectral images, thereby obtaining the final enhanced multispectral remote sensing image. .

[0025] In this embodiment, step S2.1 includes: based on a given structural block size parameter , the pseudo-exposure multispectral image sequence The pseudo-exposure multispectral image in is decomposed into The structural blocks of For multispectral images The number of bands; for each structural block, the signal strength, average strength, and signal structure are decomposed according to the following formula: , , , In the above formula, , and Respectively Pseudo-exposure multispectral image The signal strength, average strength and signal structure are three components. For the Pseudo-exposure multispectral image A collection of structural blocks (including multiple exposure image sequences), is equal to The intermediate variable express norm, Indicates the mean of the structure block. Each image in The image is divided into structural blocks of size , the size of each image block is , is the number of bands of the multispectral image, then the structural block It can be expressed as: , In the above formula, the meaning of each symbol is explained in detail in the previous text.

[0026] In step S2.2 of this embodiment, the signal intensity, average intensity, and signal structure of each pseudo-exposure multispectral image are fused to generate a fused signal intensity , signal structure and average intensity When the signal intensity of different image sequences is fused Signal strength by structure block The maximum value in determines the signal structure after fusion of different image sequences. Signal structure by structure block The average intensity after fusion of different image sequences is obtained by power weighting and normalization. The average strength of the structural block The signal strength after fusion is , the higher the contrast, the better the gain effect, and the mathematical expression is: .

[0027] In step S2.4 of this embodiment, the multispectral remote sensing image is enhanced. Enhance multispectral remote sensing images by performing multi-scale decomposition and fusion to enhance the spatial detail information of multispectral images. It can be expressed as: , , , Enhanced Multispectral Remote Sensing Image For the Image detail layer after layer decomposition and image base layer The superposition composition can be based on which multi-scale decomposition and fusion can be performed to enhance the spatial detail information of the multi-spectrum. Specifically, step S2.4 in this embodiment includes: S2.4.1, Initialization iteration number is 1, feature No. Layer Features Value as feature ,feature No. Layer Features The original feature , the feature No. Layer Features Value as feature , the feature No. Layer Features Value as feature ; S2.4.2, calculate the Image detail layer after layer decomposition and image base layer : , , In the above formula, It means summation, The kernel size is The mean filter, Represents element-by-element multiplication operation, feature This is the result obtained by aggregating the original structural blocks and then performing mean filtering; S2.4.3, the number of iterations Add 1 to determine the number of iterations Equal to the preset decomposition level threshold Is it true? If not, then the feature No. Layer Features ,feature No. Layer Features ,feature No. Layer Features ,feature No. Layer Features Downsample by a specified multiple as new features No. Layer Features ,feature No. Layer Features ,feature No. Layer Features ,feature No. Layer Features , jump to step S2.4.2; otherwise jump to step S2.4.4; where the decomposition level threshold The calculation function expression is: , In the above formula, Indicates rounding down. To obtain the minimum value, and Multispectral images The width and height of S2.4.4, the final enhanced multispectral remote sensing image is obtained according to the following formula: : + , , In the above formula, For the The base layer of the image after layer decomposition, is the upsampling (the inverse of the downsampling multiple), For the Image detail layer after layer decomposition, For the The base layer of the image after layer decomposition, For the Image detail layer after layer decomposition, For multispectral images The original size of the base layer image, For multispectral images The original size of the image detail layer, In this embodiment, the final enhanced multispectral remote sensing image is obtained like Figure 4 shown.

[0028] In this embodiment, step S3 includes: S3.1, based on reference images with the same viewpoint, scene, satellite, and no cloud or fog interference As a reference object, dark target subtraction is used to enhance the multispectral remote sensing image. Estimation of fog intensity and fog abundance to obtain enhanced multispectral remote sensing images Fog intensity map and fog abundance map : , , In the above formula, Represents the reference image The bands that do not contain information on the red tide of Enteromorpha and Represents the reference image The number of bands that do not contain information on Enteromorpha red tides, Represents multispectral remote sensing images The bands that do not contain information on the red tide of Enteromorpha and Represents multispectral remote sensing images The number of bands that do not contain information on Enteromorpha red tide; To enhance multispectral remote sensing images Water body normalization index The pixel mean of the cloud area with a specified percentage before, The fog intensity map Water body normalization index The pixel mean of the cloud area with a specified percentage before, To enhance multispectral remote sensing images Water body normalization index The average value of the pixels in the cloud-free area before the specified percentage, The fog intensity map Water body normalization index The average value of pixels in the cloud-free area of ​​the specified percentage; S3.2, the multispectral image after removing the cloud and fog interference is calculated according to the following formula : ， In the above formula, To enhance multispectral remote sensing images, is the fog abundance map, The multi-spectral image obtained in this embodiment after removing the interference of fog and cloud is like Figure 5 shown.

[0029] In step S3.1, the enhanced multispectral remote sensing image is input And reference images with the same viewing angle, scene, satellite, and without cloud or fog interference , to obtain enhanced multispectral remote sensing images The average image of the band that does not contain the information of the red tide of Enteromorpha is used to obtain the reference image The average image of the corresponding band. In particular, the reflectivity of the Enteromorpha red tide usually gradually increases in the red light band (700-750 nm), which is quite different from the seawater. Therefore, the spectrum less than 700 nm can be regarded as a band that does not contain the Enteromorpha red tide information, and the spectrum greater than 700 nm can be regarded as a band that contains the Enteromorpha red tide information. In the multispectral image used in the example, there are a total of 8 bands, namely 412 nm, 443 nm, 490 nm, 555 nm, 660 nm, 680 nm, 745 nm and 865 nm. Therefore, the first 6 bands can be regarded as bands that do not contain the Enteromorpha red tide information, and the last 2 bands can be regarded as bands that contain the Enteromorpha red tide information. Then, use the reference image The average image of the corresponding band minus the enhanced multispectral remote sensing image The enhanced multispectral remote sensing image can be obtained by averaging the image without the information band of Enteromorpha red tide. Fog intensity map , its function expression is as follows: , In the above formula, Represents the reference image The bands that do not contain information on the red tide of Enteromorpha and Represents the reference image The number of bands that do not contain information on Enteromorpha red tides, Represents multispectral remote sensing images The bands that do not contain information on the red tide of Enteromorpha and Represents multispectral remote sensing images The number of bands that do not contain information on Enteromorpha red tide.

[0030] According to the dark channel method, the enhanced multispectral remote sensing image is obtained The minimum value of each pixel in the multispectral image in all bands is calculated by the following formula: the larger the pixel value, the higher the cloud concentration. The pixels with a specified percentage (in this embodiment, the value is 1%, and other specified percentages can also be used as needed) are selected to represent the pixels in the cloud area; the water body normalization index of the multispectral image is calculated. The pixels with a specified percentage (in this embodiment, the value is 1%, and other specified percentages can also be used as needed) represent the cloud-free area. The water body normalization index The calculation function expression is as follows: , In the above formula, Represents the green band, The fog abundance map The calculation function expression is as follows: , In the above formula, , In the above formula, To enhance multispectral remote sensing images Water body normalization index The pixel mean of the cloud area with a specified percentage before, The fog intensity map Water body normalization index The pixel mean of the cloud area with a specified percentage before, To enhance multispectral remote sensing images Water body normalization index The average value of the pixels in the cloud-free area before the specified percentage, The fog intensity map Water body normalization index The average value of pixels in the cloud-free area before the specified percentage.

[0031] In step S4, the multispectral image after removing the cloud and fog interference is The input image is classified by the pre-trained classifier to obtain the distribution map of the Enteromorpha red tide. The required classifier and its training method can be used as needed. For example, the multispectral image after removing the cloud and fog interference As the input image, the red tide of Enteromorpha was annotated and divided into training set and test set at a ratio of 1:100. The support vector machine was used to classify it. The distribution map of the red tide of Enteromorpha was obtained as follows: Figure 6 As shown, (a) is the result of classification using the support vector machine before image enhancement, that is, the distribution map of the Enteromorpha red tide obtained by directly using only step S4 of the method of this embodiment, and (b) is the result of classification using the support vector machine after image enhancement, that is, the distribution map of the Enteromorpha red tide obtained by using steps S1 to S4 of the method of this embodiment. Figure 6 It can be seen that before the image enhancement of steps S1 to S3 of the method of this embodiment, the classified distribution map of the Enteromorpha red tide (a) misidentifies some noise or thin cloud boundary areas as Enteromorpha red tide, while after the image enhancement of steps S1 to S3 of the method of this embodiment, the classified distribution map of the Enteromorpha red tide (b) effectively increases the separability between classes (background, seawater and Enteromorpha) and reduces the misclassification scene. In addition, a classification method or classifier such as random forest can also be used to achieve the classified distribution map of the Enteromorpha red tide.

[0032] In order to verify the multi-spectral remote sensing enteromorpha red tide detection system of this embodiment, three indicators are selected in this embodiment to quantitatively evaluate the target segmentation method proposed in this example. TP represents true positive, that is, the number of samples that are actually positive and correctly predicted as positive; FN represents false negative, that is, the number of samples that are actually positive but are wrongly predicted as negative; FP represents false positive, that is, the number of samples that are actually negative but are wrongly predicted as positive; TN represents true negative, that is, the number of samples that are actually negative and are correctly predicted as negative. The following three indicators can be calculated based on the confusion matrix: Accuracy AA (Accuracy): refers to the proportion of samples whose model prediction results match the actual labels. The calculation formula is as follows: AA=1 / 2×(TP / (TP+FN)+TN / (TN+FP)) , Prediction accuracy OA (Overal Accuracy): refers to the proportion of samples in the model prediction results that match the actual labels. The calculation formula is as follows: OA=((TP+TN)) / (TP+TN+FP+FN), Kappa coefficient (Cohen's Kappa): takes into account the difference between expected accuracy and actual accuracy, and is used to measure the consistency of model classification. The calculation formula is as follows: Kappa=((OA-AA)) / (1-AA).

[0033] The results obtained by using two classification methods, support vector machine or random forest, are shown in Table 1.

[0034] Table 1: Quantitative evaluation results of three indicators of two classification methods

[0035] As shown in Table 1, the multispectral remote sensing enteromorpha red tide detection system of this embodiment has achieved good results in the three indicators of prediction accuracy OA, accuracy AA and Kappa coefficient based on two classification methods, namely support vector machine or random forest.

[0036] In summary, the multispectral remote sensing Entera red tide detection system of this embodiment processes low-resolution, large-scale multispectral remote sensing images including: image enhancement, image declouding and Entera red tide segmentation. Generate a pseudo-exposure sequence for the current multispectral remote sensing image, and perform fusion reconstruction from the three components of signal structure, signal intensity and average intensity respectively; refer the reconstructed enhanced image to the cloud-free remote sensing image of the same scene, perform dark target subtraction to decloud, remove the influence of thin clouds in the image, and enhance the spectral information of Entera red tide; classify the declouded image using classification methods such as random forest and support vector machine to effectively detect the distribution area of ​​Entera red tide. The multispectral remote sensing Entera red tide detection system of this embodiment combines image enhancement and declouding technology to enhance the separability of Entera red tide from seawater, land, clouds and other areas, and can effectively monitor the distribution of Entera red tide in the ocean over a long period of time, providing support for research and application in related fields.

[0037] In addition, this embodiment also provides a multi-spectral remote sensing Enteromorpha red tide detection system, including a microprocessor and a memory connected to each other, and the microprocessor is programmed or configured to execute the multi-spectral remote sensing Enteromorpha red tide detection method.

[0038] In addition, this embodiment also provides a computer-readable storage medium, in which a computer program or instruction is stored. The computer program or instruction is programmed or configured to execute the multi-spectral remote sensing Enteromorpha red tide detection method through a processor.

[0039] In addition, this embodiment also provides a computer program product, including a computer program or instructions, which are programmed or configured to execute the multi-spectral remote sensing Enteromorpha red tide detection method through a processor.

[0040] Those skilled in the art should understand that the technical solutions provided by the embodiments of the present application may be in the form of methods, systems, or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may take the form of a computer program product implemented on one or more computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes. The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the process Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing device to work in a specific way, so that the instructions stored in the computer-readable memory produce a product including an instruction device, which implements the functions specified in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0041] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A multi-spectral remote sensing enteromorpha red tide detection method, characterized in that: The steps include: S1, the multispectral image of ocean color remote sensing satellite is input Normalization is performed, and then pseudo-exposure image processing is performed to generate pseudo-exposure multispectral image sequences with different exposure levels. , and there is ,in ~ is a pseudo-exposure multispectral image sequence No. 1~ Pseudo-exposure multispectral images; S2, pseudo-exposure multispectral image sequence The pseudo-exposure multispectral image in the image is decomposed into structural blocks, and the signal intensity, average intensity and signal structure components of each structural block are decomposed and fused to reconstruct the enhanced multispectral remote sensing image. ; S3, based on reference images with the same viewpoint, scene, and satellite, without cloud or fog interference As a reference object, dark target subtraction is used to enhance the multispectral remote sensing image. The fog intensity and fog abundance are estimated and the interference of thin clouds and fog in multispectral remote sensing images is removed to obtain multispectral images after removing the interference of clouds and fog. ; S4, remove the multispectral image after cloud and fog interference The input image is classified using a pre-trained classifier to obtain the distribution map of the Enteromorpha red tide.

2. The multi-spectral remote sensing enteromorpha red tide detection method according to claim 1, characterized in that: The normalized function expression in step S1 is: ， In the above formula, is the normalized multispectral image, For multispectral images Matrix data obtained after matrix transformation The Line List, Representing Matrix Data Middle The minimum value of the column, Representing Matrix Data Middle The maximum value of the column.

3. The multi-spectral remote sensing enteromorpha red tide detection method according to claim 1, characterized in that: In step S1, pseudo-exposure image processing is performed to generate pseudo-exposure multispectral image sequences with different exposure levels. When, before The generating function expression of a pseudo-exposure multispectral image is: ， In the above formula, For the Pseudo-exposure multispectral images, is the logarithmic coefficient of variation, is the normalized multispectral image, The pseudo-exposure multispectral image is based on the normalized multispectral image It is obtained by using adaptive histogram equalization.

4. The multi-spectral remote sensing enteromorpha red tide detection method according to claim 1, characterized in that: Step S2 includes: S2.1, pseudo-exposure multispectral image sequence The pseudo-exposure multispectral image in is decomposed into structural blocks, and the signal intensity is decomposed for each structural block , average intensity , signal structure Three components; S2.2: According to the following formula, the signal intensity, average intensity and signal structure of each pseudo-exposure multispectral image are fused to generate the fused signal intensity. , signal structure and average intensity ; ， ， ， In the above formula, Indicates taking the maximum value, and is an intermediate variable, and there are: ， ， in, and is a constant, is the standard deviation of the building blocks, It means summation, is the inverse tangent function; Solve the intermediate variables according to the following formula : ， in, is the signal strength; S2.3, intermediate variables of all pseudo-exposure multispectral images The stitching is restored to the original image size and represented as a feature , intermediate variable The stitching is restored to the original image size and represented as a feature , intermediate variable The stitching is restored to the original image size and represented as a feature , the enhanced multispectral remote sensing image is reconstructed according to the following formula ; ， In the above formula, It means summation, The kernel size is The mean filter, Represents element-by-element multiplication operation, feature This is the result obtained by aggregating the original structural blocks and then performing mean filtering; S2.4, will enhance multispectral remote sensing images Multi-scale decomposition and fusion are performed to enhance the spatial detail information of multispectral images, thereby obtaining the final enhanced multispectral remote sensing image. .

5. The multi-spectral remote sensing enteromorpha red tide detection method according to claim 4, characterized in that: Step S2.1 includes: based on a given structural block size parameter , the pseudo-exposure multispectral image sequence The pseudo-exposure multispectral image in is decomposed into The structural blocks of For multispectral images The number of bands; for each structural block, the signal strength, average strength, and signal structure are decomposed according to the following formula: ， ， ， In the above formula, , and Respectively Pseudo-exposure multispectral image The signal strength, average strength and signal structure are three components. For the Pseudo-exposure multispectral image A collection of structural blocks, is equal to The intermediate variable express norm, It means to find the mean of the structure block.

6. The multi-spectral remote sensing enteromorpha red tide detection method according to claim 4, characterized in that: Step S2.4 includes: S2.4.1, Initialization iteration number is 1, feature No. Layer Features Value is a feature ,feature No. Layer Features The original feature , the feature No. Layer Features Value is a feature , the feature No. Layer Features Value is a feature ; S2.4.2, calculate the Image detail layer after layer decomposition and image base layer : ， ， In the above formula, It means summation, The kernel size is The mean filter, Represents element-by-element multiplication operation, feature This is the result obtained by aggregating the original structural blocks and then performing mean filtering; S2.4.3, the number of iterations Add 1 to determine the number of iterations Equal to the preset decomposition level threshold Is it true? If not, then the feature No. Layer Features ,feature No. Layer Features ,feature No. Layer Features ,feature No. Layer Features Downsample by a specified multiple as new features No. Layer Features ,feature No. Layer Features ,feature No. Layer Features ,feature No. Layer Features , jump to step S2.4.2; otherwise jump to step S2.4.4; where the decomposition level threshold The calculation function expression is: ， In the above formula, Indicates rounding down. To obtain the minimum value, and Multispectral images The width and height of S2.4.4, the final enhanced multispectral remote sensing image is obtained according to the following formula: : + ， ， In the above formula, For the The base layer of the image after layer decomposition, For upsampling, For the Image detail layer after layer decomposition, For the The base layer of the image after layer decomposition, For the Image detail layer after layer decomposition, For multispectral images The original size of the base layer image, For multispectral images The original size of the image detail layer, .

7. The multi-spectral remote sensing enteromorpha red tide detection method according to claim 1, characterized in that: Step S3 includes: S3.1, based on reference images with the same viewpoint, scene, satellite, and no cloud or fog interference As a reference object, dark target subtraction is used to enhance the multispectral remote sensing image. Estimation of fog intensity and fog abundance to obtain enhanced multispectral remote sensing images Fog intensity map and fog abundance map : ， ， In the above formula, Represents the reference image The bands that do not contain information on the red tide of Enteromorpha and Represents the reference image The number of bands that do not contain information on Enteromorpha red tides, Represents multispectral remote sensing images The bands that do not contain information on the red tide of Enteromorpha and Represents multispectral remote sensing images The number of bands that do not contain information on Enteromorpha red tide; To enhance multispectral remote sensing images Water body normalization index The pixel mean of the cloud area with a specified percentage before, The fog intensity map Water body normalization index The pixel mean of the cloud area with a specified percentage before, To enhance multispectral remote sensing images Water body normalization index The average value of the pixels in the cloud-free area before the specified percentage, The fog intensity map Water body normalization index The average value of pixels in the cloud-free area of ​​the specified percentage; S3.2, the multispectral image after removing the cloud and fog interference is calculated according to the following formula : ， In the above formula, To enhance multispectral remote sensing images, is the fog abundance map, This is the fog intensity map.

8. A multi-spectral remote sensing enteromorpha red tide detection system, comprising a microprocessor and a memory connected to each other, characterized in that: The microprocessor is programmed or configured to execute the multi-spectral remote sensing Enteromorpha red tide detection method according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program or instruction stored therein, characterized in that: The computer program or instruction is programmed or configured to execute the multi-spectral remote sensing enteromorpha red tide detection method as described in any one of claims 1 to 7 through a processor.

10. A computer program product comprising a computer program or instructions, characterized in that The computer program or instruction is programmed or configured to execute the multi-spectral remote sensing enteromorpha red tide detection method as described in any one of claims 1 to 7 through a processor.

Images