Noctilucent fishing boat automatic identification method combining peak detection and local detection

Abstract

The invention discloses a noctilucent fishing boat automatic identification method combining peak detection and local detection, and belongs to the technical field of fishing boat business monitoring. Comprising the following steps: acquiring and preprocessing noctilucent remote sensing image data; carrying out peak detection on the preprocessed noctilucent remote sensing image data by adopting a peak detection method to obtain a preliminary fishing boat point to be determined; performing local peak detection on the image data of the preliminary fishing boat point to be determined obtained after peak detection to obtain a fishing boat point to be determined; and performing false alarm factor elimination on the to-be-determined fishing boat point data to obtain a fishing boat identification result. According to the invention, by using VIIRS / DNB night light remote sensing data, through a detection algorithm combining local peak detection and peak detection, misjudgment pixels caused by light divergence are removed. And for the to-be-detected fishing boat point location after the local peak detection extraction, the false alarm factor is eliminated, the misjudgment is reduced, and the method has the characteristics of high resolution and high precision.

Description

Technical Field

[0001] The invention relates to the technical field of fishing vessel business monitoring, and in particular to an automatic identification method for luminous fishing vessels combining peak detection with local detection. Background Art

[0002] Light fishing boats refer to fishing boats that use artificial light sources to attract and trap marine fish by taking advantage of the fish's phototaxis. Such fishing boats usually use purse seines, trawl nets and cover nets to catch fish swarms that are attracted by lights. Ommastrephes, Scomber and Cololabis sair are the main targets of light fishing boats. Fishing boat monitoring is an important source of information on fishing boat operations. It can help fishery management departments formulate reasonable fishery policies and prevent overfishing. It can also help protect marine ecosystems and reduce the impact on the environment. The unique fishing operation mode of light fishing boats can rely on night light remote sensing technology to achieve real-time monitoring of light fishing boats. Compared with traditional fishing boat monitoring methods, it has the advantages of low cost and wide coverage, and has a good development prospect in light-induced fisheries.

[0003] Night light remote sensing data are obtained by satellite or aerial platforms to observe the distribution of lights on the earth at night. These data are mainly obtained by night spectral sensors, which can reflect different types of light sources such as urban lights, industrial activities, and agricultural irrigation. Typical night light remote sensing data sources are the "Daytime / Nighttime Cloud Top" image data (VIIRS / DNB) and "Nighttime Light Satellite" data (DMSP / OLS) provided by the National Oceanic and Atmospheric Administration (NOAA) of the United States. Since the 1970s, the Linear Scan System (Operational Linescan System, OLS) visible light band sensor carried on the Defense Meteorological Satellite Program (DMSP) platform has the ability to collect low-light images at night. Since then, low-light imaging data collected by sensors on meteorological satellites at night have been used in the study of detecting fishing boat lights. However, DMSP / OLS has a series of well-known shortcomings: coarse spatial resolution and lack of flight calibration. Compared with DMSP / OLS data, the DNB (Day / Night Band) data of the Visible Infrared Imaging Radiometer Suite (VIIRS) carried by the Suomi National Polar-orbiting Partnership (S-NPP) satellite has higher spatial and temporal resolution, lower detection limit, wider dynamic range, and solves the problem of lack of flight calibration, so it has received more widespread attention.

[0004] There are many studies on algorithms for automatically identifying luminous fishing boats based on night light remote sensing data. All algorithms are designed based on the characteristics of the huge difference between the luminous pixels and the background in the night light images. The current mainstream algorithms are the peak detection algorithm and the gradient method. For the peak detection algorithm, it mostly uses empirical thresholds in threshold division, which is relatively simple and cannot adaptively calculate thresholds according to different images. This is easy to cause misjudgment due to the divergence of light from highlight pixels to adjacent pixels; another maximum entropy threshold segmentation method can well solve the problem that the empirical threshold cannot be threshold adaptive, but it cannot play a good role in correcting the misjudgment of adjacent pixels. At the same time, night light data may also be affected by factors such as clouds and meteorological conditions, affecting the recognition accuracy; background light sources such as city lights and shipping lights may interfere with the recognition of fishing boats and need to be effectively distinguished. Summary of the invention

[0005] In view of the defects of the prior art, the present invention proposes a detection algorithm that combines local peak detection with spike detection, which uses VIIRS / DNB night light remote sensing data to realize automatic identification of luminous fishing vessels, and has the characteristics of high resolution and high precision.

[0006] In order to achieve the above object, the present invention provides a method for automatically identifying luminous fishing boats by combining peak detection with local detection, comprising the following steps:

[0007] (1) Obtain VIIRS / DNB night light remote sensing image data and perform preprocessing;

[0008] (2) using a peak detection method to perform peak detection on the pre-processed night light remote sensing image data to obtain preliminary fishing boat points to be determined;

[0009] (3) performing local peak detection on the image data of the preliminary fishing boat point to be determined after the peak detection to obtain the fishing boat point to be determined;

[0010] (4) Eliminate false alarm factors for the fishing boat point data to be determined, and obtain a fishing boat identification result.

[0011] Furthermore, the preprocessing includes: geometric correction of the longitude and latitude of the night light remote sensing image data and cropping of the sea area to be studied;

[0012] The geometric correction of the longitude and latitude of the night light remote sensing image data is used to eliminate the geometric distortion in the night light remote sensing image so that the geographic coordinates in the image are consistent with the position in the real world;

[0013] The clipping for the sea area to be studied is to clip the night light remote sensing image data using the land mask and the longitude and latitude range of the sea area to be studied to obtain the image data of the sea area to be studied.

[0014] Furthermore, the step (2) is specifically as follows:

[0015] (2.1) performing radiation amplification on the pre-processed night light remote sensing image data and performing radiation enhancement using logarithmic transformation log10 (rad);

[0016] (2.2) De-noising the image data after the radiation enhancement processing; De-noising the image data by Wiener filtering and then by median filtering;

[0017] (2.3) subtracting the image data after noise reduction processing from the original image data to obtain a difference, and screening out points above the first threshold;

[0018] (2.4) The image data that has been radiometrically amplified but not logarithmically transformed in step (2.1) is traversed in the horizontal and vertical directions respectively, and the ratio between the target pixel and the mean of the two adjacent pixels is calculated to obtain the peak height index:

[0019]

[0020] Among them: radiance of the subjecr pixel is the target pixel, average radiance of the two adjacent pixels is the average radiance of the two adjacent pixels of the target pixel;

[0021] Screening the points whose peak height index is greater than a second threshold;

[0022] (2.5) Horizontal strip lightning detection is performed on the image data after radiometric enhancement, and the pixels marked as lightning will be discarded;

[0023] (2.6) For the intersection of the results gradually screened out in steps (2.3) to (2.4) and the points not marked as lightning, the points above the third threshold are screened out and retained as the preliminary fishing boat points to be determined, and the remaining non-preliminary fishing boat points to be determined are set to 0.

[0024] Furthermore, the first threshold is 0.75; the second threshold is 0.035; and the third threshold is 0.6.

[0025] Furthermore, the Wiener filter and the median filter both use a 3×3 window.

[0026] Furthermore, the horizontal band lightning detection is as follows: for the image data that has been processed with radiation enhancement, the radiation difference between 16 rows of pixels is calculated and separated, and the area where the difference is greater than 0.1Log10DNB and lasts for 24 or more continuous pixels in the horizontal direction is marked as lightning.

[0027] Furthermore, the step (3) is specifically as follows:

[0028] (3.1) For the image data after peak detection, start from the first column and compare the radiation values ​​between two pixels vertically. The radiation value of the pixel with the larger radiation value is retained, and the smaller value is changed to 0.

[0029] (3.2) After traversing the entire column, repeat step (3.1) starting from the data in the second column until all columns are traversed;

[0030] (3.3) Then, traverse the results of the column traversal horizontally starting from the first row, compare the pixel sizes, and retain the larger pixel values ​​until all rows are traversed.

[0031] (3.4) The points that remain after column traversal and row traversal are used as the fishing boat points to be determined.

[0032] Further, the false alarm factors include position-invariant point light sources, position-variable point light sources, and gas flares;

[0033] The position-invariant point light sources include nearshore light sources, island and reef light sources, and offshore drilling platforms. A buffer zone is set as a mask to remove luminous pixels within the mask.

[0034] The position-variable point light source includes a merchant ship, and the navigation track data of non-fishing vessels is obtained through AIS, and the luminous pixels within the mask are removed using the track mask;

[0035] The gas flare was acquired using the M10 band of VIIRS and was used to remove the effects of the gas flare in the DNB image.

[0036] Furthermore, for the nearshore light source, a 2 km buffer zone is created outside the land as a continental mask, and the luminous pixels within the mask are eliminated;

[0037] For the island and reef light sources, a 2-kilometer buffer zone is created outside the sea area as an island and reef mask, and the luminous pixels within the mask are removed;

[0038] For each of the offshore drilling platforms, a 1-km buffer zone is created outside each drilling platform as a drilling platform mask, and the luminous pixels within the mask are removed.

[0039] Beneficial effects of the present invention:

[0040] The present invention utilizes VIIRS / DNB nighttime light remote sensing data, and removes misjudged pixels caused by light divergence through a detection algorithm combining local peak detection with spike detection. At the same time, the false alarm factor of the fishing boat point to be tested after local peak detection extraction is eliminated, and the misjudgment of the light fishing boat point caused by shipping, offshore drilling platforms, gas flares, lightning and other factors is removed, which has the characteristics of high resolution and high precision. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a schematic flow chart of a method for automatically identifying luminous fishing vessels by combining peak detection with local detection according to an embodiment of the present invention.

[0042] Figure 2 This is a VIIRS / DNB image after radiation enhancement processing according to an embodiment of the present invention.

[0043] Figure 3 This is a VIIRS / DNB image after noise reduction processing according to an embodiment of the present invention.

[0044] Figure 4 This is a VIIRS / DNB image after peak detection according to an embodiment of the present invention.

[0045] Figure 5 This is a graph showing the result of local peak detection according to an embodiment of the present invention.

[0046] Figure 6 1 is a comparison diagram of the results before and after local peak detection according to an embodiment of the present invention.

[0047] Figure 7 This is a graph showing high-quality coordinate matching results of fishing boat points and EOG data products obtained after local peak detection in an embodiment of the present invention. DETAILED DESCRIPTION

[0048] In order to better illustrate and understand the technical solutions and advantages of the present invention, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings and embodiments, but the content of the present invention is not limited to the following embodiments.

[0049] like Figure 1 As shown, an embodiment of the present invention provides a method for automatically identifying luminous fishing boats by combining peak detection with local detection, comprising the following steps:

[0050] S101, obtaining night light remote sensing image data and performing preprocessing.

[0051] The remote sensing image data used in this patent are all from the National Oceanic and Atmospheric Administration (NOAA; https: / / www.class.noaa.gov). VIIRS / DNB (Visible Infrared Imaging Radiometer Suite / Day-Night Band) data are obtained from the Suomi National Polar-orbiting Partnership (SuomiNPP) satellite and subsequent satellites of the National Oceanic and Atmospheric Administration (NOAA).

[0052] VIIRS is an advanced imager, a multi-band remote sensing instrument that captures visible, infrared, and mid-infrared radiation from the Earth's surface. It has high spatial resolution and a wide range of observation bands, and can provide data on climate, weather, ocean, land, and other aspects. It can monitor a variety of meteorological and environmental parameters such as clouds, aerosols, temperature, and humidity. The DNB band is a special band on VIIRS, which is mainly used to observe the Earth's surface at night. It has high sensitivity and can detect very weak light, so DNB can capture light sources such as city lights, fires, and luminescence of marine plankton at night. Compared with DMSP, VIIRS has a stronger ability to detect the characteristics of fishing vessels from low-light images and has sufficient resolution to identify the source of ship lights.

[0053] Preprocessing of the original image data of VIIRS / DNB refers to geometric correction of its longitude and latitude and cropping of the sea area to be studied in order to obtain remote sensing images that can be analyzed and processed.

[0054] The geometric correction of the longitude and latitude of night light remote sensing image data is used to eliminate the geometric distortion in night light remote sensing images so that the geographic coordinates in the image are consistent with the real-world location.

[0055] The cropping of the sea area to be studied is to crop the night light remote sensing image data using the land mask and the longitude and latitude range of the study area to obtain the image data of the sea area to be studied.

[0056] S102, using a peak detection method to perform peak detection on the pre-processed night light remote sensing image data to obtain preliminary fishing boat points to be determined.

[0057] The radiation value of the highlight pixel in the night light remote sensing data is much higher than that of the background pixel, showing the characteristics of a spike in the image, and the noise reduction process can smooth such a spike very well. Subtracting the original image with the peak and the pixel after the smooth noise reduction process can highlight the existence of the peak and segment the highlight pixel from the background, so as to achieve the preliminary detection of night light ships. The specific steps are:

[0058] (1) Radiation enhancement processing: Based on the acquired pre-processed VIIRS / DNB remote sensing data, radiation amplification is first performed, that is, the image radiation value is ×10^9, and then logarithmic transformation log10 (rad) is performed to achieve radiation enhancement processing of the image data. Figure 2 As shown in the figure, the horizontal and vertical axes represent the range of rows and columns of the area in the entire image, and the grayscale value represents the radiation range after logarithmic transformation.

[0059] (2) Noise reduction: The image data after radiation enhancement is subjected to noise reduction. It usually needs to go through two steps of noise reduction. First, Wiener filtering is used for noise reduction and then median filtering is used for noise reduction. Both use a 3×3 window to achieve noise reduction of remote sensing images. Figure 3 As shown, Figure 2 In contrast, the difference between highlight pixels and background pixels in the image after noise reduction and smoothing is reduced, preparing for the next step of peak detection.

[0060] (3) Peak detection (SMI): Peak detection (SMI) is performed on the radiation image after noise reduction processing. The sharp radiation value of the image after noise reduction processing is smoothed, and the difference obtained by subtracting it from the original image can well express the highlighted luminous pixels in the smooth background. The pixels higher than the first threshold are screened out as peaks. In the embodiment of the present invention, the first threshold is 0.035.

[0061] (4) Calculation of spike height index (SHI): In order to detect the sharpness of the spike in the adjacent pixels, it is necessary to continue to calculate the spike height index of the image that has been radiometrically amplified but not logarithmically transformed. The calculation process is traversed in the horizontal and vertical directions respectively, and the ratio between the target pixel and the mean of the two adjacent pixels is calculated (the adjacent pixels in the horizontal direction refer to the adjacent pixels on the left and right of the target pixel, and in the vertical direction refer to the adjacent pixels on the top and bottom);

[0062]

[0063] Among them: the radiance of the subject pixel is the target pixel, and the average radiance of the two adjacent pixels is the average radiance of the two adjacent pixels of the target pixel.

[0064] Pixels whose spike height index (SHI) is greater than a second threshold are screened; in the embodiment of the present invention, the second threshold is 0.75.

[0065] (5) Lightning detection: Considering the presence of lightning in the image, in order to reduce the interference of lightning on the data, lightning is detected as horizontal strip lightning in VIIRS / DNB. The characteristics of these lightning are exactly 16 rows of bright pixels. The input image after logarithmic enhancement is used to calculate the radiation difference between the separated 16 rows of pixels. The area with a difference greater than 0.1 Log10 DNB and 24 or more continuous pixels in the horizontal direction is marked as lightning, and the pixels marked as lightning are discarded.

[0066] (6) For the intersection of the results gradually screened out in steps (3) to (4) and the points not marked as lightning, the points above the third threshold are screened out as preliminary fishing boat points to be determined, and the remaining non-preliminary fishing boat points to be determined are set to 0. In the embodiment of the present invention, the third threshold is 0.6.

[0067] The image processed by the peak detection algorithm is shown in the figure Figure 4 As shown, Figure 2 , Figure 3 By comparison, it can be seen that peak detection can effectively extract bright pixels from the dark background and remove background noise to a certain extent.

[0068] S103, performing local peak detection on the image data of the preliminary fishing boat point to be determined obtained after the peak detection, to obtain the fishing boat point to be determined.

[0069] For ocean-going fishing vessels with high-power fish-attracting lights, the pixels adjacent to the pixels where the fish-attracting lights are located may also be illuminated by the high-power fish-attracting lights and have high radiation values, leading to misjudgment. During the propagation of light, its energy tends to gradually decay with increasing distance. Therefore, local detection can be performed on the pixels after threshold segmentation to remove misjudgment caused by scattered radiation of light. The specific process is as follows:

[0070] (1) For the night light remote sensing image data after peak detection, the radiation values ​​between two pixels are compared vertically starting from the first column, and the radiation value of the pixel with the larger radiation value is retained, and the smaller value is changed to 0;

[0071] (2) After traversing the entire column, repeat step (1) starting from the data in the second column until all columns have been traversed;

[0072] (3) Then, traverse the results of the column traversal horizontally starting from the first row, compare the pixel sizes, and retain the larger pixel values ​​until all rows are traversed.

[0073] (4) The points that remain after column traversal and row traversal are taken as the fishing boat points to be determined.

[0074] S104, removing false alarm factors from the to-be-determined fishing boat point data to obtain a fishing boat identification result.

[0075] Since there are different nighttime luminous targets at sea, such as nearshore light sources or offshore drilling platforms, the extracted sea area nighttime light point data set needs to be further screened and eliminated. Therefore, it is necessary to calculate the false alarm factor mask for the extracted sea area except for the fishing boat lights. False alarm factors are divided into fixed light sources, such as nearshore light sources or offshore drilling platforms; and non-fixed light sources with variable positions, such as fast-moving merchant ships at sea; and gas flares. Different types of false alarms are extracted separately.

[0076] (1) Fixed-position light source

[0077] Fixed-position light sources mainly include near-shore light sources, island and reef light sources, and offshore drilling platforms.

[0078] Nearshore light sources are generally caused by the halo of city lights or infrastructure such as coastal ports. In response to the impact of nearshore light sources, the study adopted a buffer zone method and created a 2-kilometer buffer zone outward from the land as a continental mask to eliminate the luminous pixels within the mask.

[0079] The island and reef light sources mainly refer to the coral islands and reefs widely distributed in the Northwest Pacific Ocean. These islands and reefs are either partially exposed to the sea and are not fishable areas, or partially occupied by people and have built lighthouses, airports and other light source facilities, which will cause false alarms of fishing boats' lights. Therefore, a 2-kilometer buffer zone is created outside the sea area for the island and reef flat area as an island and reef mask.

[0080] Offshore drilling platforms are fixed burning fire points at sea, and the Northwest Pacific Ocean is rich in oil and gas resources. The brightness of the burning fire points of offshore drilling platforms is the brightest target in night light remote sensing images except for noise. Considering the halo effect of burning fire points, a 1 km buffer operation was performed for each drilling platform in the dataset.

[0081] (2) Non-fixed light source with variable position

[0082] In addition to intensive fishing, the Northwest Pacific Ocean also bears important transportation responsibilities for my country's maritime trade with other Asian countries, America, Europe, and Africa. Therefore, commercial ships at sea are also one of the important false alarms of fishing boat lights. The navigation track data of non-fishing fishing vessels can be obtained through AIS, and the luminous pixels within the mask can be eliminated using the track mask.

[0083] (3) Gas flare

[0084] Gas flares can be obtained using the M10 band of VIIRS, which is used to remove the effects of gas flares in DNB images.

[0085] The following is a set of recognition results of luminous fishing boats based on VIIRS / DNB remote sensing image data. Table 1 shows the comparison of radiation values ​​of the extraction results of the embodiment method of the present invention. Table 1 (a) is the result after peak detection, and Table 1 (b) is the result after local peak detection again. The first column in the table represents the column where the pixels of the area are located in the image, and the first row represents the row where the pixels of the area are located in the image. By comparison, it can be seen that: threshold detection only uses the threshold to filter the pixel values ​​of luminous remote sensing data, but local peak detection eliminates the misjudgment caused by high radiation values ​​due to high-power fish-attracting lights as much as possible through traversal in the horizontal and vertical directions, and the result is more accurate.

[0086] Table 1

[0087]

[0088]

[0089] (a)

[0090]

[0091]

[0092] (b)

[0093] Figure 5 to Figure 7 The result map is given as an example of the cloud-free image of the area of ​​123°E~127°E, 35°N~38°N in the Yellow Sea between China and South Korea on October 9, 2020. The gray polygon in the figure is the continental mask. Figure 5 This is the result diagram of local peak detection, and the cross mark is the detection result. Figure 6 The comparison chart of the results before and after local peak detection is shown in Figure 2. The larger dots are the marks after local peak detection, and the smaller dots are the marks before local peak detection. It can be found that the error detection of adjacent pixels irradiated by the lights of fishing boats has a more obvious correction effect. Figure 7 This is the result diagram of the coordinate matching between the fishing boat points obtained by local peak detection and the high-quality fishing boat points of the EOG data product. The larger dots are the logos of the high-quality fishing boat points of the EOG data product, and the small dots are the logos after local peak detection. It can be seen from the figure that the fishing boat position obtained by local peak detection is highly matched with the fishing boat position provided by EOG.

[0094] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A method for automatically identifying luminous fishing boats by combining peak detection with local detection, characterized in that: The steps include: (1) Obtain VIIRS / DNB night light remote sensing image data and perform preprocessing; (2) using a peak detection method to perform peak detection on the pre-processed night light remote sensing image data to obtain preliminary fishing boat points to be determined; (3) performing local peak detection on the image data of the preliminary fishing boat point to be determined after the peak detection to obtain the fishing boat point to be determined; (4) Eliminate false alarm factors for the fishing boat point data to be determined, and obtain a fishing boat identification result.

2. The method for automatically identifying luminous fishing boats by combining peak detection with local detection according to claim 1 is characterized in that: The preprocessing includes: geometric correction of the longitude and latitude of the night light remote sensing image data and clipping of the sea area to be studied; The geometric correction of the longitude and latitude of the night light remote sensing image data is used to eliminate the geometric distortion in the night light remote sensing image so that the geographic coordinates in the image are consistent with the real-world position; The clipping for the sea area to be studied is to clip the night light remote sensing image data using the land mask and the longitude and latitude range of the sea area to be studied to obtain the image data of the sea area to be studied.

3. The method for automatically identifying luminous fishing boats by combining peak detection with local detection according to claim 1 is characterized in that: The step (2) is specifically: (2.1) performing radiation amplification on the pre-processed night light remote sensing image data and performing radiation enhancement using logarithmic transformation log 10 (rad); (2.2) performing noise reduction on the image data after radiation enhancement processing; The noise reduction is carried out by Wiener filtering and median filtering in sequence; (2.3) subtracting the image data after noise reduction processing from the original image data to obtain a difference, and screening out points above the first threshold; (2.4) The image data that has been over-radiation-enlarged but not logarithmically transformed in step (2.1) is traversed in the horizontal and vertical directions respectively, and the ratio between the target pixel and the mean of the two adjacent pixels is calculated to obtain the peak height index: Among them: radiance of the subject pixel is the target pixel, average radiance of the two adjacent pixels is the average radiance of the two adjacent pixels of the target pixel; Screening the points whose peak height index is greater than a second threshold; (2.5) Horizontal strip lightning detection is performed on the image data after radiometric enhancement, and the pixels marked as lightning will be discarded; (2.6) For the intersection of the results gradually screened out in steps (2.3) to (2.4) and the points not marked as lightning, the points above the third threshold are screened out and retained as the preliminary fishing boat points to be determined, and the remaining non-preliminary fishing boat points to be determined are set to 0.

4. The method for automatically identifying luminous fishing boats by combining peak detection with local detection according to claim 3 is characterized in that: The first threshold is 0.75; the second threshold is 0.035; The third threshold is 0.

6.

5. The method for automatically identifying luminous fishing boats by combining peak detection with local detection according to claim 3 is characterized in that: The Wiener filter and median filter both use a 3×3 window.

6. The method for automatically identifying luminous fishing boats by combining peak detection with local detection according to claim 3 is characterized in that: The horizontal band lightning detection is as follows: for the image data after radiation enhancement processing, the radiation difference between the separated 16 rows of pixels is calculated, and the area where the difference is greater than 0.1Log10 DNB and lasts for 24 or more continuous pixels in the horizontal direction is marked as lightning.

7. The method for automatically identifying luminous fishing boats by combining peak detection with local detection according to claim 1 is characterized in that: The step (3) is specifically: (3.1) For the image data after peak detection, start from the first column and compare the radiation values ​​between two pixels vertically. The radiation value of the pixel with the larger radiation value is retained, and the smaller value is changed to 0. (3.2) After traversing the entire column, repeat step (3.1) starting from the data in the second column until all columns are traversed; (3.3) Then, traverse the results of the column traversal horizontally starting from the first row, compare the pixel sizes, and retain the larger pixel values ​​until all rows are traversed. (3.4) The points that remain after column traversal and row traversal are used as the fishing boat points to be determined.

8. The method for automatically identifying luminous fishing boats by combining peak detection with local detection according to claim 1 is characterized in that: The false alarm factors include position-invariant point light sources, position-variable point light sources and gas flares; The position-invariant point light sources include nearshore light sources, island and reef light sources, and offshore drilling platforms. A buffer zone is set as a mask to remove luminous pixels within the mask. The position-variable point light source includes a merchant ship, and the navigation track data of non-fishing vessels is obtained through AIS, and the luminous pixels within the mask are removed using the track mask; The gas flare was acquired using the M10 band of VIIRS and was used to remove the effects of the gas flare in the DNB image.

9. The method for automatically identifying luminous fishing boats by combining peak detection with local detection according to claim 8 is characterized in that: For the nearshore light source, a 2 km buffer zone is created outside the land as a continental mask, and the luminous pixels within the mask are removed; For the island and reef light sources, a 2-kilometer buffer zone is created outside the sea area as an island and reef mask, and the luminous pixels within the mask are removed; For each of the offshore drilling platforms, a 1-km buffer zone is created outside each drilling platform as a drilling platform mask, and the luminous pixels within the mask are removed.

Images