Marine radar rainfall image recognition method based on image hierarchical processing

Abstract

The invention belongs to the technical field of image recognition and detection, and particularly relates to a marine radar rainfall image recognition method based on image hierarchical processing, which comprises the following steps: S1, acquiring an original marine radar image and converting the original marine radar image to obtain an original grey-scale map; s2, processing to obtain 8 bit layer images; s3, calculating the Pearson's correlation coefficient of each bit layer image and the original grey-scale map, and selecting the bit layer image with the highest Pearson's correlation coefficient as a feature information plane graph; s4, n images under different rainfall conditions are selected, and n images without rainfall are selected; processing the selected image, performing zero-intensity percentage calculation by using a feature information plane graph, and selecting a threshold value of the image influenced by rainfall; s5, acquiring and processing a real-time marine radar image, comparing the zero intensity percentage with a selected threshold value, and judging whether the image is an image influenced by rainfall or not; according to the method, the marine radar images which are influenced by different rainfall conditions and are not influenced by rainfall can be efficiently and accurately classified.

Description

Technical Field

[0001] The invention belongs to the technical field of image recognition and detection, and in particular relates to a method for recognizing rainfall images of marine radars based on image layering processing. Background Art

[0002] In the process of using marine radar to extract sea surface wind field information, radar echo parameters play a vital role, which can reveal the wind field information on the sea surface. However, radar echoes cannot always provide clear and accurate data as we wish, because they are affected by a variety of interference factors, among which rainfall interference is particularly significant. Rainfall has many effects on marine radar wave measurement, including but not limited to signal attenuation, multipath effect, and waveform distortion. These effects not only increase the complexity of rainfall data processing, but also make the analysis of sea surface wind field conditions particularly difficult.

[0003] At present, most of the processing methods for marine radar rainfall images adopt a screening strategy: first identify the images disturbed by rainfall, and then correct these images to retain the parts that are less disturbed by rainfall, while removing the parts that are more seriously disturbed. Although this approach can improve the interpretation and analysis accuracy of radar images to a certain extent, there are still many challenges. The key is how to efficiently and accurately identify images affected by rainfall, which is the first step and a crucial link in research in this field.

[0004] Traditional methods for identifying rainfall images of marine radar, such as the zero intensity percentage method, the K parameter method, the echo mean and the coefficient of variation method, etc., attempt to identify rainfall interference images through radar image parameters and their combination, but their essence is still to analyze and process all the data of the entire image. This method does not extract the feature information represented by the effective data very well, so the accuracy of rainfall image recognition is low and cannot meet the engineering standards of actual detection.

[0005] After an in-depth analysis of the existing technical problems, it was found that the main reason is that the traditional method does not make full use of the effective information in the image when processing radar images, but processes the entire image indiscriminately. This processing method is not only inefficient, but also difficult to accurately identify images affected by rainfall. In addition, due to the diversity and complexity of rainfall conditions, traditional recognition methods are often difficult to adapt to the image classification needs under different rainfall conditions.

[0006] Therefore, how to efficiently and accurately classify marine radar images affected by different rainfall conditions and those not affected by rainfall, so as to provide an effective way to analyze the sea surface wind field information of marine radar images under rainfall interference, has become a problem that needs to be solved urgently. Summary of the invention

[0007] In view of the above-mentioned deficiencies in the prior art, the present invention provides a method for recognizing marine radar rainfall images based on image layering processing, which can efficiently and accurately classify marine radar images affected by different rainfall conditions and those not affected by rainfall, thereby providing an effective way to analyze the sea surface wind field information of marine radar images under rainfall interference.

[0008] In order to solve the above technical problems, the present invention adopts the following technical solutions: A method for recognizing marine radar rainfall images based on image layering processing comprises the following steps: S1. Obtain the original marine radar image and convert its grayscale value to obtain the corresponding original grayscale image; S2, performing bit layering processing on the original grayscale image to obtain 8 bit layer images; S3, calculating the Pearson correlation coefficient between each bit layer image and the original grayscale image, and selecting the bit layer image with the highest Pearson correlation coefficient as the feature information plane image that best represents the feature information of the original grayscale image; S4, selecting n images of different rainfall conditions, and selecting n images without rainfall; processing each selected image to obtain a corresponding feature information plane map, using the obtained feature information plane map to calculate the zero intensity percentage, and statistically analyzing the zero intensity percentage distribution corresponding to the images without rainfall and the images with rainfall, and then selecting a threshold for identifying images affected by rainfall based on the statistical results, if the zero intensity percentage is less than the threshold, the corresponding image is judged to be a non-rainfall image, otherwise the corresponding image is judged to be an image affected by rainfall; S5, acquiring and processing a real-time marine radar image, calculating a zero intensity percentage of the obtained feature information plane map, and comparing the obtained zero intensity percentage with the threshold value selected in S4 to determine whether the marine radar image is an image affected by rainfall; S6. Perform corresponding subsequent analysis and processing according to the judgment result of S5.

[0009] Compared with the prior art, the present invention has the following beneficial effects: 1. This method first converts the grayscale value of the original marine radar image to obtain the original grayscale image. Subsequently, 8 bit layer images are formed through bit layering. This process effectively decomposes the original image into multiple layers, each of which contains different feature information of the image. Next, by calculating the Pearson correlation coefficient between each bit layer image and the original grayscale image, the bit layer image with the highest correlation coefficient is selected as the feature information plane map. This step utilizes the characteristic that the Pearson correlation coefficient can measure the degree of linear correlation between two variables, so as to accurately extract the bit layer image that best represents the feature information of the original grayscale image. After obtaining the feature information plane map, this method selects n images of different rainfall conditions and no rainfall conditions, calculates the zero intensity percentage, and counts the zero intensity percentage distribution corresponding to the no rainfall image and the rainfall image. According to the statistical results, a threshold is selected to identify images affected by rainfall. This step utilizes the characteristic that the zero intensity percentage can reflect the proportion of zero intensity pixels in the image, so as to effectively distinguish images affected by rainfall from images not affected by rainfall. Finally, for the real-time marine radar image acquired, the method also performs the above processing to obtain the feature information plane map and calculate the zero intensity percentage. By comparing the calculation result with the selected threshold, it can be determined whether the image is affected by rainfall.

[0010] Compared with the existing technology, this method can quickly identify images affected by rainfall through image layering processing and feature information extraction, greatly improving processing efficiency. Through the calculation of Pearson correlation coefficient and zero intensity percentage, this method can accurately determine whether an image is affected by rainfall, avoiding the misjudgment and missed judgment that may occur in traditional methods.

[0011] 2. This method has strong adaptability. This method can process images under different rainfall conditions, including light rain and heavy rain, and has strong adaptability. The processing flow of this method is clear and easy to understand and implement. At the same time, by adjusting relevant parameters and algorithms, the method can be further optimized and improved to adapt to different application scenarios and needs.

[0012] 3. The method has a high degree of automation. This method realizes the automatic classification and processing of marine radar images, reduces the subjectivity of manual intervention and judgment, and improves the accuracy and reliability of classification.

[0013] In summary, this method can efficiently and accurately classify marine radar images affected by different rainfall conditions and those not affected by rainfall, thus providing an effective way to analyze the sea surface wind field information of marine radar images under rainfall interference. This method helps to more efficiently and accurately identify rainfall images when marine radar is disturbed by rainfall, providing a better basis for further extracting sea surface wind field information and improving the accuracy of wind field information extraction.

[0014] Preferably, in S2, the grayscale value of each pixel in the original grayscale image is converted into an 8-bit binary number, and then each bit of the 8-bit binary number is extracted respectively to form an 8-bit layer image.

[0015] This setting allows for fine-grained layered processing of the image. Each bit-layer image represents the characteristic information of the original grayscale image within a different grayscale range. This layered approach helps capture and analyze the detailed features of the image in a more detailed manner. The generation of eight bit-layer images is equivalent to decomposing and reorganizing the information of the original grayscale image, so that each bit-layer image focuses on expressing information within a specific grayscale range. This processing method enhances the information expression capability of the image, allowing subsequent analysis and processing to more accurately capture the key features of the image. In the subsequent image recognition and processing process, by analyzing the bit-layer images, it is easier to identify image areas affected by rainfall or other interference factors.

[0016] Preferably, in S3, the calculation formula of the Pearson correlation coefficient is: r = cov(X,Y) / (std(X)std(Y)); In the formula, r represents the Pearson correlation coefficient between the original grayscale image X and the bit layer image Y; cov(X,Y) represents the covariance of X and Y, and std(X) and std(Y) represent the standard deviation of X and Y respectively.

[0017] With this setting, 1. The correlation can be quantified. The Pearson correlation coefficient r provides a quantitative indicator for measuring the degree of linear correlation between two variables (in this case, the original grayscale image and the bit layer image). The value of r is between -1 and 1. When r is close to 1, it indicates a strong positive correlation between the two variables; when r is close to -1, it indicates a strong negative correlation; when r is close to 0, it indicates that there is almost no linear relationship between the two variables. This quantitative method allows us to more accurately evaluate the ability of the bit layer image to express the feature information of the original grayscale image.

[0018] 2. Redundant information can be reduced. Since each bit layer image contains the feature information of the original grayscale image at different gray levels, there may be redundant information. By calculating the Pearson correlation coefficient and selecting the bit layer image with the highest correlation coefficient, redundant information can be effectively reduced, and only the bit layer image that is most critical to the expression of the feature information of the original grayscale image can be retained.

[0019] 3. By calculating the Pearson correlation coefficient between each bit layer image and the original grayscale image, the bit layer image with the highest correlation coefficient can be selected as the feature information plane. This step is actually feature selection, which aims to select the image that best represents the feature information of the original grayscale image from multiple bit layer images. This feature selection method helps to improve the accuracy of subsequent image recognition and processing.

[0020] Preferably, in S4, the process of selecting a threshold for identifying images affected by rainfall includes: First, according to the statistical results, the main distribution interval of the zero intensity percentage of the non-rainfall image and the main distribution interval of the zero intensity percentage of the non-rainfall image are determined; the main distribution interval is an interval where the distribution density of the zero intensity percentage is greater than a preset density; Then, the interval between the two main distribution intervals is used as the threshold range interval; Then, a threshold for identifying images affected by rainfall is selected from the threshold range.

[0021] Such a setting, 1. can accurately distinguish between rainy and non-rainy images. By counting the main distribution intervals of the zero intensity percentage of non-rainy images, a benchmark range can be determined, which represents the typical characteristics of non-rainy images in terms of zero intensity percentage. Similarly, if the main distribution intervals of the zero intensity percentage of rainy images are also counted (although the original text does not directly mention it, this step should be included logically for contrast), images affected by rain can be more accurately identified. This is because rainy images may have significantly different distribution characteristics in terms of zero intensity percentage, such as a higher proportion of zero-intensity pixels. Using the interval between the two main distribution intervals as the threshold range interval is actually looking for a boundary that can best distinguish between rainy and non-rainy images.

[0022] 2. It can improve the accuracy and robustness of recognition. The selected threshold is based on the statistical results of a large number of non-rainfall images, so it is highly representative. By setting the threshold to a value between the two main distribution intervals, the possibility of misjudgment can be minimized, that is, the situation of misjudging non-rainfall images as rainfall images (or vice versa) can be reduced. This statistical-based method has good adaptability to images of different weather conditions, different time periods and different geographical locations, and therefore has high robustness.

[0023] 3. It can simplify the recognition process. Once the threshold for identifying images affected by rainfall is determined, this threshold can be applied to the subsequent image recognition process. During the recognition process, it is only necessary to calculate the zero intensity percentage of the image to be identified and compare it with the threshold to quickly determine whether the image is affected by rainfall. This simple comparison operation greatly improves the recognition efficiency and enables the system to process large amounts of image data in real time or near real time.

[0024] Preferably, the average value of the maximum value and the minimum value of the threshold range interval is selected as the threshold.

[0025] Such a setting, 1. Selecting the average of the maximum and minimum values ​​as the threshold, actually finds a midpoint between the two extreme distributions. This midpoint is neither completely biased towards the distribution of non-rain images nor completely biased towards the distribution of rain images, so it maintains a balance to a certain extent. This balance helps reduce the possibility of misjudgment and makes the recognition result more robust.

[0026] 2. Although choosing the average value as the threshold may not be optimal (because the optimal threshold may vary depending on the data set), it provides a reasonable compromise in most cases. The average value as the threshold can usually adapt well to changes in different data sets and maintain robustness to a certain extent. Even if there are some minor changes in the data set, the average value as the threshold can usually still provide acceptable recognition performance.

[0027] Preferably, in S4, different rainfall conditions include light rain, moderate rain and heavy rain.

[0028] This setting ensures that the statistical zero intensity percentage distribution is more comprehensive and accurate by selecting images with different rainfall conditions (light rain and heavy rain) and images without rainfall, and processing these images to obtain the feature information plane map. This comprehensiveness helps to more accurately identify images affected by rainfall. Since two different rainfall conditions, light rain and heavy rain, are taken into account, the technical content can take into account their respective characteristics when setting the threshold. This adaptability enables the system to better cope with image recognition tasks under different rainfall conditions, improving the robustness and practicality of the system.

[0029] Preferably, in S4, 80≤n≤200.

[0030] With this setting, 80 to 200 images are selected for statistics, ensuring the adequacy of the sample size. Within this range, the sample size is neither too large to cause excessive processing costs nor too sparse to make the statistical results unreliable. A sufficient number of samples helps to more accurately reflect the distribution characteristics of zero-intensity percentages of non-rain images and rain images (including light, medium and heavy rain), thereby improving the accuracy of threshold setting. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to make the purpose, technical solution and advantages of the invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings, in which: Figure 1 This is the flow chart of the method; Figure 2 The original marine radar image in the example of the second embodiment; Figure 3 is the original grayscale image in the example of embodiment 2; Figure 4 A comparison diagram of the original grayscale image and its eight-bit layered image in the example of the second embodiment; Figure 5 A comparison diagram of the original grayscale image and its seventh bit layer under different rainfall conditions in the example of the second embodiment; Figure 6 It is a schematic diagram of the proportion of the seventh bit layer data under the influence of different rainfall conditions in the example of the second embodiment; Figure 7 Schematic diagram of the zero intensity percentage distribution of the 7th bit layer under different rainfall conditions in the example of the second embodiment. DETAILED DESCRIPTION

[0032] The following is a further detailed description through specific implementation methods: Embodiment 1 like Figure 1 As shown, this embodiment discloses a A method for recognizing marine radar rainfall images based on image layering processing comprises the following steps: S1. Obtain an original marine radar image and perform grayscale value conversion on it to obtain a corresponding original grayscale image.

[0033] During specific implementation, the original marine radar image can be randomly selected.

[0034] S2. Perform bit layering processing on the original grayscale image to obtain 8 bit layer images.

[0035] In the specific implementation, the grayscale value of each pixel in the original grayscale image (0-255, a total of 256 grayscale levels) is converted into an 8-bit binary number, and then each bit of the 8-bit binary number is extracted to form 8 bit layer images. In this way, the image is processed in a refined layered manner. Each bit layer image represents the feature information of the original grayscale image in different grayscale ranges. This layering method helps to capture and analyze the detailed features in the image more carefully. The generation of 8 bit layer images is equivalent to decomposing and reorganizing the information of the original grayscale image, so that each bit layer image focuses on expressing the information within a specific grayscale range. This processing method enhances the information expression ability of the image, so that subsequent analysis and processing can more accurately capture the key features in the image. In the subsequent image recognition and processing process, through the analysis of the bit layer image, it is easier to identify the image area affected by rainfall or other interference factors.

[0036] S3. Calculate the Pearson correlation coefficient between each bit layer image and the original grayscale image, and select the bit layer image with the highest Pearson correlation coefficient as the feature information plane image that best represents the feature information of the original grayscale image.

[0037] The calculation formula of Pearson correlation coefficient is: r = cov(X,Y) / (std(X)std(Y)); In the formula, r represents the Pearson correlation coefficient between the original grayscale image X and the bit layer image Y; cov(X,Y) represents the covariance of X and Y, and std(X) and std(Y) represent the standard deviation of X and Y respectively.

[0038] In this way, the correlation can be quantified. The Pearson correlation coefficient r provides a quantitative indicator for measuring the degree of linear correlation between two variables (in this case, the original grayscale image and the bit layer image). The value of r ranges from -1 to 1. When r is close to 1, it indicates a strong positive correlation between the two variables; when r is close to -1, it indicates a strong negative correlation; when r is close to 0, it indicates that there is almost no linear relationship between the two variables. This quantification method allows us to more accurately evaluate the ability of the bit layer image to express the feature information of the original grayscale image. It can also reduce redundant information. Since each bit layer image contains the feature information of the original grayscale image at different gray levels, there may be redundant information. By calculating the Pearson correlation coefficient and selecting the bit layer image with the highest correlation coefficient, the redundant information can be effectively reduced, and only the bit layer image that is most critical to the expression of the feature information of the original grayscale image can be retained. In addition, by calculating the Pearson correlation coefficient between each bit layer image and the original grayscale image, the bit layer image with the highest correlation coefficient can be selected as the feature information plane. This step is actually feature selection, which aims to select the image that best represents the feature information of the original grayscale image from multiple bit layer images. This feature selection method helps to improve the accuracy of subsequent image recognition and processing.

[0039] S4. Select n images of different rainfall conditions and select n images without rainfall; process each selected image to obtain a corresponding feature information plane map, use the obtained feature information plane map to calculate the zero intensity percentage, and count the zero intensity percentage distribution corresponding to the non-rainfall images and rainfall images, and then select a threshold for identifying images affected by rainfall based on the statistical results. If the zero intensity percentage is less than the threshold, the corresponding image is judged to be a non-rainfall image, otherwise the corresponding image is judged to be an image affected by rainfall.

[0040] Among them, different rainfall conditions include light rain and heavy rain. By selecting images of different rainfall conditions (light rain and heavy rain) and images without rainfall, and processing these images to obtain feature information plane maps, it can ensure that the statistical zero intensity percentage distribution is more comprehensive and accurate. This comprehensiveness helps to more accurately identify images affected by rainfall. Since two different rainfall conditions, light rain and heavy rain, are taken into account, the technical content can take into account their respective characteristics when setting the threshold. This adaptability enables the system to better cope with image recognition tasks under different rainfall conditions, improving the robustness and practicality of the system.

[0041] In specific implementation, 80≤n≤200. In this way, 80 to 200 images are selected for statistics to ensure the adequacy of the sample number. Within this range, the sample number is neither too large to cause excessive processing costs nor too sparse to make the statistical results unreliable. A sufficient number of samples helps to more accurately reflect the distribution characteristics of no-rainfall images and rainfall images (including light rain and heavy rain) in terms of zero intensity percentage, thereby improving the accuracy of threshold setting.

[0042] In practice, the process of selecting a threshold for identifying images affected by rainfall includes: First, according to the statistical results, the main distribution interval of the zero intensity percentage of the non-rainfall image and the main distribution interval of the zero intensity percentage of the non-rainfall image are determined; the main distribution interval is an interval where the distribution density of the zero intensity percentage is greater than a preset density; Then, the interval between the two main distribution intervals is used as the threshold range interval; Then, a threshold for identifying images affected by rainfall is selected from the threshold range.

[0043] In this embodiment, the average value of the maximum and minimum values ​​of the threshold range interval is selected as the threshold. Selecting the average value of the maximum and minimum values ​​as the threshold is actually finding a midpoint between the two extreme distributions. This midpoint is neither completely biased towards the distribution of non-rain images nor completely biased towards the distribution of rain images, so it maintains balance to a certain extent. This balance helps to reduce the possibility of misjudgment and makes the recognition results more robust. Although choosing the average value as the threshold may not be optimal (because the optimal threshold may vary depending on the data set), in most cases it provides a reasonable compromise. The average value as a threshold can usually adapt well to changes in different data sets and maintain robustness to a certain extent. Even if there are some minor changes in the data set, the average value as a threshold can usually still provide acceptable recognition performance.

[0044] Such a setting can accurately distinguish between rainy and non-rainy images. By counting the main distribution intervals of the zero intensity percentage of non-rainy images, a benchmark range can be determined, which represents the typical characteristics of non-rainy images in terms of zero intensity percentage. Similarly, if the main distribution intervals of the zero intensity percentage of rainy images are also counted (although the original text does not directly mention it, this step should be included logically for comparison), images affected by rain can be more accurately identified. Because rainy images may have significantly different distribution characteristics in terms of zero intensity percentage, such as a higher proportion of zero intensity pixels. Taking the interval between the two main distribution intervals as the threshold range interval is actually looking for a boundary that can best distinguish between rainy and non-rainy images. It can also improve the accuracy and robustness of recognition. The selected threshold is based on the statistical results of a large number of non-rainy images, so it has a high representativeness. By setting the threshold to a value between the two main distribution intervals, the possibility of misjudgment can be minimized, that is, the situation of misjudging non-rainy images as rainy images (or vice versa) can be reduced. This statistical-based method has good adaptability to images of different weather conditions, different time periods and different geographical locations, and therefore has high robustness. Furthermore, the recognition process can be simplified. Once the threshold for identifying images affected by rainfall is determined, this threshold can be applied to subsequent image recognition processes. During the recognition process, it is only necessary to calculate the zero intensity percentage of the image to be recognized and compare it with the threshold to quickly determine whether the image is affected by rainfall. This simple comparison operation greatly improves the recognition efficiency and enables the system to process large amounts of image data in real time or near real time.

[0045] S5, acquiring and processing a real-time marine radar image, calculating a zero intensity percentage of the obtained feature information plane map, and comparing the obtained zero intensity percentage with the threshold value selected in S4 to determine whether the marine radar image is an image affected by rainfall; S6. Perform corresponding subsequent analysis and processing according to the judgment result of S5.

[0046] The method first converts the grayscale value of the original marine radar image to obtain the original grayscale image. Subsequently, 8 bit layer images are formed through bit layering. This process effectively decomposes the original image into multiple layers, each of which contains different feature information of the image. Next, by calculating the Pearson correlation coefficient between each bit layer image and the original grayscale image, the bit layer image with the highest correlation coefficient is selected as the feature information plane map. This step utilizes the characteristic that the Pearson correlation coefficient can measure the degree of linear correlation between two variables, so as to accurately extract the bit layer image that best represents the feature information of the original grayscale image. After obtaining the feature information plane map, the method selects n images of different rainfall conditions and no rainfall conditions, calculates the zero intensity percentage, and counts the zero intensity percentage distribution corresponding to the no rainfall image and the rainfall image. According to the statistical results, a threshold is selected to identify images affected by rainfall. This step utilizes the characteristic that the zero intensity percentage can reflect the proportion of zero intensity pixels in the image, so as to effectively distinguish images affected by rainfall from images not affected by rainfall. Finally, for the acquired real-time marine radar image, the method also performs the above processing to obtain the feature information plane map and calculate the zero intensity percentage. By comparing the calculation result with the selected threshold, it can be determined whether the image is an image affected by rainfall. Compared with the prior art, the method can quickly identify images affected by rainfall through image layering processing and feature information extraction, greatly improving the processing efficiency. Through the calculation of the Pearson correlation coefficient and the zero intensity percentage, the method can accurately determine whether the image is affected by rainfall, avoiding the misjudgment and missed judgment that may occur in the traditional method. In addition, the method has strong adaptability. The method can process images under different rainfall conditions, including light rain and heavy rain, and has strong adaptability. The processing flow of the method is clear and easy to understand and implement. At the same time, by adjusting the relevant parameters and algorithms, the method can be further optimized and improved to adapt to different application scenarios and needs. In addition, the method has a high degree of automation. The method realizes the automatic classification and processing of marine radar images, reduces the subjectivity of manual intervention and judgment, and improves the accuracy and reliability of classification.

[0047] This method can efficiently and accurately classify marine radar images affected by different rainfall conditions and those not affected by rainfall, thus providing an effective way to analyze the sea surface wind field information of marine radar images under rainfall interference. This method helps to more efficiently and accurately identify rainfall images when marine radar is disturbed by rainfall, providing a better basis for further extracting sea surface wind field information and improving the accuracy of wind field information extraction.

[0048] Embodiment 2 In order to better understand the technical content of this method, a specific example is used for illustration.

[0049] Randomly select an original marine radar image without rainfall, such as Figure 2 As shown in Figure 1, the marine radar image is not affected by rainfall interference. The original marine radar image is converted into grayscale value to obtain the original grayscale image, as shown in Figure 1. Figure 3 shown.

[0050] After that, the grayscale values ​​of each pixel of the converted grayscale image are subjected to bit weight mapping calculation, and the bit plane layered analysis is performed. Finally, the comparison diagram of the original grayscale image and its eight-bit layered image is obtained after integration. Figure 4 shown.

[0051] pass Figure 4 It can be clearly seen that each bit layer contains different feature information, and the feature information contained in the higher-level bit layer is more accurate. In this example, the 1st, 2nd, 3rd, and 4th bit planes are low-significant bit planes, and the meaningful feature information is extremely unclear. The 5th, 6th, 7th, and 8th bit planes are high-significant bit planes, and it can be clearly seen that these four planes contain a large amount of original image feature information, especially the two higher planes.

[0052] In order to further determine the correlation between each bit layer and the feature information of the original image, the eight bit layers obtained by decomposition are compared and analyzed with the original image for feature information, and the correlation between each bit layer and the feature information contained in the original image is determined by calculating the Pearson correlation coefficient between the eight bit layers and the original image respectively. The specific calculation formula is the same as that in Example 1 and will not be repeated here.

[0053] The calculation results are shown in Table 1: Table 1 Pearson correlation coefficients between each bit layer image and the original image Bit Plane 1 2 3 4 5 6 7 8 Pearson correlation coefficient -0.003 0.004 0.006 -0.011 -0.313 0.455 0.897 0.618 The calculation results show that in this example, the Pearson correlation coefficient between the seventh bit layer and the original image is closest to 1, and its correlation coefficient is much larger than that of other bit planes, which means that the seventh bit layer contains the most and most accurate feature information. In order to eliminate irrelevant information and retain and simplify useful feature information, the seventh bit plane image is selected instead of the original image for the next step of analysis.

[0054] The bit plane layered analysis is performed on the marine radar images that are not affected by rainfall disturbance, affected by small or medium (intensity) rainfall disturbance, and affected by large (intensity) rainfall disturbance, and their seventh bit plane images are extracted as follows: Figure 5 The zero intensity percentage analysis of the 7th bit layer of the marine radar image under different rainfall conditions is shown in the following figure. Figure 6From the data in the figure, it can be preliminarily seen that the zero intensity percentage decreases with the increase of rainfall, that is, with the increase of the interference of rainfall on the navigation radar image. When there is no rain, the zero intensity percentage is the largest.

[0055] In order to further determine the relationship between the zero intensity percentage of the 7th bit layer and rainfall interference, in this example, 350 marine radar images with different rainfall intensities were selected and their 7th bit layers were analyzed for zero intensity percentage. The data distribution obtained is as follows: Figure 7 As shown. In this data distribution diagram, the first 150 marine radar images were not disturbed by rainfall, and the last 200 marine radar images were disturbed by rainfall to varying degrees. From the zero intensity percentage characteristic distribution diagram, it can be seen that the zero intensity percentage of images not disturbed by rainfall is basically greater than 65.3%, between 65.3% and 78.3% (this interval is the interval where the distribution density of zero intensity percentage is greater than the preset density), while the zero intensity percentage of images disturbed by rainfall is basically less than 62.1%, between 50.1% and 62.1%.

[0056] The analysis further proves that as the rainfall disturbance increases, the zero intensity percentage of the feature quantity will decrease accordingly, and the value is regularly distributed, so the threshold range can be selected as 62.1% to 65.3%. From the data distribution diagram, it can be seen that the proportion of images with zero intensity percentage in the range of 62.1% to 65.3% is very small, so the threshold is selected as the average of 62.1% and 65.3%, that is, the selected threshold is 63.7%.

[0057] When the zero intensity percentage of the seventh bit layer of the marine radar image is greater than the threshold value of 63.7%, it is determined to be an image not disturbed by rainfall. When the zero intensity percentage of the seventh bit layer of the marine radar image is less than the threshold value of 63.7%, it is determined to be an image disturbed by rainfall.

[0058] In order to verify the recognition accuracy in the example, 100 marine radar images not affected by rainfall interference and 200 marine radar images affected by rainfall interference were prepared again, and all the images were identified and classified according to the patented method. The final recognition accuracy reached 95%. Compared with traditional methods, this method can more efficiently and accurately identify marine radar images affected by rainfall.

[0059] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit the technical solution. Those skilled in the art should understand that those modifications or equivalent substitutions of the technical solution of the present invention that do not depart from the purpose and scope of the technical solution should be included in the scope of the claims of the present invention.

Claims

1. A method for recognizing marine radar rainfall images based on image layering processing, characterized in that: The following steps are involved: S1. Obtain the original marine radar image and convert its grayscale value to obtain the corresponding original grayscale image; S2, performing bit layering processing on the original grayscale image to obtain 8 bit layer images; S3, calculating the Pearson correlation coefficient between each bit layer image and the original grayscale image, and selecting the bit layer image with the highest Pearson correlation coefficient as the feature information plane image that best represents the feature information of the original grayscale image; S4, selecting n images of different rainfall conditions, and selecting n images without rainfall; processing each selected image to obtain a corresponding feature information plane map, using the obtained feature information plane map to calculate the zero intensity percentage, and statistically analyzing the zero intensity percentage distribution corresponding to the images without rainfall and the images with rainfall, and then selecting a threshold for identifying images affected by rainfall based on the statistical results, if the zero intensity percentage is less than the threshold, the corresponding image is judged to be a non-rainfall image, otherwise the corresponding image is judged to be an image affected by rainfall; S5, acquiring and processing a real-time marine radar image, calculating a zero intensity percentage of the obtained feature information plane map, and comparing the obtained zero intensity percentage with the threshold value selected in S4 to determine whether the marine radar image is an image affected by rainfall; S6. Perform corresponding subsequent analysis and processing according to the judgment result of S5.

2. The method for recognizing marine radar rainfall images based on image layering processing according to claim 1, characterized in that: In S2, the grayscale value of each pixel in the original grayscale image is converted into an 8-bit binary number, and then each bit of the 8-bit binary number is extracted to form an 8-bit layer image.

3. The method for recognizing marine radar rainfall images based on image layering processing as claimed in claim 1, characterized in that: In S3, the calculation formula of Pearson correlation coefficient is: r = cov(X,Y) / (std(X)std(Y)); In the formula, r represents the Pearson correlation coefficient between the original grayscale image X and the bit layer image Y; cov(X,Y) represents the covariance of X and Y, and std(X) and std(Y) represent the standard deviation of X and Y respectively.

4. The method for recognizing marine radar rainfall images based on image layering processing as claimed in claim 1, characterized in that: In S4, the process of selecting a threshold for identifying images affected by rainfall includes: First, according to the statistical results, the main distribution interval of the zero intensity percentage of the non-rainfall image and the main distribution interval of the zero intensity percentage of the non-rainfall image are determined; the main distribution interval is an interval where the distribution density of the zero intensity percentage is greater than a preset density; Then, the interval between the two main distribution intervals is used as the threshold range interval; Then, a threshold for identifying images affected by rainfall is selected from the threshold range.

5. The method for recognizing marine radar rainfall images based on image layering processing as claimed in claim 4, characterized in that: Select the average of the maximum and minimum values ​​in the threshold range as the threshold.

6. The method for recognizing marine radar rainfall images based on image layering processing as claimed in claim 1, characterized in that: In S4, different rainfall conditions include light, moderate and heavy rain.

7. The method for recognizing marine radar rainfall images based on image layering processing as claimed in claim 6, characterized in that: In S4, 80≤n≤200.

Images