An adaptive image processing method for detecting the blackness of ship black smoke

Abstract

The application discloses a self-adaptive image processing method for detecting the blackness of ship black smoke, and belongs to the technical field of image processing. In the application, a target tracking algorithm utilizes Hamming distance to determine the confidence of a target tracking model in real time, and when the confidence of the target model is lower than a threshold value, the classifier is stopped from being updated; meanwhile, a template matching algorithm is used to reposition the target and reinitialize the tracker to complete long-time tracking of the target; a virtual Lingerman color card is made for the target image, and a stable light target in the image is set as Lingerman 0-grade blackness, and a stable dark target is set as Lingerman 5-grade blackness; for a single frame image, the blackness of the image is calculated after further refinement processing based on 1% of the darkest pixels in the black smoke sub-region. The application can effectively solve the technical problems that the size of the black smoke frame region selected by an artificial selection and the floating change of the ship black smoke in a video stream will cause calculation errors.

Description

TECHNICAL FIELD

[0001] The present application mainly relates to the technical field of image processing, and particularly relates to an adaptive image processing method for detecting blackness of ship black smoke. BACKGROUND

[0002] Ship black smoke is a serious phenomenon of atmospheric pollution of ships, and should be an important object of management and control of atmospheric pollution prevention of ships in China. However, the traditional Lignin blackness method suitable for fixed pollution sources has harsh application conditions and the results are relatively subjective, which cannot meet the needs of ship black smoke detection, and restricts the development of ship black smoke supervision capability. The prior art has the following problems:

[0003] 1. The ship black smoke exhaust is a non-rigid target. In actual application, the current visual target tracking technology is easily affected by external factors, such as shape and size changes of the target, light changes, target occlusion and the like, and it is difficult to realize high-precision and robust tracking of the black smoke exhaust.

[0004] 2. When calculating the blackness of ship black smoke, the black smoke frame region needs to be selected manually. The size of the black smoke frame region selected by different people is different, and the calculated blackness value is also different. The larger the selected black smoke frame region, the smaller the calculated blackness value; on the contrary, the larger the calculated blackness value, that is, there is a calculation error.

[0005] 3. The ship black smoke in the video stream is dynamically changing. If the darkest gray value in the exhaust is used to calculate the Lignin grade of the ship black smoke, the blackness grade of the ship exhaust in each frame image will fluctuate greatly, and the finally calculated blackness value of the video stream will be too large. SUMMARY

[0006] The technical solution of the present application provides a significantly different solution from the prior art to solve the technical problem of the prior art solution being too single, and mainly provides an adaptive image processing method for detecting blackness of ship black smoke to solve the technical problem of calculation error caused by manual selection of the size of the black smoke frame region and fluctuation of the ship black smoke in the video stream.

[0007] The technical solution adopted by the present application to solve the above technical problems is as follows:

[0008] An adaptive image processing method for detecting blackness of ship black smoke, comprising the following steps:

[0009] S1. Based on initial video data, a target tracking algorithm is used to extract a tracking video;

[0010] S2. The image frame sequence data in the tracking video is analyzed to obtain image frame data as target data;

[0011] S3, calculate the blackness of the black smoke in each frame of the target data one by one, and the specific steps for calculating the blackness of the black smoke in a frame of image are:

[0012] S3-1, perform gray scale conversion on the image, take the gray scale value as the horizontal coordinate and the pixel number as the vertical coordinate, count the histogram of the gray scale value of all pixels, set the bright target in the image as the Lignermann 0 level blackness, set the dark target in the image as the Lignermann 5 level blackness, and make a virtual Lignermann color card;

[0013] S3-2, select a sub-region of the black smoke of the ship exhaust, count the histogram of the gray scale value of all pixels with the gray scale value as the horizontal coordinate and the pixel number as the vertical coordinate;

[0014] S3-3, based on the result of step S3-2, extract the 1% of the pixels with the darkest pixels in the recognized black smoke sub-region;

[0015] S3-4, based on the result of step S3-3, extract the 1% of the pixels with the darkest pixels in the recognized black smoke sub-region to the pixels with the darkest pixels, and obtain two gray scale curves connected at the head and tail;

[0016] S3-5, based on the result of step S3-4, linearly fit the two gray scale curves to obtain two fitting curves, and extract the intersection point of the linear fitting of the two fitting curves, and take the blackness parameter corresponding to the intersection point as the black smoke blackness parameter of the frame of image;

[0017] S4, based on the calculation result of step S3, calculate the average value to obtain the black smoke level calculation result of the entire video.

[0018] The above scheme is improved, in step S1, the target tracking algorithm uses Hamming distance to determine the confidence of the target tracking model in real time, when the confidence of the target model is lower than the threshold, the classifier is stopped updating; at the same time, the template matching algorithm is used to reposition the target and reinitialize the tracker to complete the long time tracking of the target.

[0019] Specifically, the target tracking algorithm takes the ridge regression classifier as the core, adopts the method of cyclic matrix to perform shift operation in the target area, forms a large number of samples to train the classifier; the similarity between the selected area and the target area is calculated through the kernel function, and the region with the maximum response is taken as the new tracking target; in addition, the algorithm cleverly transforms the training samples to have the characteristics of cyclic matrix, diagonalizes the discrete Fourier transform matrix, reduces a large number of matrix inverse operations, and significantly improves the running speed of the algorithm.

[0020] Further, in step S3-1, the 1st blackness = 0th blackness*80%+5th blackness*20%, the 2nd blackness = 0th blackness*60%+5th blackness*40%, the 3rd blackness = 0th blackness*40%+5th blackness*60%, and the 4th blackness = 0th blackness*20%+5th blackness*80%.

[0021] As an implementation, in step S3-5, the blackness parameter is a gray value; in step S4, the average value is a gray value, which is converted into a Lenneman blackness level; the calculation formula for converting a gray value into a Lenneman blackness level is: B=5*(I0-I s ) / (I0-I5), wherein I0 is the gray value of 0th blackness, I5 is the gray value of 5th blackness, and I s is the gray value to be converted.

[0022] As another implementation, in step S3-5, the blackness parameter is a Lenneman blackness level corresponding to a gray value; the calculation formula for converting a gray value into a Lenneman blackness level is: B=5*(I0-I s ) / (I0-I5), wherein I0 is the gray value of 0th blackness, I5 is the gray value of 5th blackness, and I s is the gray value to be converted; in step S4, the average value is the Lenneman blackness level.

[0023] Compared with the prior art, the present application has the following advantages:

[0024] (1) The present application is different from the method for calculating the Lenneman level of ship black smoke by using the darkest gray value in exhaust gas, and based on the 1% darkest pixels in the sub-region of black smoke, the darkest pixels in the sub-region of black smoke are extracted to the 1% darkest pixels, two gray value curves with head and tail connection are obtained; then the two gray value curves are linearly fitted to obtain two fitting curves, and the intersection of the linear fitting of the two fitting curves is extracted, and the blackness parameter corresponding to the intersection represents the blackness parameter of the black smoke in the frame image. Through the test example, it can be known that the calculation results of the present application under different frame scales are less different, so that the influence of the size of the black smoke sub-region image frame on the calculation result can be effectively reduced; the change degree of the results obtained in different frames of the present application is small, that is, the error caused by the shape change of the black smoke can be reduced. Therefore, the algorithm provided by the present application has high accuracy.

[0025] (2) The present application makes a virtual Lingerman color card for each frame of image in the target data respectively, and sets the bright target (sky and white paint on the building and ship body, etc.) in the image as Lingerman 0 level blackness, and sets the dark target (shadow and black paint on the building and ship body and window, etc.) in the image as Lingerman 5 level blackness, since the bright target and the dark target are stably present in practice, each frame of image is calculated with the same reference standard, so that accurate data can be obtained, and the calculation result can avoid the influence of external factors such as scale change and light change.

[0026] (3) The target tracking algorithm in the optimization scheme of the present application uses Hamming distance to determine the confidence of the target tracking model in real time, and when the confidence of the target model is lower than the threshold value, the classifier is stopped from updating; at the same time, the template matching algorithm is used to reposition the target and reinitialize the tracker to complete long-time tracking of the target. The problem that the tracking frame may drift or even fail in the tracking process of the original KCF algorithm if the target is shielded for a short time and then appears in the field of view again is effectively solved. Then the present application can adapt to external factors such as deformation and scale change of the exhaust gas by fusing the model confidence determination mechanism and the template matching algorithm, so as to improve the tracking robustness and precision of the algorithm and lay a good foundation for subsequent image blackness detection calculation.

[0027] The present application will be explained in detail below in combination with the drawings and specific embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0028] Figure 1 The flow chart of the adaptive image processing method for detecting the blackness of ship exhaust gas provided by the present application;

[0029] Figure 2 The flow chart of the ship exhaust gas tracking algorithm in the second embodiment of the present application;

[0030] Figure 3 The example diagram when the virtual Lingerman color card is made in the first embodiment of the present application;

[0031] Figures 4-7 The gray value histogram processed in sequence in the process of adaptively identifying the Lingerman blackness of the exhaust gas in the first embodiment of the present application;

[0032] Figure 8 The images of four situations in the first test example of the present application;

[0033] Figure 9 The result diagram in the first test example of the present application;

[0034] Figure 10 The images of four situations in the second test example of the present application;

[0035] Figure 11 Fig. 2 is a result graph of the test case two of the present application;

[0036] Figure 12 Fig. 3 is a series of images of the four cases in the test case three of the present application;

[0037] Figure 13 Fig. 4 is a result graph of the test case three of the present application, in which curve a represents the curve of the darkest gray value in the ship smoke sub-region of each frame, and curve b represents the curve of the gray value representing the blackness of the smoke calculated by the adaptive method in the ship smoke sub-region of each frame.

[0038] Figure 14 Fig. 5 is a series of images of the sample and 11 frames in the test case four of the present application;

[0039] Figure 15 Fig. 6 is a result graph of the test case four of the present application, in which curve a represents the curve of the darkest gray value in the ship smoke sub-region of each frame, and curve b represents the curve of the gray value representing the blackness of the smoke calculated by the adaptive method in the ship smoke sub-region of each frame. DETAILED DESCRIPTION

[0040] In order to facilitate the understanding of the present application, the present application will be described more fully below with reference to the accompanying drawings, in which several embodiments of the present application are given. However, the present application can be realized in different forms and is not limited to the embodiments described herein. On the contrary, these embodiments are provided to make the disclosure of the present application more thorough and comprehensive.

[0041] An adaptive image processing method for detecting the blackness of ship smoke, comprising a tracking module and a smoke blackness detection module, wherein the tracking module is mainly responsible for tracking the part of the ship body rigid region and the smoke sub-region in the video data, and the tracking result of the tracking module is used as the input data of the smoke blackness detection module; the smoke blackness detection module first analyzes the image frame sequence data in the video, selects a certain number of image frame data as target data according to a certain extraction ratio and time interval; finally, the blackness level and rendering result of the target image frame are calculated by combining the tracking result and the information of the smoke sub-region.

[0042] Embodiment one: please refer to the accompanying drawings Figure 1 In this embodiment, the following steps are specifically included:

[0043] (1) The KCF algorithm is used to process the initial video data to obtain a tracking video in which the area of the ship smoke outlet is highlighted;

[0044] (2) The image frame sequence data in the tracking video is analyzed to obtain the image frame data as target data;

[0045] (3) The blackness of smoke of each image in the target data is calculated one by one, and the process of calculating the blackness of smoke of a frame of image is as follows:

[0046] (3-1) First make a virtual Linkeiman color card:

[0047] The target area of interest in the video (i.e. target image) is converted to grayscale, and the histogram of all pixel gray values (horizontal axis for gray value, vertical axis for pixel number) is counted; set the gray value of the darkest 1% of the pixels as the Linkeiman 5th level blackness (by default, the target area of interest will have dark targets such as shadows, black paint, colored paint, and windows), and set the gray value of the brightest 1% of the pixels as the Linkeiman 0th level blackness (by default, the target area of interest will have bright targets such as the sky and white paint), as shown in Figure 3 . The data determination of 1% of the pixels is an empirical value obtained through experiments. The darkest or brightest 1% of the pixels is closer to the actual 5th level blackness and 0th level blackness, and a larger range will result in a smaller actual value; a smaller range will result in a larger actual value.

[0048] 1st level blackness = 0th level blackness x 80% + 5th level blackness x 20%, 2nd level blackness = 0th level blackness x 60% + 5th level blackness x 40%, 3rd level blackness = 0th level blackness x 40% + 5th level blackness x 60%, and 4th level blackness = 0th level blackness x 20% + 5th level blackness x 80%.

[0049] Each level of blackness corresponds to a certain gray value, i.e. the virtual Linkeiman color card is completed.

[0050] (3-2) Self-adaptive identification of the Linkeiman blackness of exhaust gas:

[0051] a) Based on the target image, select the rectangular frame area where the black smoke is located, and for the non-rigid target pixels (i.e. sub-regions of ship exhaust black smoke) in the identified rectangular area, count the gray value histogram of all pixels with gray value as the horizontal axis and pixel number as the vertical axis (as shown in Figure 4 .

[0052] b) Based on Figure 4 , extract the 1% of the pixels with the darkest gray value in the black smoke region in the identified rectangular area, as shown in Figure 5 .

[0053] c) Based on Figure 5 , extract the 1% of the pixels with the darkest gray value in the black smoke region in the identified rectangular area, as shown in Figure 6 .

[0054] d) As shown in Figure 7 , linearly fit the red gray curve A and the red gray curve B extracted based on Figure 6 to obtain the blue fitting curves A1 and B1, and the intersection C of the two linearly fitted curves, and the Linkeiman blackness level corresponding to the gray value of the intersection C represents the Linkeiman level of the ship exhaust at that moment.

[0055] The calculation formula of the Ringelmann blackness level is B = 5 * (I0 - I5) / (I0 - I5), wherein I0 is the gray value of 0-level blackness, I5 is the gray value of 5-level blackness, and I is the gray value to be converted (in this embodiment, the gray value of the plume of the target image). s s

[0056] (4) Finally, the Ringelmann blackness calculation result of each frame of image in the target data is averaged, and the black smoke level calculation result of the entire video is obtained, which is a value that can be accurate to a decimal place. (At present, in order to adapt to the actual situation, the integer value closest to the calculation result is usually taken as the final blackness level of the ship exhaust smoke, for example, if the calculation result is 3.4, the 3-level blackness is taken; if the calculation result is 3.6, the 4-level blackness is taken).

[0057] Embodiment II: The difference between this embodiment and embodiment II is that:

[0058] In this embodiment, in step (1), the tracking module adopts a ship black smoke exhaust tracking algorithm, please refer to the attached drawings for details. Figure 2 The Hamming distance is used to determine the confidence of the target tracking model in real time, if the confidence of the target model is lower than a certain threshold (the threshold value is 0.5, which is a test experience value), the classifier is stopped from updating; at the same time, the template matching algorithm is used to reposition the target and reinitialize the tracker to complete the long-time tracking of the target. The ship black smoke exhaust tracking algorithm takes the ridge regression classifier as the core, adopts the method of circular matrix to perform the shift operation in the target area, forms a large number of samples to train the classifier. The similarity between the selected area and the target area is calculated through the kernel function, and the area with the maximum response is taken as the new tracking target. In addition, the algorithm cleverly transforms the training samples to have the characteristics of the circular matrix, diagonalizes the discrete Fourier transform matrix, reduces the operation of a large number of matrix inversions, and significantly improves the running speed of the algorithm, so as to realize the real-time tracking of the target.

[0059] Specifically, a more accurate result can be obtained through the perceptual hashing algorithm, which adopts DCT (Discrete Cosine Transform) to reduce the frequency. The full name of DCT is Discrete Cosine Transform, which is mainly used for data or image compression and can convert the spatial signal to the frequency domain, having good decorrelation performance.

[0060] The working principle of the perceptual hashing algorithm is as follows:

[0061] a) Reduce the size. In order to simplify the calculation of DCT, pHash starts with a small picture. ​​

[0062] b) Simplify color. The picture needs to be converted into a grayscale image, further simplifying the amount of calculation.

[0063] c) Calculate DCT. DCT is to decompose the picture into frequency aggregation and ladder shape.

[0064] d) Reduce DCT. The result of DCT is a 32x32 matrix, but only the 8x8 matrix in the upper left corner needs to be retained, which represents the lowest frequency in the picture.

[0065] e) Calculate the average value, calculate the average value of all 64 values.

[0066] f) Further reduce DCT. According to the comparison of the 8x8 DCT matrix, set it to "1" if it is greater than or equal to the average value of DCT, and set it to "0" if it is less than the average value of DCT. The overall structure of the picture remains unchanged, and the hash result value remains unchanged.

[0067] g) Construct hash value. Combine 64 bit positions to generate hash value, the order is random, but the consistency before and after is enough.

[0068] h) Compare fingerprints: Calculate the fingerprints of two pictures, and calculate the Hamming distance.

[0069] In d) of step (3-2): the intersection point corresponds to the gray value representing the ship's black smoke gray value at that moment;

[0070] In step (4), the average value obtained is the gray value, which is converted into the Lengman blackness level, and the calculation formula is: B = 5*(I0-I s ) / (I0-I5), wherein I0 is the gray value of 0-level blackness, I5 is the gray value of 5-level blackness, and I s is the gray value to be converted (in this embodiment, it is the average value of the gray values of all images of the entire video target data).

[0071] The other is the same as embodiment one.

[0072] The following compares the adaptive image blackness detection algorithm provided by the present application (indicated by adaptive algorithm in the table) with the Lengman blackness level algorithm (indicated by 1% in the table) of directly selecting the 1% darkest part of the pixels in the black smoke sub-region to verify the accuracy of the algorithm provided by the present application.

[0073] Test example one: comparison of results of selecting four different frame scales for the same frame image:

[0074] As Figure 8As shown, (a)-(d) are illustrations of four cases where the black smoke frame area gradually increases; the grayscale values ​​obtained using the two algorithms are shown in Table 1 and... Figure 9 .

[0075] Table 1 shows the data obtained from the two different algorithms in Test Case 1.

[0076] Number Block 1 Block 2 Block 3 Block 4 1％ 94 96 98 102 Adaptive algorithm 92 93 97 98

[0077] Test Example 2: Comparison of results using four different bounding box scales for the same frame image:

[0078] like Figure 10 As shown, (a)-(d) are illustrations of four cases where the black smoke frame area gradually increases; the grayscale values ​​obtained using the two algorithms are shown in Table 2 and... Figure 11 .

[0079] Table 2 shows the data obtained from the two different algorithms in Test Example 2.

[0080] Number Block 1 Block 2 Block 3 Block 4 1％ 117 118 121 125 Adaptive algorithm 113 114 116 116

[0081] Test Example 3: Comparison of results using four different bounding box scales for another image of the same frame:

[0082] like Figure 12 As shown, (a)-(d) are illustrations of four cases where the black smoke frame area gradually increases; the grayscale values ​​obtained by the two algorithms are shown in Table 3 and Figure 13 .

[0083] Table 3 shows the data obtained from the two different algorithms in Test Case 3.

[0084] Number Block 1 Block 2 Block 3 Block 4 1％ 38 40 45 48 Adaptive algorithm 36 36 36 36

[0085] As shown in Test Examples 1 to 3, calculating the blackness of smoke using the darkest 1% of pixels in the black smoke sub-region image frame yields significantly different results at different frame sizes. Furthermore, a larger black smoke sub-region image frame results in a higher image grayscale value (i.e., a larger image grayscale value) and a lower calculated blackness value; conversely, a smaller frame size results in a higher calculated blackness value. In contrast, the adaptive method calculates a grayscale value representing black smoke with less variation across different frame sizes, effectively reducing the impact of the black smoke sub-region image frame size on the calculation results. Therefore, the algorithm provided by this invention is convenient to apply and has high accuracy.

[0086] Test Example 4: Test the algorithm's performance using a set of videos.

[0087] Selecting 1 group of challenging videos (the ship in the video stream is moving, the plume of the black smoke is dynamically changing, and the darkest gray value in the sub-region of the black smoke of the ship exhaust in different frame images is different), based on the blackness adaptive image detection algorithm provided by the present application, the video is analyzed, as shown in Figure 14 The brightest part in the video image (the white paint on the ship body) is automatically searched as the 0-level blackness, and the dark target in the video (the black paint on the ship body) is automatically searched as the 5-level blackness.

[0088] The darkest gray value in the sub-region of the black smoke of the ship exhaust in each frame is used to calculate the gray value of the image of the 1st frame, the 51st frame, the 101st frame, the 151st frame, the 201st frame, the 251st frame, the 301st frame, the 351st frame, the 401st frame, the 490th frame, and the 540th frame, and the adaptive image blackness detection algorithm provided by the present application is used to calculate the average value, and then the blackness value of the video stream of the ship black smoke is calculated. The results are shown in Table 4 and Figure 15 .

[0089] Table 4: Data table obtained by two different algorithms in test example four

[0090]

[0091] As shown in Table 4 and Figure 15 , the gray value calculated by the darkest gray value in the sub-region of the black smoke of the ship exhaust in each frame has a large fluctuation, and the gray value calculated by the adaptive method has a relatively stable change curve, so the algorithm provided by the present application can reduce the error caused by the shape change of the black smoke and the change of the environment, and improve the calculation accuracy. Moreover, the blackness value of the video stream calculated by the darkest gray value in the sub-region of the black smoke of the ship exhaust in each frame is larger.

[0092] As shown in Figure 14 , the above exemplary description of the present application in combination with the drawings is obvious, and the specific implementation of the present application is not limited by the above method. As long as the method concept and technical solution of the present application are used for such non-essential improvement, or the concept and technical solution of the present application are directly applied to other occasions without improvement, they are within the protection scope of the present application.

Claims

1. An adaptive image processing method for detecting the blackness of ship black smoke, characterized in that: The method comprises the following steps: S1, based on the initial video data, a target tracking algorithm is used to extract the tracking video; S2, the image frame sequence data in the tracking video is analyzed, and the obtained image frame data is used as target data; S3, the black smoke darkness of each image in the target data is calculated one by one, and the specific steps for calculating the black smoke darkness of an image are as follows: S3-1, the image is converted into grayscale, the histogram of the grayscale values of all pixels is counted with the grayscale value as the horizontal coordinate and the pixel number as the vertical coordinate, the bright target in the image is set as the Lignermann 0-level darkness, the dark target in the image is set as the Lignermann 5-level darkness, and a virtual Lignermann color card is made; S3-2, a ship exhaust black smoke sub-region is selected, and the histogram of the grayscale values of all pixels is counted with the grayscale value as the horizontal coordinate and the pixel number as the vertical coordinate; S3-3, based on the result of step S3-2, the 1% darkest part of the pixels in the recognized black smoke sub-region is extracted; S3-4, based on the result of step S3-3, the darkest pixel point to the 1% darkest part of the pixels in the recognized black smoke sub-region is extracted, and two head-tail connected grayscale curves are obtained; S3-5, based on the result of step S3-4, the two grayscale curves are linearly fitted to obtain two fitting curves, and the intersection point of the linear fitting of the two fitting curves is extracted, and the darkness parameter corresponding to the intersection point is taken as the black smoke darkness parameter of the image; S4, based on the calculation result of step S3, the average value is calculated to obtain the black smoke level calculation result of the entire video.

2. The self-adaptive image processing method for detecting the blackness of ship black smoke according to claim 1, characterized in that: In step S1, the target tracking algorithm uses the Hamming distance to determine the confidence of the target tracking model in real time, and when the confidence of the target model is lower than a threshold value, the classifier is stopped from being updated; meanwhile, a template matching algorithm is used to reposition the target and reinitialize the tracker to complete long-time tracking of the target.

3. The self-adaptive image processing method for detecting the blackness of ship black smoke according to claim 2, characterized in that: The target tracking algorithm takes the ridge regression classifier as the core, adopts the method of cyclic matrix to perform shift operation on the target region, forms a large number of samples to train the classifier, and calculates the similarity between the selected region and the target region through the kernel function, and the region with the maximum response is taken as the new tracking target. In addition, the algorithm cleverly transforms the training samples to have the characteristics of the cyclic matrix, diagonalizes the discrete Fourier transform matrix, reduces a large number of matrix inverse operations, and significantly improves the running speed of the algorithm.

4. The self-adaptive image processing method for detecting the blackness of ship black smoke according to claim 1 or 3, characterized in that: In step S3-1, 1-level darkness = 0-level darkness * 80% + 5-level darkness * 20%, 2-level darkness = 0-level darkness * 60% + 5-level darkness * 40%, 3-level darkness = 0-level darkness * 40% + 5-level darkness * 60%, and 4-level darkness = 0-level darkness * 20% + 5-level darkness * 80%.

5. The self-adaptive image processing method for detecting the blackness of ship black smoke according to claim 1 or 3, characterized in that: In step S3-5, the blackness parameter is a gray value; in step S4, the average value is a gray value, which is converted into a Rinkenmann blackness level; the calculation formula for converting a gray value into a Rinkenmann blackness level is: B=5*(I0-I s ) / (I0-I5), wherein I0 is the gray value of 0-level blackness, I5 is the gray value of 5-level blackness, and I s is the gray value to be converted.

6. The self-adaptive image processing method for detecting the blackness of ship black smoke according to claim 1 or 3, characterized in that: In step S3-5, the blackness parameter is the corresponding Lennert blackness level of the gray value, and the calculation formula for converting the gray value into the Lennert blackness level is B=5*(I0-I s ) / (I0-I5), wherein I0 is the gray value of 0-level blackness, I5 is the gray value of 5-level blackness, and I s is the gray value to be converted; in step S4, the average value is the Lennert blackness level.

Images