Coal spillage identification method based on information entropy

Abstract

This invention discloses a coal spill identification method based on information entropy, relating to the field of detection technology. The method includes the following steps: Step 1: For each frame image of the monitored area, convert it into a grayscale image and represent it using matrix A; Step 2: Perform edge detection on matrix A, and record the result as matrix C; Step 3: Calculate the information entropy of matrix C to obtain the frame entropy H of the grayscale image; Step 4: Process each frame of the real-time frame images generated sequentially in the monitored area through steps 1 to 3 to obtain a frame entropy sequence; Step 5: Calculate the frame entropy mean μ and variance σ of the frame entropy sequence. 2 Determine the threshold H for frame entropy alarm. gate Step 6: When the frame entropy of several consecutive frames exceeds the threshold H gate If a coal spill occurs, a shutdown command is sent. This invention utilizes a pure algorithm, avoiding the time-consuming and labor-intensive cleanup caused by coal blockage, and providing a reliable technical guarantee.

Description

Technical Field

[0001] This invention relates to the field of detection technology, and in particular to a method for identifying coal spills based on information entropy. Background Technology

[0002] Coal-fired power plants use belt conveyors to transport coal, and these belt conveyors are connected by coal drop pipes. Figure 1 As shown. Following the coal flow direction, coal flows from the head of the upper conveyor belt into the coal chute and then out to the tail of the lower conveyor belt. During this process, coal blockage can occur in the coal chute for various reasons, meaning that coal flow becomes blocked at a certain point in the chute, resulting in coal flowing in but not out, eventually filling the chute and causing coal overflow. Because the conveyor belt has a large flow rate (usually greater than 600 tons / hour), if a blockage occurs but the conveyor belt is not stopped in time, a large amount of coal will overflow, requiring significant manpower and time for subsequent cleanup. The manpower and time required for subsequent cleanup depend on the amount of overflowing coal.

[0003] The prerequisite for promptly stopping the conveyor belt after a coal blockage occurs is the ability to detect the blockage in a timely manner. Existing coal blockage detection technologies use methods such as... Figure 2 The diagram shows a rotary paddle-type coal blockage switch. This switch is installed inside the coal chute. When there is no coal around the rotating blades, it indicates no blockage and no blockage signal is issued. When the rotating blades are submerged in coal, it means a blockage has occurred, the blades stop rotating, and a blockage signal is issued. When the coal quality is good, the rotary paddle-type coal blockage switch can effectively identify blockages. However, when the coal is sticky and wet, coal bridging blockages easily occur in the coal chute, causing the rotary paddle-type coal blockage switch to fail to identify blockages. If the rotary paddle-type coal blockage switch is installed at the bridging position, which is where the coal flow passes, it can easily lead to false alarms about blockages. In conclusion, the rotary paddle-type coal blockage switch cannot meet the requirements for coal blockage identification under sticky and wet conditions.

[0004] To promptly detect coal blockages caused by bridging, high demands are placed on operators manually monitoring the video feeds at the head of the conveyor belts. Since a coal-fired power plant typically has dozens of conveyor belts, each with video monitoring at its head, manually detecting blockages not only significantly increases the workload of operators but also increases the risk of missed or incorrect detections, severely reducing the automation level of the coal conveying system.

[0005] Therefore, those skilled in the art are dedicated to developing a new method for identifying coal spills that can automatically identify coal spill phenomena from existing surveillance video monitoring. Summary of the Invention

[0006] In view of the above-mentioned deficiencies of the prior art, the technical problem to be solved by the present invention is how to design a new coal spill identification method that can automatically identify coal spill phenomena from existing surveillance video monitoring.

[0007] To achieve the above objectives, the present invention provides a coal spill identification method based on information entropy, the method comprising the following steps:

[0008] Step 1: For each frame of the monitored area, convert it to a grayscale image and represent it using matrix A;

[0009] Step 2: Perform edge detection on matrix A, and denote the result as matrix C;

[0010] Step 3: Calculate the information entropy of the matrix C to obtain the frame entropy H of the grayscale image;

[0011] Step 4: Each frame of the real-time frame image generated sequentially in the monitored area is processed by steps 1 to 3 to obtain the frame entropy sequence H(1), H(2), H(3), ...;

[0012] Step 5: Calculate the mean μ and variance σ of the frame entropy sequence. 2 Determine the threshold H for frame entropy alarm. gate ;

[0013] Step 6: When the frame entropy of several consecutive frames exceeds the threshold H gate If this occurs, it is considered that coal overflow has occurred, and a shutdown command is sent.

[0014] Further, let the matrix A in step 1 be denoted as A0. m×n The subscript m×n indicates that matrix A has m rows and n columns;

[0015] The matrix A is shown in equation (1):

[0016]

[0017] Among them, a i,j Let a be an element in matrix A. i,j The value is an integer in the range of 0 to 255. The larger the value, the higher the brightness. i is an integer from 1 to m, and j is an integer from 1 to n.

[0018] Furthermore, the size of the monitoring area shall not exceed 200×200 and shall not be less than 50×50.

[0019] Furthermore, step 2 includes the following sub-steps:

[0020] Step 2.1: Perform equalization processing on matrix A to obtain matrix B;

[0021] Step 2.2: Use the Canny() method of the cv2 library in Python to perform edge detection on the matrix B. Preferably, the lower bound threshold of the Canny() method is 127, the upper bound threshold is 128, and the Sobel kernel size is 3. The result of the edge detection is denoted as matrix C.

[0022] In this matrix, the number of rows and columns of matrix A, matrix B, and matrix C are equal, and each element of matrix C is an integer in the range of 0 to 255.

[0023] Further, step 2.1 includes the following sub-steps:

[0024] Step 2.1.1 Find the maximum value of each element in matrix A in equation (1), denoted as P; find the minimum value of each element in matrix A in equation (1), denoted as Q; find the average value of each element in matrix A in equation (1), denoted as R, and let: S=max{PR,RQ};

[0025] Step 2.1.2: For any element a in matrix A i,j Perform the transformation shown in equation (2):

[0026]

[0027] Where [x] represents the integer part of x;

[0028] All b i,j The matrix B is composed as shown in equation (3):

[0029]

[0030] Each element of matrix B is an integer in the range of 0 to 255.

[0031] Furthermore, step 3 includes the following sub-steps:

[0032] Step 3.1: Count the number of times the values ​​0, 1, 2, 3, ..., 255 appear in the matrix C, and record them as: x(0), x(1), x(2), x(3), ..., x(255);

[0033] Step 3.2: Calculate the probability of the values ​​0, 1, 2, 3, ..., 255 appearing in the matrix C, and denote them as p(0), p(1), p(2), p(3), ..., p(255), respectively.

[0034]

[0035] Step 3.3: Calculate the information entropy of the frequency sequence p(0), p(1), p(2), (3), ..., p(255), denoted as the frame entropy H of the grayscale image:

[0036]

[0037] Wherein, when p(k)=0, log2p(k)=0 is taken; H is a non-negative real number.

[0038] Furthermore, step 5 includes the following sub-steps:

[0039] Step 5.1: Calculate the mean μ and variance σ of the frame entropy sequence. 2 ;

[0040] Step 5.2: Calculate the frame entropy probability distribution density function f(H) under normal conditions:

[0041]

[0042] Where H is the frame entropy of the grayscale image, μ is the average value of the frame entropy sequence, and σ is the average value of the frame entropy sequence. 2 The variance of the frame entropy sequence;

[0043] Step 5.3: Use the erf() function provided by the special library of Python's scipy library to calculate the probability that the absolute value of the random variable is not greater than z, as shown in equation (7):

[0044]

[0045] For equation (6), the mean is μ and the variance is σ. 2 The frame entropy H, which follows a normal distribution, is transformed as shown in equation (8):

[0046]

[0047] Then we have equation (9):

[0048]

[0049] The value of z is calculated in reverse based on the inverse function erfinv() of erf() provided by the special function of Python's scipy library and the pre-calculated frame misjudgment of no coal spillage.

[0050] Step 5.4: Substitute the calculated value of z back into equation (8) to obtain the threshold H for frame entropy alarm. gate :

[0051]

[0052] Furthermore, if the frame misjudgment of coal spill is no greater than 0.2069, then the frame misjudgment of no coal spill is 0.7931; the reverse calculation yields z = 0.8025; then equation (10) is specified as: H gate =μ±1.2621σ.

[0053] Furthermore, the monitoring area in step 1 is an automatically monitored area that is manually selected.

[0054] Furthermore, the "several frames" in step 6 specifically refers to 13 frames.

[0055] The coal spill identification method based on information entropy provided by this invention has at least the following technical effects:

[0056] 1. By automatically monitoring the location of coal blockage features in the video footage that is manually judged, coal blockage can be automatically identified, saving manpower and avoiding common errors in manual monitoring;

[0057] 2. By utilizing existing video surveillance footage to identify coal blockages, no new components need to be added on-site, saving costs;

[0058] 3. Only the mean and variance of the frame entropy need to be calculated from normal videos, without needing coal blockage videos that are difficult to find in practice;

[0059] 4. Select the coal blockage feature location in the video to reduce the number of pixels and computational load in the processed image, thereby improving the processing speed;

[0060] 5. The design incorporates false alarms with a controllable probability.

[0061] The following will further explain the concept, specific structure, and technical effects of the present invention in conjunction with the accompanying drawings, so as to fully understand the purpose, features, and effects of the present invention. Attached Figure Description

[0062] Figure 1 This is a schematic diagram showing the connection between coal conveyor belts via coal drop pipes;

[0063] Figure 2 This is a real photo of a rotary paddle coal blockage switch;

[0064] Figure 3 This is a flowchart illustrating an embodiment of the coal spill identification method based on information entropy of the present invention;

[0065] Figure 4 This is a partial frame entropy trend graph;

[0066] Figure 5 The threshold H for frame entropy alarm gate Schematic diagram;

[0067] Figure 6 These are actual photos taken from a video of the coal spill.

[0068] Figure 7 This is a frame entropy trend graph when a coal spill occurs. Detailed Implementation

[0069] The following description, with reference to the accompanying drawings, illustrates several preferred embodiments of the present invention to make its technical content clearer and easier to understand. The present invention can be embodied in many different forms, and the scope of protection of the present invention is not limited to the embodiments mentioned herein.

[0070] Since a coal-fired power plant typically has dozens of conveyor belts, each with video surveillance at its head, manually detecting coal spills not only greatly increases the workload of operators but also easily leads to missed or excessive spills, severely reducing the automation level of the coal conveying system. To automatically identify coal spills, this invention provides a coal spill identification method based on information entropy. Figure 3 A flowchart of an embodiment of the present invention is shown, the method comprising the following steps:

[0071] Step 1: For each frame of the monitored area, convert it to a grayscale image and represent it using matrix A;

[0072] Step 2: Perform edge detection on matrix A, and denote the result as matrix C;

[0073] Step 3: Calculate the information entropy of matrix C to obtain the frame entropy H of the grayscale image;

[0074] Step 4: Each frame of the real-time frame image generated sequentially in the monitoring area is processed by steps 1 to 3 to obtain the frame entropy sequence H(1), H(2), H(3), ..., which approximately follows a Gaussian distribution;

[0075] Step 5: Calculate the mean μ and variance σ of the frame entropy sequence. 2 Determine the threshold H for frame entropy alarm. gate ;

[0076] Step 6: When the frame entropy of several consecutive frames exceeds the threshold H gate If this occurs, it is considered that coal overflow has occurred, and a shutdown command is sent.

[0077] In step 1, the monitoring area is a manually selected automatic monitoring area.

[0078] Let matrix A from step 1 be denoted as A. m×n The subscript m×n indicates that matrix A has m rows and n columns, where m may not be equal to n;

[0079] The specific representation of matrix A is shown in equation (1):

[0080]

[0081] Among them, a i,j Let a be an element in matrix A. i,j The value is an integer in the range of 0 to 255. The larger the value, the higher the brightness. i is an integer from 1 to m, and j is an integer from 1 to n.

[0082] To ensure that the computer can smoothly process real-time video streams from multiple cameras (no less than 20) simultaneously (25 frames per second per camera), the size of the rectangular frame of the monitoring area shall not exceed 200×200 (resolution) and shall not be less than 50×50; preferably, m×n shall be 80×80.

[0083] Step 2 includes the following sub-steps:

[0084] Step 2.1: Perform equalization on matrix A to obtain matrix B;

[0085] Step 2.2: Use the Canny() method of the cv2 library in Python to perform edge detection on matrix B. In this embodiment, the lower bound threshold of the Canny() method is set to 127, the upper bound threshold is set to 128, and the size of the Sobel kernel is set to 3. The result of edge detection is denoted as matrix C. The reason for edge detection is that under the condition of overflow coal, a large number of coal blocks in the selected area will have a large number of edges detected, while under normal conditions there is no overflow coal and therefore the edges are basically not detected.

[0086] In this matrix, the number of rows and columns of matrices A, B, and C are equal, and each element of matrix C is an integer in the range of 0 to 255.

[0087] Step 2.1 includes the following sub-steps:

[0088] Step 2.1.1 Find the maximum value of each element in matrix A in equation (1), denoted as P; find the minimum value of each element in matrix A in equation (1), denoted as Q; find the average value of each element in matrix A in equation (1), denoted as R, and let: S=max{PR,RQ};

[0089] Step 2.1.2: For any element a in matrix A i,j Perform the transformation shown in equation (2):

[0090]

[0091] Where [x] represents the integer part of x;

[0092] Remember all the b i,j The resulting matrix is ​​B, as shown in equation (3):

[0093]

[0094] Matrix B has the following characteristics: Even under varying ambient lighting conditions, the average pixel value of each frame in the video stream obtained from the same camera after the above transformation is essentially the same, meaning the brightness remains basically unchanged. Each element of Matrix B is an integer in the range of 0 to 255, satisfying the requirements of the corresponding image file format. Matrix B corresponds to the highest contrast, thus enhancing the image and facilitating signal detection.

[0095] After equalization, the mean of each element in matrix A is 127.5, and at least one of the following two results is true: the maximum value of an element in A is 255, and the minimum value of an element in A is 0. The purpose of equalization is to ensure that grayscale images in each frame of a video have the same brightness under different lighting conditions, and that the contrast between pixels with different brightness levels is as high as possible.

[0096] Step 3 includes the following sub-steps:

[0097] Step 3.1: Count the number of times the values ​​0, 1, 2, 3, ..., 255 appear in matrix C, and record them as: x(0), x(1), x(2), x(3), ..., x(255);

[0098] Step 3.2: Calculate the probabilities of the values ​​0, 1, 2, 3, ..., 255 appearing in matrix C, and denote them as p(0), p(1), p(2), p(3), ..., p(255), respectively.

[0099]

[0100] Step 3.3: Calculate the information entropy of the frequency sequence p(0), p(1), p(2), (3), ..., p(255), denoted as the frame entropy H of the grayscale image:

[0101]

[0102] Wherein, when p(k)=0, log2p(k)=0 is taken; H is a non-negative real number.

[0103] For the same camera operating normally (without coal blockage), there are at least 1 million frames of video. The more frames, the higher the accuracy. Calculate the frame entropy H of each frame.

[0104] Step 5 includes the following sub-steps:

[0105] Step 5.1: Calculate the mean μ and variance σ of the frame entropy sequence. 2 ;

[0106] Step 5.2: Calculate the frame entropy probability distribution density function f(H) under normal conditions:

[0107]

[0108] Where H is the frame entropy of the grayscale image, μ is the average value of the frame entropy sequence, and σ is the mean value of the frame entropy sequence. 2 The variance of the frame entropy sequence;

[0109] In practical engineering applications, when video footage detects coal spillage at the conveyor head due to blockage, the conveyor should be stopped as quickly as possible. The corresponding signal trip circuit takes approximately one second from receiving the signal to activation. Practice shows that spillage within one second results in a small, acceptable amount of spillage. Therefore, the video assessment time for detecting spillage is 0.5 seconds, with 0.5 seconds remaining. Typically, a video stream has 25 frames per second; any 13 consecutive frames (0.5 seconds' worth of frames) are considered spillage and used as the trigger for the conveyor trip signal. Equipment and operators can accept one false alarm of coal blockage per conveyor per year. Assuming a year has 365 days, the number of frames required for annual monitoring is:

[0110] 365 × 24 × 3600 × 25 = 788400000

[0111] In these 788,400,000 frames, let the probability of a false alarm be p. If any 13 consecutive frames indicate a coal spill and trigger a conveyor belt shutdown signal, find the range of values ​​for p that ensure the expected number of false conveyor belt shutdowns per year is no greater than 1.

[0112] Let E be the expected number of times a erroneous belt conveyor stop signal is sent. Then we have:

[0113] E = (788400000 - 13)p 13 ≈788400000p 13 ≤1

[0114] Solving for:

[0115] That is, the probability of a frame misjudgment of coal spillage should not be greater than 0.2069.

[0116] Equation (6) cannot be used to find the antiderivative by integration; it can only be used to find an approximate solution of the definite integral by Taylor expansion.

[0117] Step 5.3: Use the erf() function provided by the special library of Python's scipy library to calculate the probability that the absolute value of the random variable is not greater than z; the function of erf() is to find the probability distribution of the absolute value of the random variable not greater than z under the standard normal distribution, as shown in equation (7):

[0118]

[0119] The mathematical meaning of equation (7) is the probability that the absolute value of a normally distributed random variable with a mean of 0 and a variance of 1 is not greater than z.

[0120] For equation (6), the mean is μ and the variance is σ. 2 The frame entropy H, which follows a normal distribution, is transformed as shown in equation (8):

[0121]

[0122] Then we have equation (9):

[0123]

[0124] Based on the inverse function erfinv() of erf() provided by the special function of Python's scipy library and the pre-calculated frame misjudgment of no coal spillage, the value of z is calculated in reverse, that is, z is calculated when the cumulative distribution probability is known.

[0125] Step 5.4: Substitute the calculated value of z back into equation (8) to obtain the frame entropy alarm threshold H. gate :

[0126]

[0127] Since the probability of a frame misjudging coal spillage is no greater than 0.2069, the probability of a frame misjudging no coal spillage under the condition of no coal spillage is 0.7931; the reverse calculation yields z = 0.8025; then equation (10) is specified as: H gate =μ±1.2621σ.

[0128] For a normally operating belt conveyor monitoring video, the mean μ and variance σ of the frame entropy H under normal conditions are obtained. 2 Then, the threshold H for judging coal spill alarm is determined according to formula (10). gate When the frame entropy in online monitoring is not the H determined by equation (10) gate When the frame is within the upper or lower limit range, it is determined to be a coal blockage. Using any 13 consecutive frames as the condition for stopping the conveyor belt, it can be ensured that the number of times the conveyor belt is mistakenly stopped in continuous operation in a year does not exceed 1.

[0129] The normal video and the coal blockage video are almost identical except for the areas where coal spills or leaks. Therefore, only the local image of the area where the coal spill or leak occurs needs to be processed.

[0130] After a color image is converted to a grayscale image, the computer representation of the image is a two-dimensional numerical matrix. Each element of the matrix represents the grayscale value of the corresponding pixel, with 0 representing pure white and 255 representing pure black.

[0131] Image equalization aims to unify different lighting conditions at a scene to the same brightness and maximize contrast.

[0132] Considering that both the conveyor belt and the coal are black, the purpose of performing a 2D convolution on the grayscale image is to sharpen the edges, thereby distinguishing the coal from the background. After multiple tests, the Canny() method from Python's cv2 library was found to be the best method for edge detection on the equalized image. The optimal results were achieved with the following three parameters: a lower threshold of 127, an upper threshold of 128, and a Sobel kernel size of 3.

[0133] Calculate the frame entropy of the convolved image, accumulate a large number (no less than 1 million frames) of normally running video, and obtain the corresponding frame entropy mean and variance. Use the mathematical theory of normal distribution and the corresponding Python library functions to determine the frame entropy alarm threshold.

[0134] The frame entropy of any 13 consecutive frames exceeds the threshold H gate If this happens, the belt conveyor trip command will be triggered, thus satisfying the requirement of a small amount of coal spillage on site.

[0135] The following is an application example of a coal spill identification method based on information entropy provided by the present invention in a power plant.

[0136] The 2B conveyor belt of a certain power plant frequently experiences coal blockage. The existing rotary paddle coal blockage switch cannot reliably identify it. The manual monitoring method not only greatly increases the workload of the operators, but also easily leads to errors and omissions, resulting in a large amount of coal spillage and time-consuming and laborious cleanup.

[0137] Under normal operating conditions, after setting the automatic video monitoring area, the resulting frame entropy trend graph (partial) is as follows: Figure 4 As shown.

[0138] The mean μ = 2.6504297 is obtained by calling the mean() function of the NumPy library in Python, and the variance σ is obtained by calling the var() function of the NumPy library. 2 =0.024911199, and therefore σ = 0.1579. From equation (10), we obtain:

[0139] H gate =2.6504297±0.1992

[0140] The corresponding thresholds are as follows Figure 5 As shown.

[0141] Monitoring began at 10:30 AM on September 3, 2022, and ended at 3:30 PM on September 13, 2022, totaling 22,050,000 frames of video, of which:

[0142] At 19:57:32 on September 8th, the video system issued a simulated belt conveyor trip command (the belt conveyor did not actually trip). The corresponding surveillance footage was retrieved, and the image appeared as follows. Figure 6 As shown.

[0143] The corresponding background data was retrieved, and the resulting frame entropy trend chart is as follows: Figure 7 As shown.

[0144] At other times, the video system did not issue any tripping commands, and the conveyor belt did not actually experience coal blockage.

[0145] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A coal spill identification method based on information entropy, characterized in that, The method comprises the following steps: Step 1, for each frame image of the monitoring area, convert it into a gray image and represent it with a matrix A; Step 2, perform edge detection on the matrix A, and the result is recorded as a matrix C; Step 3, calculate the information entropy of the matrix C to obtain the frame entropy H of the gray image; Step 4, each frame of the real-time frame images successively generated by the monitoring area is processed through steps 1 to 3 to obtain a frame entropy sequence H(1), H(2), H(3), …; Step 5, calculate the frame entropy mean μ and variance σ of the frame entropy sequence 2 According to the inverse function erfinv() of erf() provided by special of scipy library of Python and the frame false alarm coal probability obtained by pre-computation, the value of z is reversely calculated to determine the threshold H of frame entropy alarm gate : Step 6, when the frame entropy of consecutive frames exceeds the threshold H gate a stop command is sent.

2. The information entropy-based coal spill identification method of claim 1, wherein, Let the matrix A in step 1 be denoted as A m×n , where the subscript m x n indicates that the number of rows of the matrix A is m and the number of columns is n; The matrix A is shown in formula (1): wherein a i,j is an element in the matrix A, a i,j is an integer in the range of 0 to 255, and the larger the value, the higher the brightness, i is an integer from 1 to m, and j is an integer from 1 to n.

3. The information entropy-based coal spill identification method of claim 2, wherein, The size of the monitoring area is not more than 200x200, and not less than 50x50.

4. The information entropy-based coal spill identification method of claim 2, wherein, The step 2 comprises the following sub-steps: Step 2.1, perform equalization processing on the matrix A to obtain a matrix B; Step 2.2, use the Canny() method of the cv2 library in Python to perform the edge detection on the matrix B; the result of the edge detection is recorded as the matrix C; Wherein, the number of rows and the number of columns of the matrix A, the matrix B and the matrix C are equal, and each element of the matrix C is also an integer in the range of 0 to 255.

5. The information entropy-based coal spill identification method of claim 4, wherein, The step 2.1 comprises the following sub-steps: Step 2.1.1, find the maximum value of each element in the matrix A in formula (1), recorded as P; find the minimum value of each element in the matrix A in formula (1), recorded as Q; find the average value of each element in the matrix A in formula (1), recorded as R, and let S = max{P-R, R-Q}; Step 2.1.

2. For any one element a in the matrix A i,j a transformation as shown in equation (2): Wherein, [x] represents the integer part of x; All of b i,j The matrix B is composed as shown in equation (3): Each element of the matrix B is an integer in the range of 0 to 255.

6. The information entropy-based coal spill identification method of claim 4, wherein, The step 3 comprises the following sub-steps: Step 3.1, count the number of times that the values 0, 1, 2, 3, …, 255 appear in the matrix C, and record them in sequence as x(0), x(1), x(2), x(3), …, x(255); Step 3.2, calculate the probabilities of the values 0, 1, 2, 3, …, 255 appearing in the matrix C, and record them in sequence as p(0), p(1), p(2), (3), …, p(255), that is: Step 3.3, calculate the information entropy of the frequency sequence p(0), p(1), p(2), (3), …, p(255), recorded as the frame entropy H of the gray image: Wherein, when p(k) = 0, take log2p(k) = 0; H is a non-negative real number.

7. The information entropy-based coal spill identification method of claim 6, wherein, The step 5 comprises the following sub-steps: Step 5.1, compute the frame entropy mean μ and variance σ of the frame entropy sequence 2 ; Step 5.2, calculate the frame entropy probability distribution density function f(H) under the normal state: where H is the frame entropy of the grayscale image, μ is the mean value of the frame entropy sequence, σ 2 is the variance of the frame entropy sequence; Step 5.3, use the erf() function provided by the special of Python's scipy library to calculate the probability that the absolute value of a random variable is not greater than z, as shown in formula (7): For the frame entropy H described by equation (6) with mean μ and variance σ 2 , which is normally distributed, the transformation is as equation (8): Then formula (9) is obtained: According to the inverse function erfinv() of the erf() provided by the special of Python's scipy library and the pre-calculated frame misjudgment overflow coal probability, the value of z is calculated reversely; Step 5.

4. Substitute the calculated value of z back into equation (8) to obtain the threshold H for the frame entropy alarm gate : 8.The information entropy-based coal spill identification method of claim 7, wherein, If the frame misjudgment overflow coal probability is not greater than 0.2069, then the frame misjudgment no overflow coal probability is 0.7931; reversely calculating z=0.8025; then formula (10) is specified as: H gate = μ ± 1.2621σ. 9.The information entropy-based coal spill identification method of claim 1, wherein, The monitoring area in the step 1 is an automatic monitoring area manually framed. 10.The information entropy-based coal spill identification method of claim 1, wherein, The number of frames in the step 6 is 13.

Images