Wind turbine generator cabin fire detection method and system based on image processing

Abstract

The invention relates to the technical field of fire detection, in particular to a wind turbine generator cabin fire detection method and system based on image processing, and the method comprises the following steps: extracting a motion region from a current cabin video stream image, and obtaining a suspected flame block and a suspected smoke block; performing feature extraction on the two blocks to obtain first feature information and second feature information; performing true and false judgment based on the first feature information and the second feature information, when two or more features in the first feature information are judged to be true, judging that the suspected flame block is true, and when any one of the first feature information and the second feature information is judged to be true, judging that the fire event is true; and performing fire event detection output according to a true and false judgment result. According to the invention, through video detection and motion area extraction, feature extraction is carried out on the motion area, and flame and smoke image features are integrated to carry out true and false determination of the fire event, the misjudgment rate can be effectively reduced, and higher fire identification accuracy is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of fire detection, and more particularly, to a method and system for detecting a fire in a nacelle of a wind turbine based on image processing. Background Art

[0002] A wind power generating unit is the core equipment of wind power generation. Due to the high concentration of technical equipment and the use of flammable materials in the nacelle, there are certain fire hazards in lubrication and heat dissipation systems, gearboxes, braking systems, nacelle bases, transmission cables, control cabinets, etc. Once a fire occurs, it will spread rapidly. When the nacelle is completely damaged, the repair cost may reach the value of the wind power generating unit itself.

[0003] In order to prevent and reduce fire hazards, the demand for intelligent fire detection systems is becoming more urgent. At present, relatively mature fire detection systems include temperature sensing, light sensing, smoke sensing, etc. These sensor-based detection technologies are widely used in real life, but they must be installed near the fire starting point and the detection data is single. Due to the diversity of fires, the data collected by these traditional fire detection technologies is inevitably affected by the surrounding environment. Therefore, the application of this type of detector is greatly limited. In this context, vision-based fire detection technology has received more and more attention because video intelligent monitoring technology is applicable to a wider range of places and can provide more reliable data information.

[0004] At present, vision-based fire detection technologies are divided into two types: flame detection and smoke detection; by analyzing information such as the color, movement, and contour of the flame, it is then determined whether a fire has occurred.

[0005] To design an efficient fire detection algorithm, it is first necessary to consider establishing an effective method for detecting fire image features, where flame and smoke are the core of fire detection. After the flame is generated, its temperature and brightness increase, the area gradually expands, and it is accompanied by continuous and irregular pulsations, forming obvious flame image features; while smoke, as one of the important features in the initial stage of a fire, is formed by the incomplete combustion of objects during a fire, and its color mainly shows from grayish white to black, and the area continuously spreads. Existing technologies often only focus on one of the flame or smoke, and it is crucial to integrate the two types of information to obtain reliable flame and smoke image features for improving the accuracy of image-based fire identification.

[0006] Based on this, there is an urgent need for a method for detecting a fire in a nacelle of a wind turbine that integrates flame and smoke image features to achieve a higher fire identification accuracy. Summary of the Invention

[0007] The object of the present invention is to provide a method and system for detecting the fire in the nacelle of a wind turbine based on image processing, which integrates the image features of flames and smoke to achieve a higher accuracy rate of fire identification.

[0008] The present invention is realized through the following technical solutions: A method for detecting the fire in the nacelle of a wind turbine based on image processing, comprising the following steps:

[0009] Based on the current nacelle video stream image, extract the moving area from the video stream image to obtain a suspected flame block and a suspected smoke block;

[0010] Perform feature extraction on the suspected flame block and the suspected smoke block respectively to obtain first feature information and second feature information. The first feature information includes the flame contour feature obtained based on edge detection, the flame movement feature obtained based on the difference in flame area between adjacent frames, and the flame diffusion feature obtained based on the distance between the centroids of flames in adjacent frames. The second feature information includes the smoke diffusion feature obtained based on the distance between the centroids of smoke in adjacent frames;

[0011] Perform true or false determination based on the first feature information and the second feature information. Among them, when two or more features in the first feature information are determined to be true, the suspected flame block is determined to be true. When any one of the first feature information and the second feature information is determined to be true, the fire event is determined to be true;

[0012] Perform fire event detection output according to the true or false determination result.

[0013] According to a preferred implementation manner, the specific process of extracting the suspected smoke block from the video stream image is as follows;

[0014] Obtain the transmittance of the current frame and the previous frame, and calculate the dark channel images of the current frame and the previous frame based on the transmittance and the pre-acquired atmospheric light component. The expression is as follows:

[0015] DARK = (1 - t(x))·A c / ω

[0016] In the above formula, DARK represents the dark channel image, t(x) represents the transmittance, A represents the atmospheric light component, the subscript c represents three channels, and ω represents the adjustment factor;

[0017] Perform differential operation on the dark channel images of the current frame and the previous frame to obtain the suspected smoke block.

[0018] According to a preferred implementation manner, the specific process of extracting the suspected flame block from the video stream image is as follows;

[0019] Convert the pixels of the video stream image from the RGBA color space to the YCbCR color space, and extract features from the video stream image in the YCbCR color space based on a preset flame extraction rule to obtain a suspected flame block. The expression of the preset flame extraction rule is as follows:

[0020]

[0021] In the above formula, Y(x,y) represents the pixel value in the Y channel of the YCbCR color space, Cb(x,y) represents the pixel value in the Cb channel of the YCbCR color space, Cr(x,y) represents the pixel value in the Cr channel of the YCbCR color space, F(x,y) represents the pixel value of the video stream image, and Y mean represents the average value of brightness, Cb mean represents the average value of blue chrominance, Cr mean represents the average value of red chrominance, M represents the width of the video stream image, and N represents the height of the video stream image.

[0022] According to a preferred embodiment, the flame contour feature obtained based on edge detection specifically includes:

[0023] Taking the binary image of the suspected flame block as the input, obtaining the area of the suspected flame block and using the Sobel edge detection operator to detect the input binary image to obtain the perimeter of the suspected flame block. Based on the area and perimeter of the suspected flame block, obtain the flame contour feature, and the expression is as follows:

[0024]

[0025] In the above formula, C k represents the circularity of the suspected flame block, k represents the number of connected regions in the suspected flame block, S k represents the area of the suspected flame block, and L k represents the perimeter of the suspected flame block.

[0026] According to a preferred embodiment, the specific process of the flame movement feature obtained based on the difference in flame area between adjacent frames is as follows;

[0027] Obtain the centroid coordinates C1 of the suspected flame block in the current frame and the centroid coordinates C2 of the suspected flame block in the previous frame. Calculate the centroid distance D1 of the suspected flame block in adjacent frames based on the centroid coordinates C1 and C2, and determine whether D1 satisfies the movement determination condition. When the movement determination condition is satisfied, obtain the flame movement feature, and the movement determination condition is that D1 is within a preset movement range.

[0028] According to a preferred embodiment, the specific process of the flame diffusion feature obtained based on the distance between the centroids of adjacent frames of flames is as follows;

[0029] Obtain the area S1 of the suspected flame block in the current frame and the area S2 of the suspected flame block in the previous frame, determine whether S1 is greater than S2, and determine whether the area change rate SR of the suspected flame block in the current frame is greater than a preset diffusion threshold. When the diffusion determination condition is satisfied, obtain the flame diffusion feature, where the diffusion determination condition is that S1 is greater than S2 and SR is greater than the preset diffusion threshold.

[0030] The present invention also provides a wind turbine nacelle fire detection system based on image processing, including:

[0031] A motion area extraction module, configured to extract a motion area from the current nacelle video stream image to obtain a suspected flame block and a suspected smoke block;

[0032] A feature extraction module, configured to perform feature extraction on the suspected flame block and the suspected smoke block respectively to obtain first feature information and second feature information. The first feature information includes a flame contour feature obtained based on edge detection, a flame movement feature obtained based on the difference in flame area between adjacent frames, and a flame diffusion feature obtained based on the distance between the centroids of adjacent frames of flames. The second feature information includes a smoke diffusion feature obtained based on the distance between the centroids of adjacent frames of smoke;

[0033] A true / false determination module, configured to perform true / false determination based on the first feature information and the second feature information. Among them, when two or more features in the first feature information are determined to be true, the suspected flame block is determined to be true. When any of the first feature information and the second feature information is determined to be true, the fire event is determined to be true;

[0034] An output module, configured to perform fire event detection output according to the true / false determination result.

[0035] The technical solution of the wind turbine nacelle fire detection method and system based on image processing provided by the present invention has at least the following advantages and beneficial effects: The present invention detects and extracts the motion area through video, further extracts features from the motion area, and integrates the flame and smoke image features to perform true / false determination of the fire event, which can effectively reduce the false positive rate and achieve a higher fire identification accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 It is a schematic flowchart of the wind turbine nacelle fire detection method based on image processing provided in Embodiment 1 of the present invention. DETAILED DESCRIPTION

[0037] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Usually, the components of the embodiments of the present invention described and illustrated in the accompanying drawings here can be arranged and designed in various different configurations.

[0038] Embodiment 1

[0039] Figure 1 It is a schematic flowchart of the method for detecting wind turbine nacelle fires based on image processing provided by the embodiments of the present invention. Refer to Figure 1 As shown, the implementation process of the method for detecting wind turbine nacelle fires based on image processing includes the following detailed steps:

[0040] Step 1: Detect the video and extract the moving region.

[0041] In the embodiments of the present invention, based on the current nacelle video stream image, the moving region is extracted from the video stream image to obtain suspected flame blocks and suspected smoke blocks.

[0042] Step 1.1: Extract the suspected smoke blocks. The specific process is as follows;

[0043] Obtain the transmittance of the current frame and the previous frame, and calculate the dark channel images of the current frame and the previous frame based on the transmittance and the pre-obtained atmospheric light components. The expression is as follows:

[0044] DARK = (1 - t(x))·A c / ω

[0045] In the above formula, DARK represents the dark channel image, t(x) represents the transmittance, A represents the atmospheric light components, the subscript c represents three channels, and ω represents the adjustment factor;

[0046] Perform a differential operation on the dark channel images of the current frame and the previous frame to obtain the suspected smoke blocks.

[0047] Step 1.2: Extract the suspected flame blocks. The specific process is as follows;

[0048] Convert the pixels of the video stream image from the RGBA color space to the YCbCr color space, and perform feature extraction on the video stream image in the YCbCr color space based on the preset flame extraction rules to obtain the suspected flame blocks. Among them, the expression of the preset flame extraction rules is as follows:

[0049]

[0050] In the above formula, Y(x, y) represents the pixel value in the Y channel of the YCbCR color space, Cb(x, y) represents the pixel value in the Cb channel of the YCbCR color space, Cr(x, y) represents the pixel value in the Cr channel of the YCbCR color space, and F(x, y) represents the pixel value of the video stream image, where Y mean represents the average value of luminance, Cb mean represents the average value of blue chrominance, Cr mean represents the average value of red chrominance, M represents the width of the video stream image, and N represents the height of the video stream image.

[0051] It should be noted that in the YCbCR color space, the extracted suspected flame blocks are more complete and accurate compared to those in the RGBA color space, and the recognizable color range is wider, so as to avoid the interference of other objects in the environment of the wind turbine nacelle and reduce the internal holes in the blocks. Therefore, it is applicable to the extraction of suspected flame blocks in complex environments such as the wind turbine nacelle.

[0052] Step 2: Perform feature extraction.

[0053] In the embodiment of the present invention, feature extraction is respectively performed on the suspected flame blocks and the suspected smoke blocks to obtain first feature information and second feature information. The first feature information includes a flame contour feature obtained based on edge detection, a flame movement feature obtained based on the difference in flame area between adjacent frames, and a flame diffusion feature obtained based on the distance between the centroids of flames in adjacent frames. The second feature information includes a smoke diffusion feature obtained based on the distance between the centroids of smoke in adjacent frames.

[0054] Step 2.1: The flame contour feature obtained based on edge detection specifically includes:

[0055] Taking the binary image of the suspected flame block as the input, obtaining the area of the suspected flame block and using the Sobel edge detection operator to detect the input binary image to obtain the perimeter of the suspected flame block. Based on the area and perimeter of the suspected flame block, the flame contour feature is obtained, and the expression is as follows:

[0056]

[0057] In the above formula, C k represents the circularity of the suspected flame block, k represents the number of connected regions in the suspected flame block, S k represents the area of the suspected flame block, and L k represents the perimeter of the suspected flame block.

[0058] Step 2.2: The flame movement feature obtained based on the difference in flame area between adjacent frames, the specific process is as follows;

[0059] Obtain the centroid coordinates C1 of the suspected flame block in the current frame and the centroid coordinates C2 of the suspected flame block in the previous frame. Calculate the centroid distance D1 of the suspected flame block in adjacent frames based on the centroid coordinates C1 and C2, and determine whether D1 meets the movement determination condition. When the movement determination condition is met, obtain the flame movement feature, where the movement determination condition is that D1 is within a preset movement range.

[0060] Step 2.3, the flame diffusion feature obtained based on the centroid distance of adjacent frame flames, the specific process is as follows;

[0061] Obtain the area S1 of the suspected flame block in the current frame and the area S2 of the suspected flame block in the previous frame. Determine whether S1 is greater than S2 and whether the area change rate SR of the suspected flame block in the current frame is greater than a preset diffusion threshold. When the diffusion determination condition is met, obtain the flame diffusion feature, where the diffusion determination condition is that S1 is greater than S2 and SR is greater than the preset diffusion threshold.

[0062] Step 3, perform true or false determination of the feature.

[0063] In the embodiment of the present invention, true or false determination is performed based on the first feature information and the second feature information. Among them, when two or more features in the first feature information are determined to be true, the suspected flame block is determined to be true. When any of the first feature information and the second feature information is determined to be true, the fire event is determined to be true.

[0064] Step 4, result output.

[0065] In this embodiment, fire event detection output is performed according to the true or false determination result. For example, when the flame diffusion feature and the flame movement feature in the first feature information are determined to be true, the fire event is determined to be true. At this time, the output result is that there is a fire event. Compared with relying on a single flame or smoke feature, integrating flame and smoke features can significantly reduce the misjudgment rate.

[0066] In summary, the technical solution of the wind turbine nacelle fire detection method and system based on image processing provided by the present invention at least has the following advantages and beneficial effects: The present invention detects and extracts the moving area through video, further extracts features from the moving area, and integrates the flame and smoke image features to perform true or false determination of the fire event, which can effectively reduce the misjudgment rate and achieve a higher fire identification accuracy.

[0067] Embodiment 2

[0068] Based on the technical solution provided in Embodiment 1, this embodiment provides a wind turbine nacelle fire detection system based on image processing. The system includes a moving area extraction module, a feature extraction module, a true or false determination module, and an output module;

[0069] Among them, the motion area extraction module is used to extract the motion area from the current cabin video stream image to obtain a suspected flame block and a suspected smoke block; the feature extraction module is used to perform feature extraction on the suspected flame block and the suspected smoke block respectively to obtain first feature information and second feature information. The first feature information includes a flame contour feature obtained based on edge detection, a flame movement feature obtained based on the difference in flame area between adjacent frames, and a flame diffusion feature obtained based on the distance between the centroids of flames in adjacent frames. The second feature information includes a smoke diffusion feature obtained based on the distance between the centroids of smoke in adjacent frames; the true / false determination module is used to perform true / false determination based on the first feature information and the second feature information. Among them, when two or more features in the first feature information are determined to be true, the suspected flame block is determined to be true. When any one of the first feature information and the second feature information is determined to be true, the fire event is determined to be true; the output module is used to perform fire event detection output according to the true / false determination result.

[0070] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for detecting the fire in the nacelle of a wind turbine based on image processing, characterized in that, The steps include the following: Based on the current cabin video stream image, extract the moving regions from the video stream image to obtain suspected flame blocks and suspected smoke blocks; Perform feature extraction on the suspected flame blocks and suspected smoke blocks respectively to obtain first feature information and second feature information. The first feature information includes the flame contour feature obtained based on edge detection, the flame movement feature obtained based on the difference in flame area between adjacent frames, and the flame diffusion feature obtained based on the distance between the centroids of flames in adjacent frames. The second feature information includes the smoke diffusion feature obtained based on the distance between the centroids of smoke in adjacent frames; Perform true / false determination based on the first feature information and the second feature information. Among them, when two or more features in the first feature information are determined to be true, the suspected flame block is determined to be true. When any of the first feature information and the second feature information is determined to be true, the fire event is determined to be true; Perform fire event detection output according to the true / false determination result.

2. The method for detecting the fire in the nacelle of a wind turbine based on image processing according to claim 1, characterized in that, The specific process of extracting the suspected smoke block from the video stream image is as follows; Obtain the transmittance of the current frame and the previous frame, and calculate the dark channel images of the current frame and the previous frame based on the transmittance and the pre-acquired atmospheric light component. The expression is as follows: DARK = (1 - t(x))·A c / ω In the above formula, DARK represents the dark channel image, t(x) represents the transmittance, A represents the atmospheric light component, the subscript c represents three channels, and ω represents the adjustment factor; Perform a difference operation on the dark channel images of the current frame and the previous frame to obtain the suspected smoke block.

3. The method for detecting the fire in the nacelle of a wind turbine based on image processing according to any one of claims 1 to 2, characterized in that, The specific process of extracting the suspected flame block from the video stream image is as follows; Convert the pixels of the video stream image from the RGBA color space to the YCbCR color space, and perform feature extraction on the video stream image in the YCbCR color space based on a preset flame extraction rule to obtain the suspected flame block. Among them, the expression of the preset flame extraction rule is as follows: In the above formula, Y(x, y) represents the pixel value in the Y channel of the YCbCR color space, Cb(x, y) represents the pixel value in the Cb channel of the YCbCR color space, Cr(x, y) represents the pixel value in the Cr channel of the YCbCR color space, F(x, y) represents the pixel value of the video stream image, and Y mean represents the average value of brightness, Cb mean represents the average value of blue chrominance, Cr mean represents the average value of red chrominance, M represents the width of the video stream image, and N represents the height of the video stream image.

4. The method for detecting the fire in the nacelle of a wind turbine based on image processing according to claim 3, characterized in that The flame contour feature obtained based on edge detection specifically includes: Using the binary image of the suspected flame block as the input, obtain the area of the suspected flame block, and use the Sobel edge detection operator to detect the input binary image to obtain the perimeter of the suspected flame block. Based on the area and perimeter of the suspected flame block, obtain the flame contour feature. The expression is as follows: In the above formula, C k represents the circularity of the suspected flame block, k represents the number of connected regions in the suspected flame block, S k represents the area of the suspected flame block, and L k represents the perimeter of the suspected flame block.

5. The method for detecting the fire in the nacelle of a wind turbine based on image processing according to claim 3, wherein, The specific process of the flame movement feature obtained based on the difference in flame area between adjacent frames is as follows; Obtain the centroid coordinates C1 of the suspected flame block in the current frame and the centroid coordinates C2 of the suspected flame block in the previous frame. Calculate the centroid distance D1 of the suspected flame block in adjacent frames based on the centroid coordinates C1 and C2, and determine whether D1 meets the movement determination condition. When the movement determination condition is met, obtain the flame movement feature. The movement determination condition is that D1 is within a preset movement range.

6. The method for detecting the fire in the nacelle of a wind turbine based on image processing according to claim 3, wherein The specific process of the flame diffusion feature obtained based on the distance between the centroids of flames in adjacent frames is as follows; Obtain the area S1 of the suspected flame block in the current frame and the area S2 of the suspected flame block in the previous frame, determine whether S1 is greater than S2 and determine whether the area change rate SR of the suspected flame block in the current frame is greater than a preset diffusion threshold. When the diffusion determination condition is satisfied, obtain the flame diffusion feature, and the diffusion determination condition is that S1 is greater than S2 and SR is greater than the preset diffusion threshold.

7. A wind turbine nacelle fire detection system based on image processing, characterized in that, Including: A motion area extraction module, configured to extract a motion area from the current cabin video stream image based on the video stream image to obtain a suspected flame block and a suspected smoke block; A feature extraction module, configured to perform feature extraction on the suspected flame block and the suspected smoke block respectively to obtain first feature information and second feature information. The first feature information includes a flame contour feature obtained based on edge detection, a flame movement feature obtained based on the difference in flame area between adjacent frames, and a flame diffusion feature obtained based on the distance between the centroids of flames in adjacent frames. The second feature information includes a smoke diffusion feature obtained based on the distance between the centroids of smoke in adjacent frames; A true / false determination module, configured to perform true / false determination based on the first feature information and the second feature information. Among them, when two or more features in the first feature information are determined to be true, the suspected flame block is determined to be true. When any of the first feature information and the second feature information is determined to be true, the fire event is determined to be true; An output module, configured to perform fire event detection output according to the true / false determination result.

Images