Forest fire prevention detection method based on artificial intelligence

Abstract

The invention discloses a forest fire prevention detection method based on artificial intelligence, and relates to the technical field of forest fire prevention, and the method comprises the following steps: S1, carrying out the pixel point set extraction of an initial detection image of a forest region, and obtaining a head pixel point set and an accompanying pixel point set; s2, obtaining a stable suppression coefficient of the initial detection image; s3, constructing a global suppression processing model for the initial detection image; and S4, processing the initial detection image by using the global suppression processing model according to the stability suppression coefficient of the initial detection image to obtain a forest fire prevention detection result. According to the invention, whether a fire area exists in the initial detection image can be determined, and timely and accurate information support is provided for fire early warning and fire fighting.

Description

Technical Field

[0001] The present invention relates to the technical field of forest fire prevention, and in particular to a forest fire prevention detection method based on artificial intelligence. Background Art

[0002] With global climate change and the increase of human activities, the frequency and severity of forest fires are increasing year by year. Traditional forest fire detection methods mainly rely on manual patrols and ground monitoring stations, which have disadvantages such as limited monitoring range and slow response speed. Therefore, it is particularly important to develop an intelligent method that can monitor forest fires in real time. Summary of the invention

[0003] In order to solve the above problems, the present invention proposes a forest fire detection method based on artificial intelligence.

[0004] The technical solution of the present invention is: a forest fire detection method based on artificial intelligence comprises the following steps:

[0005] S1. Extract pixel point sets from the initial detection image of the forest area to obtain head pixel points and companion pixel point sets respectively;

[0006] S2, obtaining a stable suppression coefficient of the initial detection image through the pixel value of the head pixel and the pixel value of the pixel in the companion pixel set;

[0007] S3, constructing a global inhibition processing model for the initial detection image;

[0008] S4. According to the stable suppression coefficient of the initial detection image, the initial detection image is processed using the global suppression processing model to obtain the forest fire detection result.

[0009] Furthermore, S1 includes the following sub-steps:

[0010] S11, collecting an initial detection image of the forest area through an image sensor, and determining the head pixel points of the initial detection image;

[0011] S12. Determine a set of companion pixels of a head pixel in the initial detection image.

[0012] The beneficial effect of the above further scheme is: in the present invention, the initial detection image contains a large number of pixels with low pixel quality, which are easy to interfere with image processing. Therefore, before the initial detection image is input into the global suppression processing model, the pixels with low quality are first screened, and the set of companion pixels obtained by screening is used to obtain a stable suppression coefficient as the characteristic value of the initial detection image, which is conducive to the subsequent global suppression processing model to complete the image processing in a targeted manner.

[0013] Furthermore, in S11, the pixel point with the maximum pixel value is used as the head pixel point.

[0014] Furthermore, in S12, the companion pixel set of the head pixel The calculation formula is:

[0015] ;

[0016] In the formula, Indicates the number of companion pixels in the set The pixel value of a pixel, represents the pixel value of the head pixel of the initial detection image, Represents an exponent.

[0017] Furthermore, in S2, the stable suppression coefficient of the initial detection image is The calculation formula is:

[0018] ;

[0019] In the formula, Represents the pixel value of the first pixel in the companion pixel set, Indicates the number of companion pixels in the set The pixel value of a pixel, Indicates the number of companion pixels in the set The pixel value of a pixel, represents the pixel value of the head pixel of the initial detection image, represents the number of pixels in the companion pixel set, Indicates the maximum value operation, Represents the minimum operation.

[0020] Further, in S3, the global suppression processing model includes a first convolutional layer, a second convolutional layer, a scaling layer, an adder, and a feature enhancement layer;

[0021] The input end of the first convolutional layer serves as the first input end of the global inhibition processing model; the output end of the first convolutional layer is connected to the input end of the scaling layer; the output end of the scaling layer is connected to the first input end of the adder; the input end of the second convolutional layer serves as the second input end of the global inhibition processing model; the first output end of the second convolutional layer is connected to the second input end of the adder; the second output end of the second convolutional layer is connected to the first input end of the feature enhancement layer; the output end of the adder is connected to the second input end of the feature enhancement layer; the output end of the feature enhancement layer serves as the output end of the global inhibition processing model.

[0022] The beneficial effect of the above further scheme is: in the present invention, the first convolution layer is used to extract features from the initial detection image to obtain a feature map of each pixel, each feature map can be used as the image feature of the pixel, and the feature map is scaled using the scaling layer. The second convolution layer is used to extract the Sobel operator of each pixel in the initial detection image, and the Sobel operator can be used for edge detection to minimize the influence of edge pixels in the input of the adder. The adder is used to fuse the features extracted by the two convolution layers and input them into the feature enhancement layer for pixel enhancement. The feature enhancement layer is used to train the features of the two convolution layers to determine whether there is an abnormal fire area in the initial detection image.

[0023] Furthermore, the expression of the scaling layer is:

[0024] ;

[0025] In the formula, represents the output of the scaling layer, represents the stable suppression coefficient of the initial detection image, represents the logarithmic function, represents the first output of the first convolutional layer, represents the first convolutional layer outputs, represents the first convolutional layer outputs, Indicates rounding up.

[0026] Furthermore, the expression of the adder is:

[0027] ;

[0028] In the formula, represents the output of the adder, represents the output of the scaling layer, Indicates the maximum value operation, Indicates rounding up. represents the first output of the second convolutional layer, represents the second convolutional layer outputs, represents the second convolutional layer outputs.

[0029] In the expression of the adder, the first output terminal and the second output terminal of the second convolutional layer both output the same content. The second convolutional layer contains multiple convolution kernels, and each convolution kernel corresponds to an output.

[0030] Furthermore, the expression of the feature enhancement layer is:

[0031] ;

[0032] In the formula, represents the output of the feature enhancement layer, represents the output of the adder, Indicates rounding up operation. represents the first output of the second convolutional layer, represents the second convolutional layer outputs, represents the second convolutional layer outputs, represents the weight of the first output of the second convolutional layer, represents the second convolutional layer The weight of the output, represents the weight of the Kth output of the second convolutional layer, Represents the activation function.

[0033] The weight of an output in the second convolutional layer is the ratio of that output to the sum of the total outputs.

[0034] The beneficial effects of the present invention are as follows: the present invention proposes a forest fire detection method based on artificial intelligence. Before the initial detection image of the forest area is input into the global suppression processing model, the companion pixel point set of the initial detection image is extracted to obtain a stable suppression coefficient characterizing the situation of more important pixel points; using the stable suppression coefficient as one of the input parameters of the global suppression processing model, the model can be used to process the initial detection image in a targeted manner, thereby determining whether there is a fire area in the initial detection image, and providing timely and accurate information support for fire warning and fire fighting. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 is a flow chart of a forest fire detection method based on artificial intelligence;

[0036] Figure 2 Schematic diagram of the structure of the global inhibition processing model. DETAILED DESCRIPTION

[0037] The embodiments of the present invention are further described below in conjunction with the accompanying drawings.

[0038] like Figure 1 As shown, the present invention provides a forest fire detection method based on artificial intelligence, comprising the following steps:

[0039] S1. Extract pixel point sets from the initial detection image of the forest area to obtain head pixel points and companion pixel point sets respectively;

[0040] S2, obtaining a stable suppression coefficient of the initial detection image through the pixel value of the head pixel and the pixel value of the pixel in the companion pixel set;

[0041] S3, constructing a global inhibition processing model for the initial detection image;

[0042] S4. According to the stable suppression coefficient of the initial detection image, the initial detection image is processed using the global suppression processing model to obtain the forest fire detection result.

[0043] In this embodiment of the present invention, S1 includes the following sub-steps:

[0044] S11, collecting an initial detection image of the forest area through an image sensor, and determining the head pixel points of the initial detection image;

[0045] S12. Determine a set of companion pixels of a head pixel in the initial detection image.

[0046] In the present invention, the initial detection image contains a large number of pixels with low pixel quality, which are easy to interfere with image processing. Therefore, before the initial detection image is input into the global suppression processing model, the pixels with low quality are first screened, and the set of companion pixels obtained by screening is used to obtain a stable suppression coefficient as the characteristic value of the initial detection image, which is conducive to the subsequent global suppression processing model to complete the image processing in a targeted manner.

[0047] In the embodiment of the present invention, in S11, the pixel point with the maximum pixel value is used as the head pixel point.

[0048] In the embodiment of the present invention, in S12, the companion pixel point set of the head pixel point The calculation formula is:

[0049] ;

[0050] In the formula, Indicates the number of companion pixels in the set The pixel value of a pixel, represents the pixel value of the head pixel of the initial detection image, Represents an exponent.

[0051] In the embodiment of the present invention, in S2, the stable suppression coefficient of the initial detection image The calculation formula is:

[0052] ;

[0053] In the formula, Represents the pixel value of the first pixel in the companion pixel set, Indicates the number of companion pixels in the set The pixel value of a pixel, Indicates the number of companion pixels in the set The pixel value of a pixel, represents the pixel value of the head pixel of the initial detection image, represents the number of pixels in the companion pixel set, Indicates the maximum value operation, Represents the minimum operation.

[0054] In the embodiment of the present invention, Figure 2 As shown, in S3, the global inhibition processing model includes a first convolutional layer, a second convolutional layer, a scaling layer, an adder, and a feature enhancement layer;

[0055] The input end of the first convolutional layer serves as the first input end of the global inhibition processing model; the output end of the first convolutional layer is connected to the input end of the scaling layer; the output end of the scaling layer is connected to the first input end of the adder; the input end of the second convolutional layer serves as the second input end of the global inhibition processing model; the first output end of the second convolutional layer is connected to the second input end of the adder; the second output end of the second convolutional layer is connected to the first input end of the feature enhancement layer; the output end of the adder is connected to the second input end of the feature enhancement layer; the output end of the feature enhancement layer serves as the output end of the global inhibition processing model.

[0056] In the present invention, the first convolution layer is used to extract features from the initial detection image to obtain a feature map for each pixel. Each feature map can be used as the image feature of the pixel, and the feature map is scaled using the scaling layer. The second convolution layer is used to extract the Sobel operator of each pixel in the initial detection image. The Sobel operator can be used for edge detection to minimize the influence of edge pixels in the input of the adder. The adder is used to fuse the features extracted by the two convolution layers and input them into the feature enhancement layer for pixel enhancement. The feature enhancement layer is used to train the features of the two convolution layers to determine whether there is an abnormal fire area in the initial detection image.

[0057] In the embodiment of the present invention, the expression of the scaling layer is:

[0058] ;

[0059] In the formula, represents the output of the scaling layer, represents the stable suppression coefficient of the initial detection image, represents the logarithmic function, represents the first output of the first convolutional layer, represents the first convolutional layer outputs, represents the first convolutional layer outputs, Indicates rounding up.

[0060] In the embodiment of the present invention, the expression of the adder is:

[0061] ;

[0062] In the formula, represents the output of the adder, represents the output of the scaling layer, Indicates the maximum value operation, Indicates rounding up. represents the first output of the second convolutional layer, represents the second convolutional layer outputs, represents the second convolutional layer outputs.

[0063] In the expression of the adder, the first output terminal and the second output terminal of the second convolutional layer both output the same content. The second convolutional layer contains multiple convolution kernels, and each convolution kernel corresponds to an output.

[0064] In the embodiment of the present invention, the expression of the feature enhancement layer is:

[0065] ;

[0066] In the formula, represents the output of the feature enhancement layer, represents the output of the adder, Indicates rounding up operation. represents the first output of the second convolutional layer, represents the second convolutional layer outputs, represents the second convolutional layer outputs, represents the weight of the first output of the second convolutional layer, represents the second convolutional layer The weight of the output, represents the weight of the Kth output of the second convolutional layer, Represents the activation function.

[0067] The weight of an output in the second convolutional layer is the ratio of that output to the sum of the total outputs.

[0068] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific variations and combinations that do not deviate from the essence of the present invention based on the technical revelations disclosed by the present invention, and these variations and combinations are still within the protection scope of the present invention.

Claims

1. A forest fire detection method based on artificial intelligence, characterized in that: The following steps are involved: S1. Extract pixel point sets from the initial detection image of the forest area to obtain head pixel points and companion pixel point sets respectively; S2, obtaining a stable suppression coefficient of the initial detection image through the pixel value of the head pixel and the pixel value of the pixel in the companion pixel set; S3, constructing a global inhibition processing model for the initial detection image; S4. According to the stable suppression coefficient of the initial detection image, the initial detection image is processed using the global suppression processing model to obtain the forest fire detection result.

2. The artificial intelligence-based forest fire detection method according to claim 1, characterized in that: The S1 comprises the following sub-steps: S11, collecting an initial detection image of the forest area through an image sensor, and determining the head pixel points of the initial detection image; S12. Determine a set of companion pixels of a head pixel in the initial detection image.

3. The artificial intelligence-based forest fire detection method according to claim 2, characterized in that: In S11, the pixel point with the maximum pixel value is used as the head pixel point.

4. The artificial intelligence-based forest fire detection method according to claim 2, characterized in that: In S12, the companion pixel set of the head pixel The calculation formula is: ； In the formula, Indicates the number of companion pixels in the set The pixel value of a pixel, represents the pixel value of the head pixel of the initial detection image, Represents an exponent.

5. The artificial intelligence-based forest fire detection method according to claim 1, characterized in that: In S2, the stable suppression coefficient of the initial detection image The calculation formula is: ； In the formula, Represents the pixel value of the first pixel in the companion pixel set, Indicates the number of companion pixels in the set The pixel value of a pixel, Indicates the number of companion pixels in the set The pixel value of a pixel, represents the pixel value of the head pixel of the initial detection image, represents the number of pixels in the companion pixel set, Indicates the maximum value operation, Represents the minimum operation.

6. The artificial intelligence-based forest fire detection method according to claim 1, characterized in that: In S3, the global suppression processing model includes a first convolutional layer, a second convolutional layer, a scaling layer, an adder, and a feature enhancement layer; The input end of the first convolutional layer is used as the first input end of the global inhibition processing model; the output end of the first convolutional layer is connected to the input end of the scaling layer; the output end of the scaling layer is connected to the first input end of the adder; The input end of the second convolutional layer serves as the second input end of the global inhibition processing model; the first output end of the second convolutional layer is connected to the second input end of the adder; the second output end of the second convolutional layer is connected to the first input end of the feature enhancement layer; the output end of the adder is connected to the second input end of the feature enhancement layer; the output end of the feature enhancement layer serves as the output end of the global inhibition processing model.

7. The artificial intelligence-based forest fire detection method according to claim 6, characterized in that: The expression of the scaling layer is: ； In the formula, represents the output of the scaling layer, represents the stable suppression coefficient of the initial detection image, represents the logarithmic function, represents the first output of the first convolutional layer, represents the first convolutional layer outputs, represents the first convolutional layer outputs, Indicates rounding up.

8. The artificial intelligence-based forest fire detection method according to claim 6, characterized in that: The expression of the adder is: ； In the formula, represents the output of the adder, represents the output of the scaling layer, represents the maximum value operation, Indicates rounding up. represents the first output of the second convolutional layer, represents the second convolutional layer outputs, represents the second convolutional layer outputs.

9. The artificial intelligence-based forest fire detection method according to claim 6, characterized in that: The expression of the feature enhancement layer is: ； In the formula, represents the output of the feature enhancement layer, represents the output of the adder, Indicates rounding up operation. represents the first output of the second convolutional layer, represents the second convolutional layer outputs, represents the second convolutional layer outputs, represents the weight of the first output of the second convolutional layer, represents the second convolutional layer The weight of the output, represents the weight of the Kth output of the second convolutional layer, Represents the activation function.

Images