A power transmission line mountain fire smoke detection method and system

Abstract

The application discloses a power transmission line mountain fire smoke detection method and system, comprising the following steps: for the power transmission channel picture to be detected, a pre-trained semantic segmentation model is used to segment out the mountain fire smoke candidate area in the picture and calculate the semantic segmentation confidence score; for each mountain fire smoke candidate area, the shape similarity score is calculated through a shape matching method, the position confidence score is calculated through a position function, and the color confidence score is calculated through color distribution comparison; the semantic segmentation confidence score, the shape similarity score, the position confidence score and the color confidence score are weighted and summed to obtain the final smoke confidence score, which is compared with a preset threshold to determine whether the corresponding mountain fire smoke candidate area is mountain fire smoke. The technical scheme provided by the application can effectively identify the mountain fire smoke condition in the power transmission line channel picture, and significantly reduce the misidentification conditions such as pond smoke, chimney smoke and cloud and mist, and has a wide application scenario.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of image recognition, in particular to a power transmission line forest fire smoke detection method and system. BACKGROUND

[0002] With the rapid development of China's power industry, the power grid scale continues to expand, and the safe and stable operation of the power transmission line, as the main artery of power transmission, is of great importance. However, the power transmission line often passes through complex geographical environments such as mountains and forests, and is threatened by various natural risks. Among them, forest fires, due to their strong suddenness and great destructiveness, have become one of the main factors endangering the safety of power transmission lines. The thick smoke produced by forest fires can cause flashover of insulator strings, causing line tripping, and can also cause large-scale power outages, posing a serious challenge to power grid stability and social and economic activities. Therefore, timely and accurate monitoring of forest fire smoke in the power transmission line corridor is an important part of disaster prevention and mitigation for the power system.

[0003] At present, there are many deficiencies in the detection method of forest fire smoke on the power transmission line, such as the poor adaptability of the traditional threshold-based detection method to complex environments, the difficulty of fixed threshold to cope with complex and changing natural environment, the possibility of false positives due to the general rise in environmental temperature in summer, the decline in sensitivity of sensors due to dust and rain pollution, causing false negatives, and the fact that single-point sensors can only cover a limited area, requiring the deployment of a large number of devices to achieve full-line monitoring, which is costly and difficult to maintain. In particular, forest fires usually start in remote locations and are difficult to capture effectively by a sparsely distributed sensor network. With the popularity of video monitoring, image analysis methods using cameras have been applied, but relying solely on image recognition methods is susceptible to environmental interference such as clouds, fog, water vapor, and shadows, resulting in low recognition accuracy. Therefore, there is an urgent need for a method and system that can accurately and reliably detect forest fire smoke on the power transmission line. SUMMARY

[0004] The technical problem to be solved by the present application is to provide a power transmission line forest fire smoke detection method that can accurately identify forest fire disasters on the power transmission line and has low cost, suitable for wide application.

[0005] To solve the above technical problems, the technical solution adopted by the present application comprises the following steps:

[0006] S1, for the power transmission channel picture to be detected, a pre-trained semantic segmentation model is used to segment the forest fire smoke candidate region in the picture and calculate the semantic segmentation confidence score Score1;

[0007] S2, for each wild fire smoke candidate region, shape similarity score Score2 is calculated by shape matching method, position confidence score Score3 is calculated by position function, color confidence score Score4 is calculated by color distribution comparison;

[0008] S3, semantic segmentation confidence score Score1, shape similarity score Score2, position confidence score Score3 and color confidence score Score4 are weighted and summed to obtain final smoke confidence score Score, and the final smoke confidence score Score is compared with a preset threshold to determine whether the corresponding wild fire smoke candidate region is wild fire smoke.

[0009] Further, in step S1, the wild fire smoke candidate region in the picture is segmented by using the pre-trained semantic segmentation model, and the semantic segmentation confidence score Score1 is calculated, and the function expression for calculating the semantic segmentation confidence score Score1 is:

[0010] ;

[0011] wherein, represents the wild fire smoke candidate region, represents a pixel point in the wild fire smoke candidate region, represents the pixel point belongs to the wild fire smoke category, represents the total number of pixel points of the wild fire smoke candidate region .

[0012] Further, in step S2, the shape similarity score Score2 is calculated by shape matching method, which includes:

[0013] S2.1, the wild fire smoke candidate region is divided into a plurality of sub-regions according to a specified angle around the region center of the wild fire smoke candidate region, for each sub-region i, a direction vector is constructed according to its two edge vertices , so as to obtain the vector feature composed of the direction vectors of all sub-regions . ;

[0014] S2.2, the vector feature is compared with the vector feature of each wild fire smoke region in the given wild fire smoke region vector feature library to calculate the similarity S:

[0015] ;

[0016] The wild fire smoke region vector feature library includes vector features of a plurality of known wild fire smoke regions .

[0017] S2.3, select the maximum similarity S as the shape similarity score Score2 of the wildfire smoke candidate region.

[0018] Further, the calculation function expression of the position confidence score Score3 in step S2 is:

[0019] ;

[0020] wherein, represents the height of the power transmission channel picture to be detected, represents the uppermost point of the wildfire smoke candidate region, represents the lowermost point of the wildfire smoke candidate region, and satisfies .

[0021] Further, the calculation function expression of the color confidence score Score4 in step S2 is:

[0022] ;

[0023] wherein, represents the covariance of the color values of the RGB three channels of the wildfire smoke candidate region, represents the average value of the color values of the RGB three channels of the wildfire smoke candidate region, represents the mean value of the color values of the RGB three channels of the wildfire smoke candidate region.

[0024] Further, the calculation formula of the final smoke confidence score Score in step S3 is as follows:

[0025] ;

[0026] wherein, , , , are the weight scores of the semantic segmentation confidence score Score1, the shape similarity score Score2, the position confidence score Score3, and the color confidence score Score4, respectively, and satisfy:

[0027] .

[0028] Further, the function expression of the loss function used by the pre-trained semantic segmentation model in step S1 during training is:

[0029] ;

[0030] wherein, represents the loss function, denotes the number of samples, and denote the height and width of the power transmission channel picture to be detected, respectively, denotes the total number of classes, the classes including at least a smoke class and a non-smoke class, denotes the sample coordinates in the sample pixel belonging to the class is 1, otherwise 0; denotes the sample coordinates in the sample pixel belonging to the class probability.

[0031] The present application also proposes a forest fire smoke detection system, comprising a microprocessor and a memory connected to each other, the microprocessor being programmed or configured to perform the power transmission line forest fire smoke detection method described above.

[0032] The present application also proposes a computer readable storage medium, the computer readable storage medium storing a computer program, the computer program being executed by a processor to implement the steps of the method described above.

[0033] The present application also proposes a computer program product, comprising a computer program, the computer program being executed by a processor to implement the steps of the method described above.

[0034] Compared with the prior art, the present application has the following advantages:

[0035] The present application can effectively identify forest fire smoke conditions in power transmission line channel pictures by comprehensively considering semantic segmentation confidence scores, shape similarity scores, position confidence scores, and color confidence scores, and significantly reduces the misidentification of conditions such as pond smoke, chimney smoke, and clouds, and has a wide range of application scenarios. BRIEF DESCRIPTION OF DRAWINGS

[0036] Figure 1 Flowchart of the present embodiment. DETAILED DESCRIPTION

[0037] The present application will be further described below in conjunction with the accompanying drawings and specific preferred embodiments, but the protection scope of the present application is not limited thereby.

[0038] The technical solution adopted by the present embodiment is shown in Figure 1 , comprising the following steps:

[0039] ​S1, for the power transmission channel picture to be detected, using a pre-trained semantic segmentation model to segment the picture to obtain a forest fire smoke candidate region and calculate a semantic segmentation confidence score Score1;

[0040] S2, for each forest fire smoke candidate region, respectively calculating a shape similarity score Score2 by a shape matching method, a position confidence score Score3 by a position function, and a color confidence score Score4 by color distribution comparison;

[0041] S3, weighting and summing the semantic segmentation confidence score Score1, the shape similarity score Score2, the position confidence score Score3, and the color confidence score Score4 to obtain a final smoke confidence score Score, and comparing the final smoke confidence score Score with a preset threshold to determine whether the corresponding forest fire smoke candidate region is forest fire smoke.

[0042] The embodiment can effectively identify forest fire smoke in the power transmission channel picture by comprehensively considering the semantic segmentation confidence score, the shape similarity score, the position confidence score, and the color confidence score, and significantly reduces the misidentification of, for example, pond smoke, chimney smoke, and cloud, thereby having a wide range of application scenarios.

[0043] Preferably, in step S1, when the pre-trained semantic segmentation model is used to segment the picture to obtain the forest fire smoke candidate region and calculate the semantic segmentation confidence score Score1, the function expression for calculating the semantic segmentation confidence score Score1 is:

[0044] ;

[0045] wherein, represents the forest fire smoke candidate region, represents a pixel point in the forest fire smoke candidate region, represents the pixel point belonging to the forest fire smoke category, represents the total number of pixel points in the forest fire smoke candidate region .

[0046] Preferably, in step S2, the shape similarity score Score2 is calculated by the shape matching method, which includes:

[0047] S2.1, dividing the forest fire smoke candidate region into a plurality of sub-regions according to a specified angle around the region center of the forest fire smoke candidate region, and for each sub-region i, constructing a direction vector according to two edge vertices thereof, so as to obtain a vector feature composed of the direction vectors of all the sub-regions.

[0048] S2.2, calculate the vector feature of the vector feature library of the given wildfire smoke region and the vector feature of each wildfire smoke region in the vector feature library of the given wildfire smoke region Calculate the similarity S:

[0049] ;

[0050] The vector feature library of the wildfire smoke region includes vector features of a plurality of known wildfire smoke regions ;

[0051] S2.3, select the maximum similarity S as the shape similarity score Score2 of the wildfire smoke candidate region.

[0052] In this embodiment, for each wildfire smoke candidate region, the region edge is divided into 360 equal parts around the region center, and each pair of edge vertexes can obtain a direction vector , and the vertical vertex is selected as the starting point around the region center, and the direction vector list[ , ,... ] is obtained in the clockwise direction, thereby obtaining a vector feature V1 containing 360 elements; the vector feature of the known wildfire smoke region is the vector feature generated according to the above vector feature V1 generation method in the image data set of the labeled wildfire region.

[0053] Preferably, the calculation function expression of the position confidence score Score3 in step S2 is:

[0054] ;

[0055] wherein, represents the height of the power transmission channel picture to be detected, represents the uppermost point of the wildfire smoke candidate region, represents the lowermost point of the wildfire smoke candidate region, and satisfies .

[0056] That is, taking the 1 / 2 of the height of the power transmission channel picture as the separation line, for the wildfire smoke candidate region, if the region is entirely below the separation line, the position confidence score is 1; if the separation line passes through the region, the position confidence score is the proportion of the separation line located in the height of the wildfire smoke candidate region from bottom to top; if the region is entirely above the separation line, the position confidence score is 0. Preferably, the calculation function expression of the color confidence score Score4 in step S2 is:

[0057] ;

[0058] wherein, denotes the covariance of the color values of the RGB three channels of the wildfire smoke candidate region, denotes the mean value of the color values of the RGB three channels of the wildfire smoke candidate region, denotes the mean value of the color values of the RGB three channels of the wildfire smoke candidate region.

[0059] Preferably, the final smoke confidence score Score in step S3 is calculated according to the following formula:

[0060] ;

[0061] wherein, , , , are weight scores of the semantic segmentation confidence score Score1, the shape similarity score Score2, the position confidence score Score3, and the color confidence score Score4, respectively, and satisfy:

[0062] .

[0063] Preferably, the loss function used in the training of the pre-trained semantic segmentation model in step S1 has the following functional expression:

[0064] ;

[0065] wherein, denotes the loss function, denotes the number of samples, and denote the height and width of the power transmission channel picture to be detected, respectively, denotes the total number of categories, including at least the smoke category and the non-smoke category, denotes the indication variable that the pixel at coordinate in sample belongs to category , which is 1 when the pixel at coordinate in sample belongs to category , and 0 otherwise; denotes the probability that the pixel at coordinate in sample predicted by the semantic segmentation model belongs to category .

[0066] The embodiment also provides a forest fire smoke detection system, comprising a microprocessor and a memory connected with each other, the microprocessor is programmed or configured to execute the power line forest fire smoke detection method.

[0067] The embodiment also provides a computer readable storage medium, wherein a computer program is stored in the computer readable storage medium, and the computer program is executed by a processor to implement the steps of the method.

[0068] The embodiment also provides a computer program product, comprising a computer program, and the computer program is executed by a processor to implement the steps of the method.

[0069] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system or a computer program product. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer readable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage devices, etc.) containing computer usable program code. The present application is described with reference to flowcharts and / or block diagrams of the methods, apparatus (systems) and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and a combination of flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing apparatus to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing apparatus produce a device implemented in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 an apparatus with a function specified in one or more flows and / or blocks. These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to work in a specific manner, so that the instructions stored in the computer readable memory produce a manufactured product including instruction apparatus, which implements the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 an apparatus with a function specified in one or more flows and / or blocks. These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to work in a specific manner, so that the instructions stored in the computer readable memory produce a manufactured product including instruction apparatus, which implements the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 an apparatus with a function specified in one or more flows and / or blocks. These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to work in a specific manner, so that the instructions stored in the computer readable memory produce a manufactured product including instruction apparatus, which implements the flowcharts and / or block diagrams.

[0070] The above merely describes the preferred embodiments of the present application, and the protection scope of the present application is not limited to the above-described embodiments. Any technical solution falling within the concept of the present application shall fall within the protection scope of the present application. It should be noted that, for ordinary skilled persons in the art, some improvements and refinements without departing from the principles of the present application shall also be considered as falling within the protection scope of the present application.

Claims

1. A method for detecting smoke from wildfires on power transmission lines, characterized in that, The method comprises the following steps: S1, for the picture of the power transmission channel to be detected, a pre-trained semantic segmentation model is used to segment the wild fire smoke candidate region in the picture and calculate a semantic segmentation confidence score Score1; S2, for each wild fire smoke candidate region, a shape similarity score Score2 is calculated by a shape matching method, a position confidence score Score3 is calculated by a position function, and a color confidence score Score4 is calculated by color distribution comparison; S3, the semantic segmentation confidence score Score1, the shape similarity score Score2, the position confidence score Score3 and the color confidence score Score4 are weighted and summed to obtain a final smoke confidence score Score, and the final smoke confidence score Score is compared with a preset threshold to determine whether the corresponding wild fire smoke candidate region is wild fire smoke; In step S2, the shape similarity score Score2 is calculated by a shape matching method, which comprises: S2.1, divide the wild fire smoke candidate region into multiple sub-regions according to a specified angle around the region center of the wild fire smoke candidate region, for each sub-region i, construct a direction vector according to its two edge vertices , so as to obtain the direction vectors of all sub-regions , and the vector features composed of the direction vectors of all sub-regions ; S2.2, vector features of the given wildfire smoke region and vector features of each wildfire smoke region in the given library of vector features of wildfire smoke regions calculate the similarity S: ； The vector feature library of the mountain fire smoke area includes vector features of a plurality of known mountain fire smoke areas ; S2.3, the maximum similarity S is selected as the shape similarity score Score2 of the wild fire smoke candidate region; The calculation function expression of the position confidence score Score3 in step S2 is: ； wherein, represents a height of a picture of a power transmission channel to be detected, represents an uppermost point of the forest fire smoke candidate region, represents a lowermost point of the forest fire smoke candidate region, and satisfies .

2. The power line right-of-way wildfire smoke detection method of claim 1, wherein, In step S1, the pre-trained semantic segmentation model is used to segment the wild fire smoke candidate region in the picture and calculate the semantic segmentation confidence score Score1, and the function expression for calculating the semantic segmentation confidence score Score1 is: ； wherein, represents a wild fire smoke candidate region, represents a pixel point within the wild fire smoke candidate region, represents a pixel point belongs to the wild fire smoke class, represents a wild fire smoke candidate region total number of pixel points.

3. The power line right-of-way wildfire smoke detection method of claim 1, wherein, The calculation function expression of the color confidence score Score4 in step S2 is: ； wherein, a covariance of color values of the RGB three channels of the bushfire smoke candidate region, a mean of color values of the RGB three channels of the bushfire smoke candidate region, a mean of color values of the RGB three channels of the bushfire smoke candidate region.

4. The power line right-of-way wildfire smoke detection method of claim 1, wherein, The calculation formula of the final smoke confidence score Score in step S3 is as follows: ； wherein, , , , are weight scores of semantic segmentation confidence score Scorel, shape similarity score Score2, position confidence score Score3, color confidence score Score4, respectively, and satisfy: 。 5. The power line right-of-way wildfire smoke detection method of claim 1, wherein, The function expression of the loss function used by the pre-trained semantic segmentation model in step S1 during training is: ； in, Represents the loss function. Indicates the number of samples. and These represent the height and width of the power transmission channel image to be inspected, respectively. This indicates the total number of categories, which includes at least smoke categories and non-smoke categories. Indicates sample median coordinate The pixel at that location belongs to the category Indicator variable, when sample median coordinate The pixel at that location belongs to the category The value is 1 if it is true, and 0 otherwise. This represents the samples predicted by the semantic segmentation model. median coordinate The pixel at that location belongs to the category The probability of.

6. A power line bushfire smoke detection system comprising a microprocessor and memory interconnected, characterised in that, The microprocessor is programmed or configured to perform the power transmission line wild fire smoke detection method of any one of claims 1-5.

7. A computer-readable storage medium having stored therein a computer program or instructions, characterized in that, The computer program or instructions are programmed or configured to perform the power transmission line wild fire smoke detection method of any one of claims 1-5 by the processor.

8. A computer program product comprising computer programs or instructions, characterized in that, The computer program or instructions are programmed or configured to perform the power transmission line wild fire smoke detection method of any one of claims 1-5 by the processor.

Images