Satellite remote sensing-based drop zone positioning method

Abstract

The present application relates to the technical field of drop area positioning, and particularly relates to a drop area positioning method based on satellite remote sensing, which acquires environmental monitoring indexes in a target area; sensors are arranged in each grid sub-area according to the environmental monitoring indexes; the patrol area type of each grid sub-area is set according to historical data in each grid sub-area, and the patrol route of a fire-fighting unmanned aerial vehicle is determined according to the patrol area type; a video acquisition terminal is installed on the fire-fighting unmanned aerial vehicle, flame image recognition is performed based on video data, and flame distribution area recognition is acquired to determine a water bomb drop area; the fire-fighting unmanned aerial vehicle loaded with a water bomb device drops water bombs in the flame distribution area; the cloud monitoring center compares the patrol area of a non-flame distribution area with the flame distribution area in terms of feature data similarity to generate an early warning area, and arranges the fire-fighting unmanned aerial vehicle to go to the early warning area to drop water mist for cooling; the ecological environment can be protected to the greatest extent, and economic losses can be reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of drop area positioning, and particularly relates to a drop area positioning method based on satellite remote sensing. BACKGROUND

[0002] Forest, mountain and other terrain areas are seriously damaged by fire, which can easily cause loss of life and property and make rescue missions relatively difficult. However, at the initial stage of a forest fire, accurate fire detection and fire identification can provide early warning and reduce loss of life and property. Satellite remote sensing and unmanned aerial vehicle visual monitoring of fire images can assist in rescue and improve rescue efficiency and ensure the safety of personnel during rescue.

[0003] Existing technologies focus on accurately analyzing the geometric properties of flames in a single picture to obtain a determination result, which has a relatively fast analysis speed, but ignores the analysis of flame trends between consecutive frames of pictures, and the determination result cannot simultaneously take into account the analysis of the dynamic spreading trend of the fire. In addition, the fires in each region follow certain characteristics and rules in the formation process, and the importance of historical fire data of each region for the selection of unmanned aerial vehicle patrol areas is ignored in the existing technologies. In order to solve the above technical problems, the present application provides a drop area positioning method based on satellite remote sensing. SUMMARY

[0004] In order to solve the above technical problems, the present application aims to provide a drop area positioning method based on satellite remote sensing, comprising the following steps:

[0005] Step S1: obtaining the terrain features, climate conditions and historical data in the target area, performing gridding processing on the target area, and extracting environmental monitoring indicators of each grid sub-region according to the terrain features, climate conditions and historical data in the target area;

[0006] Step S2: arranging sensor terminals in each grid sub-region according to the environmental monitoring indicators, obtaining environmental monitoring indicator data of each grid sub-region through the sensor terminals and uploading the data to a cloud monitoring center;

[0007] Step S3: setting the patrol area type of each grid sub-region according to the historical data in each grid sub-region, and determining the patrol route of the fire-fighting unmanned aerial vehicle according to the patrol area type;

[0008] Step S4: installing a video acquisition terminal on the fire-fighting unmanned aerial vehicle for obtaining image data when the fire-fighting unmanned aerial vehicle patrols in the sub-region; performing flame image recognition based on the video data, obtaining flame distribution area recognition, determining a water bomb drop area, and dropping water bombs in the flame distribution area by using the fire-fighting unmanned aerial vehicle loaded with a water bomb device;

[0009] Step S5: The cloud monitoring center converts the environmental monitoring index data and the terrain features in each patrol area into feature vector data and feature terrain data, compares the feature data of the patrol area of the non-flame distribution area with the flame distribution area, generates a warning area according to the comparison result, and arranges the fire-fighting unmanned plane loaded with the water mist device to go to the warning area to perform the water mist cooling operation.

[0010] Further, the terrain features, climate conditions and historical data in the target area are acquired, the target area is gridded, and the process of extracting the environmental monitoring index of each grid sub-area according to the terrain features, climate conditions and historical data in the target area includes:

[0011] The target area is gridded to obtain a plurality of grid sub-areas, GIS geographic data of the plurality of grid sub-areas is acquired by means of GIS, terrain features and climate conditions are acquired according to the GIS geographic data, and climate features corresponding to the climate conditions are extracted according to the environmental index data involved in the climate conditions.

[0012] The historical data of fire occurring in the plurality of grid sub-areas is acquired, the climate features and terrain features in the historical data of fire occurring in each grid sub-area are statistically analyzed to acquire the occurrence times of each climate feature and terrain feature in each grid sub-area, the occurrence times of each climate feature and terrain feature are screened, and the screened climate features and terrain features are summarized to acquire the environmental monitoring index in each grid sub-area.

[0013] Further, the process of arranging a sensor terminal in each grid sub-area according to the environmental monitoring index and acquiring the environmental monitoring index data of each grid sub-area by the sensor terminal and uploading to the cloud monitoring center includes:

[0014] The corresponding sensor category is determined according to the environmental monitoring index in each grid sub-area, the arrangement range of each category of sensor is determined according to the terrain features of each grid sub-area, the number of fires occurring in each grid sub-area is acquired according to the historical data, and the number of sensor arrangements is determined according to the number of fires occurring in the arrangement range.

[0015] The environmental data acquisition terminal of each grid sub-area is constructed according to the arrangement range and the number of sensor arrangements of each category of sensor of each grid sub-area, and the environmental data acquisition terminal acquires the environmental monitoring index data of the grid sub-area and uploads to the cloud monitoring center.

[0016] Further, the process of setting the patrol area type of each grid sub-area according to the historical data in each grid sub-area and determining the patrol route of the fire-fighting unmanned plane according to the patrol area type includes:

[0017] acquire the number of fire occurrences in each grid sub-region, set a fire occurrence threshold, mark the grid sub-regions with the number of fire occurrences greater than or equal to the fire occurrence threshold as necessary patrol regions, and mark the grid sub-regions with the number of fire occurrences less than the fire occurrence threshold as random patrol regions;

[0018] determine the necessary patrol route of the fire unmanned aerial vehicle according to the distribution region of the sequential patrol region;

[0019] acquire the total number of random patrol regions, generate a random number D according to the number of random patrol regions, and D is less than the total number value;

[0020] randomly select D random regions from the random patrol regions, and determine the random patrol route of the fire unmanned aerial vehicle according to the D random regions;

[0021] After the fire unmanned aerial vehicle completes the random patrol route, mark the D random regions selected in this round as a forbidden selection state, randomly select D non-forbidden selection state random regions from the random patrol regions and determine the new random patrol route of the fire unmanned aerial vehicle.

[0022] Further, through the video data acquisition terminal installed on the fire unmanned aerial vehicle, real-time video data of the fire unmanned aerial vehicle during the patrol process in the patrol route is acquired, and the process of flame image recognition based on the video data includes:

[0023] set a preset period, convert the obtained video data into corresponding video frame images, and mark the acquisition time t and the patrol region of the video frame image;

[0024] pre-set reference images of each patrol region, compare the video frame images of the patrol region acquired by the fire unmanned aerial vehicle with the reference images of the corresponding patrol region;

[0025] subtract the pixel values of the video frame image at the t time and the corresponding position of the reference video frame image to obtain a difference value, convert the difference value into a pixel value of a binary image, and generate a binary image;

[0026] set two pixel point traversal pointers and a preset pixel threshold, and start traversing from the first pixel point and the last pixel point of the binary image region at the same time; mark the pixel point region in the binary image with a pixel value greater than the preset pixel threshold as a suspicious region; mark the pixel point region in the binary image with a pixel value less than or equal to the preset pixel threshold as a normal region;

[0027] When the suspicious region in the binary image of the video frame image at the t time is generated, the binary image of the video frame image at the t-1 time and the binary image of the video frame image at the t-2 time are acquired, the upper limit of the similarity threshold and the lower limit of the similarity threshold are set, the binary image of the video frame image at the t time is compared with the binary image of the video frame image at the t-1 time and the binary image of the video frame image at the t-2 time in terms of similarity, and the overall similarity is acquired;

[0028] When the overall similarity is less than or equal to the lower limit of the similarity threshold, the video frame image is marked as existing a small non-fire disturbance source.

[0029] When the overall similarity is greater than or equal to the upper limit of the similarity threshold, the video frame image is marked as existing other lighting device disturbance source.

[0030] When the overall similarity is greater than the lower limit of the similarity threshold and less than the upper limit of the similarity threshold, the video frame image is marked as to be verified, and the secondary determination of the flame feature is performed on the video frame image.

[0031] Further, the process of performing the secondary determination of the flame feature on the video frame image to be verified includes:

[0032] The perimeter of the suspicious region and the area of the suspicious region in the binary image of the video frame image to be verified are acquired, the shape irregularity threshold is set, and the shape irregularity is acquired according to the perimeter of the suspicious region and the area of the suspicious region in the binary image of the video frame image to be verified.

[0033] When the shape irregularity is greater than the shape irregularity threshold, the video frame image is marked as a flame image.

[0034] When the shape irregularity is less than or equal to the shape irregularity threshold, the video frame image is marked as existing a non-fire disturbance source.

[0035] Further, the process of performing the secondary determination of the flame feature on the video frame image to be verified includes:

[0036] The video frame image marked as a flame image is acquired, the suspicious region in the binary image of the video frame image is extracted as a flame distribution region, and the actual patrol region corresponding to the suspicious region is marked as a flame distribution region, the acting area of a single water bomb is acquired, the water bomb dropping position and the water bomb dropping quantity are determined according to the flame distribution region and the acting area of the single water bomb, and the corresponding quantity of water bombs is dropped by the fire-fighting unmanned aerial vehicle to the flame distribution region.

[0037] Further, the cloud monitoring center converts the environmental monitoring index data and the terrain features in each patrol area into feature vector data and feature terrain data, compares the feature data similarity of the patrol area in the non-flame distribution area with the flame distribution area, generates an early warning area according to the comparison result, and arranges a water mist device-loaded fire-fighting unmanned aerial vehicle to go to the early warning area to perform a water mist dropping cooling operation process, which comprises:

[0038] The cloud monitoring center obtains the environmental monitoring index data and the terrain features collected by the data collection terminal of the patrol area, converts the environmental monitoring index data and the terrain features into feature vector data and feature terrain data, and obtains the feature vector data and the feature terrain data of the flame distribution area;

[0039] A feature vector similarity threshold is set, the patrol area with the same feature terrain data as the flame distribution area is selected from the patrol area, the feature vector data of the patrol area is compared with the feature vector data of the flame distribution area one by one, and the feature vector similarity of each patrol area is obtained;

[0040] The patrol area with the feature vector similarity greater than the feature vector similarity threshold is screened out, and the patrol area is marked as an early warning area;

[0041] The cloud monitoring center arranges a fire-fighting unmanned aerial vehicle to go to the early warning area to perform a water mist dropping cooling operation.

[0042] Compared with the prior art, the present application has the following beneficial effects:

[0043] 1. According to the historical fire occurrence data of each area, the patrol type of each area is determined, the characteristics and laws of fire occurrence are followed, and the selection of the unmanned aerial vehicle patrol is more scientific and reasonable.

[0044] 2. The present application firstly compares the similarity of continuous flame images, eliminates the interference factors affecting flame recognition, compares the environmental data of the fire occurrence area with the environmental data of other non-fire areas, arranges a fire-fighting unmanned aerial vehicle to go to the early warning area to perform a water mist dropping cooling operation according to the comparison result, and can realize the maximum protection of the ecological environment and the reduction of economic loss. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 It is a principle diagram of a water mist dropping area positioning method based on satellite remote sensing for the embodiments of the present application. DETAILED DESCRIPTION

[0046] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0047] like Figure 1 As shown, a method for positioning a delivery area based on satellite remote sensing includes the following steps:

[0048] Step S1: Obtain the terrain features, climate conditions and historical data of the target area, perform grid processing on the target area, and extract environmental monitoring indicators for each grid sub-area based on the terrain features, climate conditions and historical data of the target area.

[0049] Step S2: Deploy sensor terminals in each grid area according to the environmental monitoring indicators, acquire environmental monitoring indicator data of each grid area through the sensor terminals and upload them to the cloud monitoring center;

[0050] Step S3: Set the patrol area type for each grid area based on historical data within each grid area, and determine the patrol route for the fire-fighting drone based on the patrol area type;

[0051] Step S4: Install a video acquisition terminal on the firefighting drone to acquire image data of the firefighting drone during patrols in the patrol sub-area; perform flame image recognition based on the video data, and acquire flame distribution area recognition to determine the water bomb delivery area; use the firefighting drone equipped with a water bomb device to deliver water bombs in the flame distribution area.

[0052] Step S5: The cloud monitoring center converts the environmental monitoring index data and terrain features in each patrol area into feature vector data and feature terrain data. It compares the feature data similarity between the patrol area in the non-flame distribution area and the flame distribution area. Based on the comparison results, it generates an early warning area and arranges fire-fighting drones equipped with water mist devices to go to the early warning area to carry out water mist cooling operations.

[0053] It should be further explained that, in the specific implementation process, the process of acquiring the terrain features, climate conditions, and historical data of the target area, performing grid-based processing on the target area, and extracting environmental monitoring indicators for each grid sub-area based on the terrain features, climate conditions, and historical data of the target area includes:

[0054] The target area is gridded to obtain a plurality of grid sub-areas, GIS geographic data of the plurality of grid sub-areas is obtained through GIS means, terrain features and climate conditions are obtained according to the GIS geographic data, and climate features corresponding to the climate conditions are extracted according to environmental index data involved in the climate conditions;

[0055] Historical data of fire occurring in the plurality of grid sub-areas is obtained, statistical analysis is performed on the climate features and the terrain features in the historical data of fire occurring in each grid sub-area, occurrence times corresponding to each climate feature and terrain feature in each grid sub-area are obtained, the occurrence times corresponding to each climate feature and terrain feature are screened, and environmental monitoring indexes in each grid sub-area are obtained by aggregating the screened climate features and terrain features.

[0056] It should be further explained that, in the specific implementation process, the process of arranging sensor terminals in each grid sub-area according to the environmental monitoring indexes and uploading the environmental monitoring index data of each grid sub-area to the cloud monitoring center by the sensor terminals includes:

[0057] The corresponding sensor categories are determined according to the environmental monitoring indexes in each grid sub-area, the arrangement range of each category of sensors is determined according to the terrain features of each grid sub-area, the number of fires occurring in each grid sub-area is obtained according to the historical data, and the number of sensor arrangements is determined according to the number of fires occurring in the arrangement range.

[0058] The environmental data collection terminal of each grid sub-area is constructed according to the arrangement range and the number of sensor arrangements of each category of sensors in each grid sub-area, and the environmental data collection terminal obtains the environmental monitoring index data of the grid sub-area and uploads the data to the cloud monitoring center.

[0059] It should be further explained that, in the specific implementation process, the process of setting the patrol area type of each grid sub-area according to the historical data in each grid sub-area and determining the patrol route of the fire-fighting unmanned aerial vehicle according to the patrol area type includes:

[0060] The number of fires occurring in each grid sub-area is obtained, a fire occurrence threshold is set, a grid sub-area with a number of fires greater than or equal to the fire occurrence threshold is marked as a necessary patrol area, and a grid sub-area with a number of fires less than the fire occurrence threshold is marked as a random patrol area.

[0061] The necessary patrol route of the fire-fighting unmanned aerial vehicle is determined according to the distribution area of the sequential patrol area.

[0062] The total number of random patrol areas is obtained, a random number D is generated according to the number of random patrol areas, and D is less than the total number value.

[0063] D random areas are randomly selected from the random patrol areas, and a random patrol route of the fire-fighting unmanned aerial vehicle is determined according to the D random areas;

[0064] After the fire-fighting unmanned aerial vehicle completes the random patrol route, the D random areas selected in this round are marked as forbidden selection states, D non-forbidden selection state random areas are randomly selected from the random patrol areas, and a new random patrol route of the fire-fighting unmanned aerial vehicle is determined.

[0065] It needs to be further explained that, in the specific implementation process, the video data of the fire-fighting unmanned aerial vehicle in the patrol process on the patrol route is obtained in real time through the video data acquisition terminal installed on the fire-fighting unmanned aerial vehicle, and the process of flame image recognition based on the video data includes:

[0066] A preset period is set, the obtained video data is converted into corresponding video frame images, and the acquisition time t and the patrol area of the video frame image are marked;

[0067] The reference images of each patrol area are set in advance, and the video frame image of the patrol area obtained by the fire-fighting unmanned aerial vehicle is compared with the reference image of the corresponding patrol area;

[0068] The pixel values of the video frame image at the t time and the corresponding positions of the reference video frame image are subtracted to obtain a difference value, the difference value is converted into a pixel value of a binary image, and a binary image is generated;

[0069] Two pixel point traversal pointers and a preset pixel threshold value are set, and traversal starts from the first pixel point and the last pixel point of the binary image area at the same time; the pixel point area with a pixel value greater than the preset pixel threshold value in the binary image is marked as a suspicious area; the pixel point area with a pixel value less than or equal to the preset pixel threshold value in the binary image is marked as a normal area;

[0070] When the binary image of the video frame image at the t time generates a suspicious area, the binary images of the video frame images at the t-1 time and the t-2 time are obtained, a similarity threshold upper limit and a similarity threshold lower limit are set, the binary image of the video frame image at the t time is compared with the binary images of the video frame images at the t-1 time and the t-2 time in terms of similarity, and an overall similarity is obtained;

[0071] When the overall similarity is less than or equal to the similarity threshold lower limit, the video frame image is marked as existing a small non-fire disaster interference source;

[0072] When the overall similarity is greater than or equal to the similarity threshold upper limit, the video frame image is marked as existing other lighting device interference sources;

[0073] When the total similarity is greater than the lower similarity threshold and less than the upper similarity threshold, the video frame image is marked as to be verified, and a second determination of the flame feature is performed on the video frame image.

[0074] It needs to be further explained that, in the specific implementation process, the process of performing a second determination of the flame feature on the video frame image to be verified includes:

[0075] The perimeter of the suspicious area and the area of the suspicious area in the binary image of the video frame image to be verified are obtained, a shape irregularity threshold is set, and the shape irregularity is obtained according to the perimeter of the suspicious area and the area of the suspicious area in the binary image of the video frame image to be verified.

[0076] When the shape irregularity is greater than the shape irregularity threshold, the video frame image is marked as a flame image.

[0077] When the shape irregularity is less than or equal to the shape irregularity threshold, the video frame image is marked as having a non-fire disturbance source.

[0078] It needs to be further explained that, in the specific implementation process, the process of identifying the flame distribution area based on the flame image and using the fire-fighting unmanned aerial vehicle loaded with the water bomb device to drop the water bomb in the flame distribution area includes:

[0079] The video frame image marked as the flame image is obtained, the suspicious area in the binary image of the video frame image is extracted as the flame distribution area, and the actual patrol area corresponding to the suspicious area is marked as the flame distribution area. The action area of a single water bomb is obtained, the water bomb dropping position and the water bomb dropping quantity are determined according to the flame distribution area and the action area of the single water bomb, and the fire-fighting unmanned aerial vehicle drops the corresponding quantity of water bombs to the flame distribution area.

[0080] It needs to be further explained that, in the specific implementation process, the cloud monitoring center converts the environmental monitoring index data and the terrain features in each patrol area into feature vector data and feature terrain data, compares the feature data similarity between the patrol area of the non-flame distribution area and the flame distribution area, generates a warning area according to the comparison result, and arranges the fire-fighting unmanned aerial vehicle loaded with the water mist device to go to the warning area to perform the water mist dropping cooling operation process includes:

[0081] The cloud monitoring center obtains the environmental monitoring index data and the terrain features collected by the data collection terminal of the patrol area, converts the environmental monitoring index data and the terrain features into feature vector data and feature terrain data, and obtains the feature vector data and the feature terrain data of the flame distribution area;

[0082] A feature vector similarity threshold is set, a patrol area consistent with the feature terrain data of the flame distribution area is selected from the patrol areas, the feature vector data of the patrol area is compared with the feature vector data of the flame distribution area one by one, and the feature vector similarity of each patrol area is obtained;

[0083] The patrol area with the feature vector similarity greater than the feature vector similarity threshold is screened out, and the patrol area is marked as a pre-warning area;

[0084] The cloud monitoring center arranges a fire-fighting unmanned aerial vehicle to go to the pre-warning area to perform a water mist cooling operation.

[0085] The above embodiments are only used to illustrate the technical method of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical method of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical method of the present application.

Claims

1. A satellite remote sensing-based drop zone positioning method, characterized in that, The method comprises the following steps: Step S1: Obtain the terrain features, climate conditions and historical data in the target area, grid the target area, and extract the environmental monitoring indicators of each grid sub-area according to the terrain features, climate conditions and historical data in the target area; Grid the target area to obtain a plurality of grid sub-areas, obtain GIS geographic data of the plurality of grid sub-areas through GIS means, obtain terrain features and climate conditions according to the GIS geographic data, and extract climate features corresponding to each item of environmental indicator data involved in the climate conditions; Obtain historical data of fire occurring in the plurality of grid sub-areas, statistically analyze the climate features and terrain features in the historical data of fire occurring in each grid sub-area, obtain the occurrence frequency of each climate feature and terrain feature in each grid sub-area, filter the occurrence frequency of each climate feature and terrain feature, and aggregate the filtered climate features and terrain features to obtain the environmental monitoring indicators in each grid sub-area; Step S2: According to the environmental monitoring indicators, arrange sensor terminals in each grid sub-area, obtain the environmental monitoring indicator data of each grid sub-area through the sensor terminals, and upload the data to the cloud monitoring center; Step S3: Set the patrol area type of each grid sub-area according to the historical data in each grid sub-area, and determine the patrol route of the fire-fighting unmanned aerial vehicle according to the patrol area type; Obtain the number of fire occurrences in each grid sub-area, set a fire occurrence threshold, mark the grid sub-areas with a fire occurrence number greater than or equal to the fire occurrence threshold as necessary patrol areas, and mark the grid sub-areas with a fire occurrence number less than the fire occurrence threshold as random patrol areas; Determine the necessary patrol route of the fire-fighting unmanned aerial vehicle according to the distribution of the sequential patrol areas; Obtain the total number of random patrol areas, generate a random number D according to the number of random patrol areas, and D is less than the total number value; Randomly select D random areas from the random patrol areas, and determine the random patrol route of the fire-fighting unmanned aerial vehicle according to the D random areas; After the fire-fighting unmanned aerial vehicle completes the random patrol route, mark the D random areas selected in this round as a non-selected state, randomly select D non-forbidden state random areas from the random patrol areas, and determine the new random patrol route of the fire-fighting unmanned aerial vehicle; Step S4: Install a video acquisition terminal on the fire-fighting unmanned aerial vehicle to obtain image data when the fire-fighting unmanned aerial vehicle patrols in the sub-area; perform flame image recognition based on the video data, obtain flame distribution area recognition, determine a water bomb dropping area, and drop water bombs in the flame distribution area by using the fire-fighting unmanned aerial vehicle loaded with a water bomb device; Step S5: The cloud monitoring center converts the environmental monitoring indicator data and terrain features in each patrol area into feature vector data and feature terrain data, compares the feature data similarity of the patrol areas in the non-flame distribution area and the flame distribution area, generates an early warning area according to the comparison result, and arranges the fire-fighting unmanned aerial vehicle loaded with a water mist device to go to the early warning area to perform water mist dropping operation.

2. The method according to claim 1, wherein, The process of laying out sensor terminals in each grid sub-region according to the environmental monitoring indicators, acquiring environmental monitoring indicator data of each grid sub-region by the sensor terminals, and uploading the data to the cloud monitoring center comprises: determining the corresponding sensor categories according to the environmental monitoring indicators in each grid sub-region, determining the layout range of each category of sensors according to the topographic features of each grid sub-region, acquiring the number of fire occurrences in each grid sub-region according to the historical data, and determining the number of sensors to be laid out according to the number of fire occurrences in the layout range; constructing an environmental data acquisition terminal for each grid sub-region according to the layout range and the number of sensors of each category of sensors in each grid sub-region, and the environmental data acquisition terminal acquiring environmental monitoring indicator data of the grid sub-region where it is located and uploading the data to the cloud monitoring center.

3. The method according to claim 2, wherein, The process of real-time acquisition of video data by the video data acquisition terminal installed on the fire-fighting unmanned aerial vehicle during the patrol process of the fire-fighting unmanned aerial vehicle on the patrol route, and flame image recognition based on the video data comprises: setting a preset period, converting the obtained video data into corresponding video frame images, and marking the acquisition time t and the patrol area of the video frame images; pre-setting reference images of each patrol area, comparing the video frame images of the patrol area acquired by the fire-fighting unmanned aerial vehicle with the reference images of the corresponding patrol area; subtracting the pixel values of the video frame image at time t from the pixel values of the corresponding positions of the reference video frame image to obtain a difference value, converting the difference value into a pixel value of a binary image, and generating a binary image; setting two pixel point traversal pointers and a preset pixel threshold, and starting traversal from the first pixel point and the last pixel point of the binary image area at the same time; marking the pixel point area in the binary image where the pixel value is greater than the preset pixel threshold as a suspicious area; marking the pixel point area in the binary image where the pixel value is less than or equal to the preset pixel threshold as a normal area; when the binary image of the video frame image at time t produces a suspicious area, acquiring the binary images of the video frame images at times t-1 and t-2, setting an upper limit of the similarity threshold and a lower limit of the similarity threshold, comparing the binary image of the video frame image at time t with the binary images of the video frame images at times t-1 and t-2 in terms of similarity, and acquiring an overall similarity; when the overall similarity is less than or equal to the lower limit of the similarity threshold, the video frame image is marked as existing a small non-fire disturbance source; when the overall similarity is greater than or equal to the upper limit of the similarity threshold, the video frame image is marked as existing other lighting device disturbance sources; when the overall similarity is greater than the lower limit of the similarity threshold and less than the upper limit of the similarity threshold, the video frame image is marked as to be verified, and a secondary determination of flame features is performed on the video frame image.

4. The method according to claim 3, wherein, The process of performing a secondary determination of flame features on the to-be-verified video frame image comprises: Obtain the suspicious region perimeter and the suspicious region area in the binary image of the video frame image to be verified, set a shape irregularity threshold, and obtain the shape irregularity according to the suspicious region perimeter and the suspicious region area in the binary image of the video frame image to be verified; When the shape irregularity is greater than the shape irregularity threshold, mark the video frame image as a flame image; When the shape irregularity is less than or equal to the shape irregularity threshold, mark the video frame image as existing non-fire disturbance sources.

5. The method according to claim 4, wherein, Based on the flame image, the process of identifying the flame distribution area and using the fire-fighting unmanned aerial vehicle loaded with the water bomb device to drop the water bomb in the flame distribution area includes: Obtain the video frame image marked as the flame image, extract the suspicious region in the binary image of the video frame image as the flame distribution area, mark the actual patrol area corresponding to the suspicious region as the flame distribution area, obtain the action area of a single water bomb, determine the water bomb dropping position and the water bomb dropping quantity according to the flame distribution area and the action area of a single water bomb, and the fire-fighting unmanned aerial vehicle drops the corresponding quantity of water bombs to the flame distribution area.

6. The method according to claim 5, wherein, The cloud monitoring center converts the environmental monitoring index data and the terrain features in each patrol area into feature vector data and feature terrain data, compares the feature data similarity between the patrol area of the non-flame distribution area and the flame distribution area, generates a warning area according to the comparison result, and arranges the fire-fighting unmanned aerial vehicle loaded with the water mist device to go to the warning area to perform the water mist dropping cooling operation process, which includes: The cloud monitoring center obtains the environmental monitoring index data and the terrain features collected by the data collection terminal of the patrol area, converts the environmental monitoring index data and the terrain features into feature vector data and feature terrain data, and obtains the feature vector data and the feature terrain data of the flame distribution area; Set a feature vector similarity threshold, select the patrol area consistent with the feature terrain data of the flame distribution area from the patrol area, compare the feature vector data of the patrol area with the feature vector data of the flame distribution area one by one, and obtain the feature vector similarity of each patrol area; Screen out the patrol area with the feature vector similarity greater than the feature vector similarity threshold, and mark the patrol area as a warning area; The cloud monitoring center arranges the fire-fighting unmanned aerial vehicle to go to the warning area to perform the water mist dropping cooling operation.

Images