Power transmission line forest fire prediction method and device based on multi-modal data fusion

Abstract

The application relates to the technical field of power transmission line mountain fire monitoring and early warning, and discloses a power transmission line mountain fire prediction method and device based on multi-modal data fusion. Fixed images are acquired based on multiple fixed cameras and spliced into a full-width image, global mountain fire features are identified from the full-width image, the aggregation degree value of the global mountain fire position is calculated, and the generation speed value, the disappearance speed value and the development degree value within a preset first time period are calculated, and then a mountain fire prediction value is obtained. If the prediction value is greater than a reference prediction value, the center position of the aggregation area is calculated and a mobile camera is called to move to the position, multiple mobile area images are shot during the movement, local mountain fire features are identified from the multiple mobile area images, local aggregation areas are calculated, the aggregation area movement vector is calculated according to the multiple local aggregation areas, and an early warning is issued when the local aggregation area corresponds to the center position and the center position is on the aggregation area movement vector. The method improves the monitoring accuracy and reduces the false alarm rate through multi-modal data fusion and dynamic monitoring.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power transmission line forest fire monitoring and early warning, and in particular to a power transmission line forest fire prediction method and device based on multi-modal data fusion. BACKGROUND

[0002] Power transmission line forest fires are a major hidden danger to the safe operation of power systems. When a forest fire occurs, high-temperature flames and thick smoke can damage the insulating layer of the line, causing the conductor to burn out and triggering line tripping or even large-scale power outages. In addition, forest fires also endanger the safety of surrounding residents' lives and property, and are difficult to control, resulting in long recovery periods and significant economic losses.

[0003] Currently, there are two main conventional methods for predicting power transmission line forest fires. One is based on meteorological data, which collects temperature, humidity, wind speed and other meteorological parameters, and establishes a model combining historical forest fire data to predict the probability of forest fires. The other is based on video monitoring, which uses cameras installed on towers to capture real-time images of the surrounding area, and uses artificial or intelligent methods to identify fire conditions. The former can assess the risk of a large area in advance, and the latter can directly capture signs of fire on site.

[0004] However, both of these conventional methods have obvious general shortcomings. The prediction based on meteorological data is easily affected by local topography, vegetation and other factors, making it difficult to accurately reflect the specific risks around the line. The monitoring based on video surveillance is limited by weather conditions (such as heavy fog, heavy rain) and lighting conditions, and is prone to delayed or incorrect identification. Both methods rely on a single source of data, and the accuracy of the forest fire data needs to be improved. SUMMARY

[0005] To improve the accuracy of forest fire data, the present application provides a power transmission line forest fire prediction method and device based on multi-modal data fusion.

[0006] In a first aspect, the present application provides a power transmission line forest fire prediction method based on multi-modal data fusion, which adopts the following technical solution:

[0007] A power transmission line forest fire prediction method based on multi-modal data fusion, comprising the following steps:

[0008] Obtaining a plurality of fixed images based on a plurality of fixed cameras, the fixed cameras being located at fixed positions in different locations; and splicing and fusing the plurality of fixed images into a fixed full-width image;

[0009] identify a global wildfire feature from the fixed full-width image, output the identified global wildfire feature and a corresponding global wildfire position, and calculate an aggregation degree value of the global wildfire position; calculate a generation speed value and a disappearance speed value of the global wildfire feature within a preset first time period, calculate a development degree value according to the generation speed value and the disappearance speed value; calculate a wildfire prediction value according to the aggregation degree value and the development degree value;

[0010] If the wildfire prediction value is greater than a preset reference prediction value, calculate a center position of an aggregation area of the global wildfire position, and call a mobile camera to move to the center position;

[0011] The mobile camera moves to capture a plurality of mobile area images with different positions, and the positions of the mobile area images are located within the area corresponding to the fixed full-width image;

[0012] Identify a local wildfire feature from the mobile area image, output the identified local wildfire feature and a corresponding local wildfire position, and calculate a local aggregation area of the local wildfire feature according to the local wildfire position;

[0013] Calculate a plurality of local aggregation areas based on a plurality of mobile area images, and calculate an aggregation area movement vector according to the positions of the plurality of local aggregation areas; the direction of the aggregation area movement vector is opposite to the direction of the movement vector of the mobile camera;

[0014] If the local aggregation area corresponds to the center position and the center position is on the aggregation area movement vector, issue a wildfire warning.

[0015] By using the above technical solution, through multi-modal data fusion and dynamic monitoring mechanism, a double verification system is constructed from global early warning to local accurate judgment, which improves monitoring accuracy, reduces false alarm rate and optimizes resource allocation; specifically, a fixed camera is used to splice a full-width image to realize large-scale monitoring, and a wildfire feature is quantified to calculate and predict risks; when the risk is high, a mobile camera is called to focus on the core area, and accurate data is obtained through local feature recognition and aggregation area analysis; the direction of the aggregation area movement vector is opposite to the direction of the movement vector of the mobile camera, and double verification is performed combined with the position coincidence condition, which greatly reduces the false alarm rate. At the same time, the fixed and mobile devices work together, resources are activated as needed, and costs are effectively saved.

[0016] Optionally, before comparing the wildfire prediction value, the following sub-steps are further included:

[0017] Search for a fiber temperature measuring device within a preset distance range from the center position, and obtain a detection temperature value generated by the fiber temperature measuring device;

[0018] calculating a center temperature according to the detected temperature value, and calculating a temperature ratio between the center temperature and a preset reference temperature value;

[0019] if the temperature ratio is greater than a preset reference ratio, then adjusting the wildfire prediction value according to a positive correlation of the temperature ratio.

[0020] By adopting the above technical solution, the accuracy of wildfire prediction is significantly improved through data fusion of the optical fiber temperature measurement device, through multi-source data complementation and dynamic adjustment mechanism. The optical fiber temperature measurement can capture temperature abnormalities in visual blind areas, and the judgment universality is enhanced by the quantitative temperature ratio; when the temperature ratio exceeds the threshold value, the wildfire prediction value is adjusted in a positive correlation, and dynamic correction of risk assessment is realized. On the basis of vision, temperature data is supplemented for double verification, reduces the false alarm rate, and accurately locates the risk area.

[0021] Optionally, the method further comprises the following steps:

[0022] calculating a shortest distance value between the optical fiber temperature measurement device and the center position;

[0023] adjusting the reference ratio according to a positive correlation of the shortest distance value; the smaller the shortest distance, the smaller the reference ratio; the larger the shortest distance, the larger the reference ratio;

[0024] adjusting the distance range according to a positive correlation of the development degree value; the smaller the development degree value, the smaller the distance range; the larger the development degree value, the larger the distance range.

[0025] By adopting the above technical solution, when the optical fiber temperature measurement device is close to the center position, the reference ratio is reduced, which means that the sensitivity to temperature abnormalities is improved, and potential fire can be captured in time; when the distance is far, the reference ratio is increased, which avoids misjudgment caused by temperature measurement error at a long distance, and makes the temperature verification more in line with the actual risk. If the wildfire development degree value is small, the fire spreads slowly, the distance range is reduced, the high-risk core area is focused, and invalid data processing is reduced; if the development degree value is large, the fire spreads rapidly, and the distance range is expanded, so as to monitor the surrounding potential affected area in advance and gain more time for emergency decision-making.

[0026] Optionally, the calculation steps of the generation speed value, the disappearance speed value and the development degree value comprise the following sub-steps:

[0027] statistically counting a global wildfire feature new quantity in a preset first time period T1 as Q1, and calculating a generation speed value V1: V1=Q1 / T1;

[0028] statistically counting a global wildfire feature reduction quantity in a preset first time period T1 as Q2, and calculating a disappearance speed value V2: V2=Q2 / T1;

[0029] The development degree value = V1 / V2;

[0030] The calculation step of the aggregation degree value comprises the following sub-steps:

[0031] Taking the center position of the aggregation area of the global wildfire position as the center of a circle, the number Q3 of data points within a set neighborhood radius R is counted, and if the number Q3 of data points exceeds a preset number threshold, the aggregation degree value is calculated: aggregation degree value = Q3 / (π×R 2 ).

[0032] By adopting the above technical solution, the spark generation / disappearance speed value of the wildfire reflects the feature increase / decrease rate of the wildfire in real time, the development degree value predicts the evolution direction of the fire through the ratio of the two, and the turning point of the fire can be sensitively captured, such as the transition from sporadic fire points to large-area spread. The density formula is used to calculate the aggregation degree value, the spatial aggregation of the wildfire position is converted into the point density in the unit area, and the concentration degree of the risk is directly reflected.

[0033] Optionally, the method further comprises the following sub-steps:

[0034] The length of the first time period is inversely related to the aggregation degree value, that is, the smaller the aggregation degree value, the longer the length of the first time period, and the greater the aggregation degree value, the shorter the length of the first time period;

[0035] The reference prediction value is inversely related to the development degree value, that is, the smaller the development degree value, the greater the reference prediction value, and the greater the development degree value, the smaller the reference prediction value.

[0036] By adopting the above technical solution, when the aggregation degree value is low, the first time period T1 is extended, the data fluctuation is smoothed through a longer time window, and false positives triggered by sporadic isolated fire points are avoided. When the aggregation degree value is high, T1 is shortened, and the response speed to rapidly developing fire is improved. When the development degree value is low (the fire is weakened), the reference prediction value is increased, and the warning is triggered only when the risk is significantly increased, thereby avoiding excessive response. When the development degree value is high (the fire is enhanced), the reference prediction value is reduced, and the potential risk is more sensitive.

[0037] Optionally, the method further comprises the following steps:

[0038] The optical fiber temperature measuring device is arranged by an unmanned device driving the movement of the mobile camera, the optical fiber temperature measuring device comprises an optical fiber line body and an optical fiber sensing body arranged at an end point of the optical fiber line body, the optical fiber sensing body is used to generate the detection temperature value and transfer optical fiber data on the optical fiber line body, and the optical fiber sensing body has a wireless communication module and an optical fiber communication module;

[0039] The unmanned device has a fiber communication module and a wireless communication module connected with the fiber sensing body;

[0040] In an initial state, the unmanned device establishes a fiber communication connection with the fiber sensing body;

[0041] The unmanned device obtains a first deployment instruction from the background through the wireless communication module and forwards the first deployment instruction to the fiber sensing body;

[0042] The fiber sensing body establishes a wireless connection with the background in response to the first deployment instruction, and the unmanned device deploys one of the fiber sensing bodies and establishes a fiber communication connection with another fiber sensing body that has not been deployed;

[0043] The unmanned device obtains a second deployment instruction from the background and deploys another fiber sensing body.

[0044] By adopting the above technical solution, the unmanned device deploys the fiber sensing body in stages according to the instruction, can quickly build a distributed temperature measurement point in a high-risk area, improves the laying efficiency, and adapts to the demand for rapid spread of forest fires; the fiber sensing body has both fiber and wireless communication modules, in a normal state, through fiber direct connection with the unmanned aerial vehicle to ensure low-delay data transmission, after deployment, switch to wireless communication to maintain connection with the background, and the dual-link design avoids communication interruption.

[0045] Optionally, in the step of obtaining the second deployment instruction from the background by the unmanned device through the wireless communication module, the following sub-steps are further included:

[0046] If the unmanned device is connected with the background through the wireless communication module, the signal strength of the wireless connection is detected, and if the signal strength is greater than or equal to a preset reference strength value, the unmanned device obtains the second deployment instruction through the wireless communication module;

[0047] If the unmanned device is not connected with the background through the wireless communication module or the signal strength is less than the preset reference strength value, the unmanned device establishes a connection with the background through the wireless temperature measurement module, the background sends the second deployment instruction to the fiber sensing body that has been deployed and is connected wirelessly, the fiber sensing body forwards the second deployment instruction to the fiber sensing body to be deployed through the fiber line body, and the fiber sensing body to be deployed forwards the second deployment instruction to the unmanned device through the fiber communication module.

[0048] By adopting the technical scheme, the multi-link redundancy mechanism of wireless communication and fiber relay is constructed, the instruction transmission reliability in a complex environment is significantly improved, when the unmanned device and the background wireless signal strength meet the standard, the second deployment instruction is directly received wirelessly, and low-delay response is ensured, if the wireless connection is interrupted or the signal is weak, the fiber relay link is automatically switched to, the deployed fiber sensing body is used as a signal relay station, the distance and terrain restrictions of wireless communication are broken, and the instruction transmission success rate is improved.

[0049] Optionally, the step of deploying another fiber sensing body further includes the following sub-steps:

[0050] If the unmanned device is not connected to the background through the wireless communication module or the signal strength is less than the preset reference strength value, after the unmanned device and the fiber sensing body establish wireless connection through the wireless communication module, the connection of the fiber communication module is disconnected, and another fiber sensing body is deployed.

[0051] By adopting the technical scheme, the risk of fiber line body winding is avoided, and the deployment success rate is improved.

[0052] In a second aspect, the application provides a power transmission line forest fire prediction device based on multi-modal data fusion, which adopts the following technical scheme:

[0053] A power transmission line forest fire prediction device based on multi-modal data fusion includes a processor, and the processor executes the steps of the power transmission line forest fire prediction method based on multi-modal data fusion according to any one of the above.

[0054] In summary, the application includes at least one of the following beneficial technical effects:

[0055] By using the multi-angle image stitching technology of the fixed camera, a high-definition monitoring picture covering the whole power transmission line is constructed, and the deep learning algorithm is combined to accurately identify the global forest fire features, so that the weak electric spark phenomenon in the initial stage of the power transmission line fire can be captured. On this basis, the temperature data of the fiber temperature measuring device is introduced to realize the multi-modal data fusion of vision and temperature perception, make up for the blind area defect of single visual monitoring, and significantly improve the forest fire prediction accuracy. At the same time, based on the quantitative calculation of the forest fire features, such as the dynamic analysis of the generated speed value, the disappeared speed value and the development degree value, the fire evolution trend can be tracked in real time.

[0056] A double verification system from global early warning to local accurate judgment is constructed. First, the global forest fire features are identified through fixed full-image recognition to calculate the forest fire prediction value. When the prediction value exceeds the threshold, the mobile camera is enabled to focus on the high-risk area for local feature recognition. With the constraint condition that the moving vector direction of the gathering area is opposite to the moving vector direction of the mobile camera, combined with the coincidence judgment of the local gathering area and the center position, the fire situation is confirmed twice, which greatly reduces the false alarm rate caused by environmental interference or feature misjudgment. In addition, in the temperature data verification link, through the comparison of the temperature ratio of the optical fiber temperature measuring device and the reference ratio, the authenticity of the fire situation is further verified. BRIEF DESCRIPTION OF DRAWINGS

[0057] Figure 1 is a step diagram of a power transmission line forest fire prediction method based on multi-modal data fusion.

[0058] Figure 2 is a sub-step diagram before comparing the forest fire prediction value. DETAILED DESCRIPTION

[0059] The embodiments of the present application will be described in detail below, and examples of the embodiments are shown in the accompanying drawings.

[0060] In the description of the present specification, the description of the terms "certain embodiments", "one embodiment", "some embodiments", "illustrative embodiments", "example", "specific example" or "some examples" means that the specific features, structures, materials or characteristics described in connection with the described embodiments or examples are included in at least one embodiment or example of the present application. In the present specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0061] The embodiments of the present application disclose a power transmission line forest fire prediction method based on multi-modal data fusion, referring to Figure 1 , comprising the following steps:

[0062] Through the deployment of multiple fixed cameras at different towers, mountain heads and other fixed positions, multi-angle image data is collected in real time. By using image stitching algorithm, the feature points (such as corner points, edge points, etc.) in the image are extracted, the features of the overlapping area are matched, and the transformation matrix is used to align the image, so that the dispersed fixed images are seamlessly fused into a complete fixed full-image. In order to ensure the image stitching quality, the resolution, color depth and other parameters of the image are uniformly calibrated during the collection process.

[0063] The global forest fire features are identified from the fixed full-frame image by using computer vision and deep learning technology (forest fires caused by power transmission lines, initial forest fires caused by power transmission lines are caused by sporadic electric sparks, and the electric sparks are bright and dim, which can be reflected as changes in pixel size or pattern features in the image), and the identified global forest fire features and corresponding global forest fire positions are output. The aggregation degree value of the global forest fire position is calculated, which directly reflects the spatial concentration trend of the forest fire risk. The generation speed value and the disappearance speed value of the global forest fire feature are calculated in a preset first time period, i.e. in a preset first time period (such as the past 30 minutes), the number of occurrences and disappearances of the global forest fire feature are counted based on time series analysis method; the development degree value is calculated by using the ratio of the generation speed value and the disappearance speed value, which shows the dynamic evolution trend of the fire. The weighted average method is used to calculate the forest fire prediction value according to the aggregation degree value and the development degree value, i.e. forest fire prediction value = α × aggregation degree value + β × development degree value, wherein α and β are weight coefficients optimized according to historical data.

[0064] If the forest fire prediction value is greater than the preset reference prediction value, the center position of the aggregation area of the global forest fire position is calculated by using the centroid algorithm (by weighted averaging the coordinates of all forest fire position points, the risk core area is quickly located), and a mobile camera (such as a UAV equipped with a high-definition camera) is called to move to the center position.

[0065] The mobile camera continuously captures multiple moving area images at different positions during flight, and these moving area images are located in the area corresponding to the fixed full-frame image, focusing on the high-risk area in the fixed full-frame image to capture local forest fire details with higher resolution.

[0066] Based on the moving area image, local forest fire features such as color change of flame and type of burning object are further identified, and corresponding local forest fire positions are determined. By analyzing the local forest fire positions in multiple moving area images, the local aggregation area of the local forest fire feature is calculated, and then the aggregation area movement vector is derived according to the position changes of multiple local aggregation areas. The direction of the aggregation area movement vector is opposite to the movement vector of the mobile camera. Only when the local aggregation area coincides with the center position calculated in the early stage and the center position is located on the aggregation area movement vector, the forest fire threat is finally confirmed and a warning is issued.

[0067] A large-scale monitoring field of view is formed by image stitching technology using a fixed camera array, and initial risk assessment is achieved by combining this with quantitative analysis of local wildfire characteristics. When the risk value exceeds the threshold, mobile cameras are automatically triggered to conduct targeted monitoring of the core area, obtaining high-precision data through local feature recognition and dynamic analysis of clustered areas. An overlay location overlap verification mechanism is added to reduce the false alarm rate. The collaborative working mode of fixed equipment and mobile terminals enables dynamic scheduling of monitoring resources and saves on operation and maintenance costs.

[0068] Reference Figure 2 To mitigate limitations of image recognition, such as poor nighttime detection and smoke obstruction, the following sub-steps are included before comparing wildfire predictions:

[0069] The system searches for fiber optic temperature sensors within a preset distance range from the center location and acquires the detected temperature values ​​generated by these sensors. These sensors can sense changes in ambient temperature in real time, making them particularly suitable for scenarios where image recognition is ineffective. In nighttime or smoky environments, when fixed and mobile cameras cannot clearly capture the flame pattern on power lines, the fiber optic temperature sensor can accurately detect temperature anomalies around the line using its highly sensitive fiber optic sensor. For example, Yangtze Optical Fibre and Cable's DTS-MMF fiber has optimized attenuation and bandwidth at 1550nm, enabling simultaneous temperature measurement and high-speed communication.

[0070] The center temperature is calculated based on the detected temperature value, and then the temperature ratio between the center temperature and a preset reference temperature value is calculated. By converting the detected temperature value into a quantifiable temperature ratio, wildfire risk assessment becomes more scientific and universal. First, the center temperature of the area covered by the fiber optic temperature measurement device is calculated, and then it is compared with a preset reference temperature value (ambient temperature) to form a temperature ratio. Compared to a single temperature threshold judgment, this ratio calculation method can flexibly adapt to differences in temperature baselines in different seasons and regions. For example, in the high-temperature environment of summer, even if the ambient temperature is high, abnormal temperature rises can still be accurately identified through ratio calculation; while in spring and autumn, this mechanism can also avoid misjudgments caused by ambient temperature fluctuations, ensuring high accuracy of temperature analysis in various scenarios.

[0071] If the temperature ratio is greater than the preset reference ratio, the wildfire prediction value will be adjusted according to the positive correlation of the temperature ratio. If the temperature ratio is much higher than the reference value, it means that there is a significant high temperature anomaly in the monitored area. At this time, the system will significantly increase the wildfire prediction value, accelerate the triggering of subsequent mobile camera monitoring and early warning processes, and shorten the response time.

[0072] The method also includes the following steps:

[0073] Calculate the shortest distance between the fiber optic temperature measuring device and the center position.

[0074] The reference ratio is adjusted according to the shortest distance value. The smaller the shortest distance, the smaller the reference ratio. The larger the shortest distance, the larger the reference ratio. When the shortest distance between the optical fiber temperature measuring device and the center position of the mountain fire gathering area is small, such as less than 50 meters, the system automatically reduces the reference ratio (table method), such as from 1.5 to 1.2, which is equivalent to reducing the trigger threshold of temperature anomaly. For example, if the reference temperature value is 50°C, the reference ratio 1.2 corresponds to a trigger temperature of 60°C, and the temperature measuring device detects 58°C to trigger temperature ratio adjustment, which can capture potential fire in the near distance in advance. By measuring the temperature in the near distance, the reliability of monitoring is improved. The closer the distance, the smaller the temperature conduction loss, so the threshold can be lowered to improve the early warning sensitivity, so as to facilitate the discovery of fire in the budding stage.

[0075] The distance range is adjusted according to the development degree value. The smaller the development degree value, the smaller the distance range. The larger the development degree value, the larger the distance range. When the temperature measuring device is far away from the center position, such as more than 200 meters, the system positively adjusts the reference ratio, such as from 1.5 to 1.8, to increase the threshold of temperature anomaly. For example, long-distance temperature measurement may be affected by environmental air flow, vegetation shielding and other factors, and the error can be expanded to ±3°C. If the reference temperature value is 50°C, the reference ratio 1.8 corresponds to a trigger temperature of 90°C, which can filter false alarms caused by long-distance temperature measurement errors, such as local high temperature caused by direct sunlight. By the logic that the threshold is higher when the distance is farther, the temperature verification standard is dynamically matched with the actual temperature measurement accuracy, and the false alarm rate of long-distance is reduced.

[0076] The calculation steps of the generation speed value, the disappearance speed value and the development degree value include the following sub-steps:

[0077] In a preset first time period T1 (such as 10 minutes), the generation speed value V1=Q1 / T1 and the disappearance speed value V2=Q2 / T1 are calculated by counting the number of new global mountain fire features Q1 and the number of reduced Q2. For example, if 5 bright sparks are added in 10 minutes, then V1=0.5 per minute, which intuitively reflects the intensity of the sudden outbreak of fire; if the disappearance speed value V2=0.3 per minute, it indicates that the net growth of fire is 0.2 per minute in this time period.

[0078] The development degree value is calculated by V1 / V2, which eliminates unit differences and makes the trend analysis in different scenarios comparable. For example:

[0079] When V1 / V2=3, the generation speed is 3 times the disappearance speed, and the fire is in a rapid spread period.

[0080] When V1 / V2=0.8, the fire tends to weaken, and the disappearance speed exceeds the generation speed.

[0081] The calculation steps of the aggregation degree value include the following sub-steps:

[0082] With the center position of the global fire position aggregation area as the center of the circle, the number Q3 of global fire features within the neighborhood radius R (such as 50 meters) of the center is counted, and when Q3 exceeds a preset threshold (such as 3), the aggregation density is calculated by the formula aggregation degree value = Q3 / (π×R 2 ) to calculate the fire point density per unit area, such as "pieces per square meter". For example, R = 50 meters, Q3 = 10, the aggregation density is 10 / (3.14×50 2 )≈0.00127 pieces / m 2 , and the higher the value, the more concentrated the risk. By filtering isolated points through a preset number threshold (such as Q3≥3), the interference of occasional false positives (such as a single thermal imaging false point) on the results is avoided. For example, if there is only one data point in the neighborhood of a certain point, even if the density formula result is not zero, it will be excluded because it does not meet the threshold, reducing the spatial false positive rate to less than 5%. Only when the number of global fire features in the neighborhood is sufficient, the density is calculated to ensure that the aggregation degree value reflects the true risk.

[0083] The method further includes the following sub-steps:

[0084] According to the inverse correlation of the aggregation degree value, the length of the first time period is adjusted, the smaller the aggregation degree value, the longer the first time period; the greater the aggregation degree value, the shorter the first time period. When the aggregation degree value is low, the electric spark is dispersed, and the first time period T1 is automatically extended, such as from 10 minutes to 30 minutes by table lookup method, and the data fluctuations are smoothed through a longer time window. For example, 1-2 fire points occasionally appear, and if a short period T1 is used, it may be misjudged as fire initiation; after extending to 30 minutes, if the number of fire points does not continue to increase, it is determined as an occasional event, avoiding false positives.

[0085] When the aggregation degree value is high, such as more than 5 electric sparks within a 5-meter range, the system automatically shortens T1 to 5 minutes or even 2 minutes through table lookup method, to improve the response speed to rapidly developing fire. For example, high-density aggregation of fire points often indicates that the fire will break out soon, and shortening the time window can capture the trend of new fire points in real time; if 2 new electric sparks are added within 2 minutes, the generation speed value V1 immediately reaches 60 pieces / hour, triggering an emergency warning.

[0086] According to the inverse correlation of the development degree value, the reference prediction value is adjusted; the smaller the development degree value, the greater the reference prediction value; the greater the development degree value, the smaller the reference prediction value. When the development degree value V1 / V2 is less than 1.0, the fire weakens or stabilizes, and the system automatically increases the reference prediction value, such as from 70 minutes to 85 minutes, to increase the warning trigger threshold.

[0087] When the development degree value V1 / V2 is greater than 1.5, the fire spreads rapidly, and the system automatically reduces the reference prediction value to 55 points, which is more sensitive to potential risks. For example, in windy weather, the fire development degree value increases from 1.2 to 2.0, and at this time, reducing the warning threshold can make the system issue an alarm when the mountain fire prediction value reaches 55 points.

[0088] When the electric spark is dispersed, the statistical period T1 is extended to filter the interference of accidental fire points and reduce false positives; when the fire points are concentrated, T1 is shortened to capture fire mutations and improve response timeliness. At the same time, the warning threshold is dynamically adjusted according to the strength of the fire: when the fire weakens, the reference prediction value is increased to avoid frequent warnings; when the fire strengthens, the reference prediction value is reduced to achieve dynamic adjustment.

[0089] In order to further improve the timeliness of mountain fire monitoring, the unmanned equipment cooperates with the optical fiber temperature measurement device, and the method further includes the following steps:

[0090] The unmanned equipment (such as a drone) carries multiple optical fiber sensing bodies and is deployed in stages according to the background instructions, which can quickly build a distributed temperature measurement point in the high-risk area of mountain fire. For example, after receiving the first deployment instruction, the drone deploys the first optical fiber sensing body 50 meters away from the center of the gathering area to form a core temperature measurement node; when the fire development degree value exceeds 1.5, the second deployment instruction is received, and the second optical fiber sensing body is deployed 200 meters away in the wind direction to build an optical fiber-based temperature measurement module.

[0091] After the first optical fiber sensing body is deployed, the first optical fiber sensing body is connected with the background data through wireless communication or wired communication, and the second optical fiber sensing body on the unmanned equipment is directly connected with the unmanned equipment through the optical fiber communication module to realize low-delay transmission of the control data of the unmanned equipment. The optical fiber line body can withstand the traction force during the flight of the unmanned aerial vehicle and has anti-electromagnetic interference capability, and can still maintain the transmission of control data around the ultra-high voltage power transmission line, which significantly improves the reliability compared with traditional wireless communication.

[0092] The first optical fiber sensing body that has been deployed can serve as a relay node for communication between the background and the unmanned equipment, and forward subsequent deployment instructions and data.

[0093] When the unmanned aerial vehicle enters a wireless signal blind area, such as the bottom of a valley, the background can forward the second deployment instruction to the to-be-deployed optical fiber sensing body through the optical fiber line body via the first deployed optical fiber sensing body, forming a relay link, and extending the communication coverage from 3 kilometers in traditional wireless communication to more than 10 kilometers.

[0094] After the second optical fiber sensing body is deployed, the second optical fiber sensing body and the unmanned equipment are automatically switched to the wireless communication module connection, and the unmanned equipment is indirectly connected with the background; the optical fiber sensing body can still continuously return data after being separated from the unmanned aerial vehicle.

[0095] Multiple optical fiber sensing bodies are connected in series through optical fiber lines, and a chain temperature measurement network can be constructed. For example, 4 or more optical fiber sensing bodies are deployed on both sides of a 4-kilometer or longer power transmission line corridor to form a linear temperature monitoring belt to capture temperature changes along the line in real time.

[0096] The optical fiber sensing bodies operate the optical fiber and wireless communication modules simultaneously to form a hot backup link, i.e., the wireless communication transmission link and the optical fiber relay transmission link are hot backups for each other.

[0097] Under normal circumstances, optical fiber communication is preferred, and when a fiber line break is detected, wireless communication is automatically switched to ensure that data is not lost.

[0098] The wireless communication module has a Beidou positioning function, and after deployment, it automatically returns the position information, and the background can real-time calibrate the temperature measurement point position.

[0099] The unmanned device can recover the optical fiber sensing body after the task is completed:

[0100] The optical fiber sensing body releases the anchoring device through wireless instructions, and the unmanned aerial vehicle takes it back to the base after grabbing.

[0101] Through the deep cooperation between the unmanned device and the optical fiber sensing body, the traditional static temperature measurement network is converted into a dynamic and reconfigurable intelligent monitoring system. The phased deployment mechanism realizes the precise layout of the temperature measurement points, and the dual communication link design guarantees the reliable transmission of data in complex environments.

[0102] In the step of the unmanned device obtaining the second deployment instruction from the background through the wireless communication module, the following sub-steps are included:

[0103] When the unmanned device and the background establish a connection through the wireless communication module, the system will monitor the wireless signal strength in real time. If the signal strength reaches or exceeds the preset reference threshold, the unmanned device will directly receive the second deployment instruction through the wireless communication module, ensuring low-latency response under ideal communication conditions.

[0104] If the wireless connection is interrupted or the signal strength is insufficient, the system will automatically start the optical fiber relay transmission link. At this time, the unmanned device re-establishes a connection with the background through the wireless temperature measurement module, and the background sends the second deployment instruction to the optical fiber sensing body that has completed the deployment and is in a wireless connection state. The deployed optical fiber sensing body will transmit the instruction to the optical fiber sensing body to be deployed in the form of relay forwarding through the optical fiber line, and finally the optical fiber sensing body to be deployed will pass the instruction to the unmanned device through the optical fiber communication module.

[0105] Through the multi-link redundancy mechanism, the transmission path is dynamically switched, and the influence of the complex environment on the instruction transmission is effectively overcome. The wireless direct connection mode guarantees efficient communication in a conventional scene, and the fiber relay link breaks through the limitation of the wireless signal coverage range and the terrain condition, and improves the stability and success rate of the instruction transmission.

[0106] The step of launching another optical fiber sensing body further includes the following sub-steps:

[0107] If the unmanned device is not connected to the background through the wireless communication module or the signal strength is less than the preset reference strength value, the unmanned device establishes a wireless connection with the optical fiber sensing body through the wireless communication module, disconnects the optical fiber communication module, and then launches another optical fiber sensing body. By switching to the wireless communication link, the control interruption of the unmanned device caused by the sudden disconnection of another optical fiber sensing body during the launching process is effectively avoided. By dynamically adjusting the communication mode, the safety of the device operation in a complex environment is ensured.

[0108] The embodiment of the present application also discloses a power transmission line mountain fire prediction device based on multi-modal data fusion, which comprises a processor, and the processor executes the steps of the power transmission line mountain fire prediction method based on multi-modal data fusion according to any one of the above.

[0109] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the above embodiments within the scope of the present application.

Claims

1. A power transmission line forest fire prediction method using multi-modal data fusion, characterized by, The method comprises the following steps: obtaining a plurality of fixed images based on a plurality of fixed cameras, the fixed cameras being located at fixed positions in different locations; and fusing the plurality of fixed images to obtain a fixed full-frame image; identifying a global forest fire feature from the fixed full-frame image, outputting the identified global forest fire feature and a corresponding global forest fire position, and calculating an aggregation degree value of the global forest fire position; calculating a generation speed value and a disappearance speed value of the global forest fire feature within a preset first time period, and calculating a development degree value based on the generation speed value and the disappearance speed value; calculating a forest fire prediction value based on the aggregation degree value and the development degree value; The calculation steps of the generation speed value, the disappearance speed value and the development degree value comprise the following sub-steps: counting the number of newly added global forest fire features within the preset first time period T1 as Q1, and calculating the generation speed value V1: V1=Q1 / T1; counting the number of reduced global forest fire features within the preset first time period T1 as Q2, and calculating the disappearance speed value V2: V2=Q2 / T1; the development degree value=V1 / V2; The calculation steps of the aggregation degree value comprise the following sub-steps: Taking the center position of the aggregation area of the global wildfire position as the center of a circle, the number Q3 of data points within a set neighborhood radius R is counted, and if the number Q3 of data points exceeds a preset number threshold, a clustering degree value is calculated: Clustering degree value = Q3 / (π×R 2 ). if the forest fire prediction value is greater than a preset reference prediction value, calculating the center position of the aggregation area of the global forest fire position, and calling a mobile camera to move to the center position; The mobile camera moves to capture a plurality of mobile area images with different positions, and the positions of the mobile area images are located within the corresponding area of the fixed full-frame image; identifying a local forest fire feature based on the mobile area image, outputting the identified local forest fire feature and a corresponding local forest fire position, and calculating a local aggregation area of the local forest fire feature based on the local forest fire position; calculating a plurality of local aggregation areas based on a plurality of mobile area images, and calculating an aggregation area movement vector based on the positions of the plurality of local aggregation areas; the direction of the aggregation area movement vector is opposite to the direction of the movement vector of the mobile camera; if the local aggregation area corresponds to the center position and the center position is on the aggregation area movement vector, a forest fire warning is issued.

2. The method for predicting forest fire on power transmission line using multi-modal data fusion as claimed in claim 1, wherein, Before comparing the forest fire prediction value, the following sub-steps are further included: searching for a fiber temperature measuring device within a preset distance range from the center position, and obtaining a detection temperature value generated by the fiber temperature measuring device; calculating a center temperature based on the detection temperature value, and calculating a temperature ratio between the center temperature and a preset reference temperature value; if the temperature ratio is greater than a preset reference ratio, the forest fire prediction value is adjusted in positive correlation with the temperature ratio.

3. The method for predicting forest fire on power transmission line using multi-modal data fusion as claimed in claim 2, wherein, The method further comprises the following steps: calculating a shortest distance value between the fiber temperature measuring device and the center position; adjusting the reference ratio in positive correlation with the shortest distance value; the smaller the shortest distance, the smaller the reference ratio; the larger the shortest distance, the larger the reference ratio; adjusting the distance range in positive correlation with the development degree value; the smaller the development degree value, the smaller the distance range; the larger the development degree value, the larger the distance range.

4. The method for predicting forest fire on power transmission line using multi-modal data fusion as claimed in claim 1 wherein, The method further comprises the following sub-steps: The duration of the first time period is inversely adjusted according to the aggregation degree value, and the smaller the aggregation degree value is, the longer the duration of the first time period is; The duration of the first time period is inversely adjusted according to the aggregation degree value, and the smaller the aggregation degree value is, the longer the duration of the first time period is; The reference prediction value is inversely adjusted according to the development degree value, and the smaller the development degree value is, the larger the reference prediction value is; the larger the development degree value is, the smaller the reference prediction value is.

5. The method for predicting forest fire on power transmission line using multi-modal data fusion as claimed in claim 2, wherein, The method further comprises the following steps: The optical fiber temperature measuring device is arranged by the unmanned device driving the mobile camera to move, the optical fiber temperature measuring device comprises an optical fiber line body and an optical fiber sensing body arranged at the end of the optical fiber line body, the optical fiber sensing body is used to generate the detection temperature value and transfer the optical fiber data on the optical fiber line body, and the optical fiber sensing body has a wireless communication module and an optical fiber communication module; The unmanned device has an optical fiber communication module and a wireless communication module connected with the optical fiber sensing body; In an initial state, the unmanned device establishes an optical fiber communication connection with the optical fiber sensing body; The unmanned device obtains a first launching instruction from the background through the wireless communication module and forwards the first launching instruction to the optical fiber sensing body; The optical fiber sensing body establishes a wireless connection with the background in response to the first launching instruction, the unmanned device launches one of the optical fiber sensing bodies, and establishes an optical fiber communication connection with another optical fiber sensing body which is not launched; The unmanned device obtains a second launching instruction from the background.

6. The method for predicting forest fire on power transmission line using multi-modal data fusion as claimed in claim 5 wherein, In the step of obtaining the second launching instruction from the background by the unmanned device through the wireless communication module, the following sub-steps are further included: If the unmanned device is connected with the background through the wireless communication module, the signal strength of the wireless connection is detected, if the signal strength is greater than or equal to a preset reference strength value, the unmanned device obtains the second launching instruction through the wireless communication module; If the unmanned device is not connected with the background through the wireless communication module or the signal strength is less than the preset reference strength value, the unmanned device establishes a connection with the background through the wireless temperature measuring module, the background sends the second launching instruction to the optical fiber sensing body which is connected wirelessly and has been launched, the optical fiber sensing body forwards the second launching instruction to the optical fiber sensing body to be launched through the optical fiber line body, and the optical fiber sensing body to be launched forwards the second launching instruction to the unmanned device through the optical fiber communication module.

7. The method for predicting forest fire on power transmission line using multi-modal data fusion as claimed in claim 6 wherein, In the step of launching another optical fiber sensing body, the following sub-steps are further included: If the unmanned device is not connected with the background through the wireless communication module or the signal strength is less than the preset reference strength value, after the unmanned device establishes a wireless connection with the optical fiber sensing body through the wireless communication module, the connection of the optical fiber communication module is disconnected, and another optical fiber sensing body is launched.

8. A power transmission line forest fire prediction device using multi-modal data fusion, characterized by, A processor is included, and the processor executes the steps of the power line mountain fire prediction method of multi-modal data fusion according to any one of claims 1-7.

Images