Animal migration route prediction method and device based on infrared camera monitoring data
By combining infrared camera monitoring data, fire source information, and terrain features, animal migration routes are dynamically corrected, solving the problems of prediction bias and misjudgment in the event of emergencies such as fires, and achieving efficient coordination between ecological protection and public safety.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- LUDONG UNIVERSITY
- Filing Date
- 2026-05-08
- Publication Date
- 2026-07-31
AI Technical Summary
Existing technologies are ill-equipped to predict animal migration routes and cannot adequately address the dynamic impacts of emergencies such as fires. They suffer from high misjudgment rates in fire source identification, inaccurate facility risk assessment, and an inability to dynamically adapt to changes in migration paths, leading to a disconnect between ecological protection and disaster emergency response.
By combining infrared camera monitoring data with fire source information, topographic features and animal behavior strategies, the system uses support vector machine algorithm to identify fire source types, cellular automata algorithm to predict the fire impact area, dynamically corrects migration routes, and provides real-time response through multiple types of alarm signals.
It improves the accuracy of animal migration route prediction and emergency response efficiency, accurately distinguishes between natural and man-made fires, reduces misjudgments, optimizes resource allocation, and achieves efficient synergy between ecological protection and public safety.
Smart Images
Figure CN122477950A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of animal migration route technology, specifically to a method and apparatus for predicting animal migration routes based on infrared camera monitoring data. Background Technology
[0002] In the field of animal ecological protection and disaster emergency response, existing technologies mostly rely on static historical data or single environmental parameters to predict animal migration routes, making it difficult to cope with the dynamic impact of emergencies such as fires. Traditional methods have the following limitations: First, fire monitoring and animal behavior analysis are disconnected, failing to integrate fire source spread trends and species avoidance rules in real time, leading to predicted paths deviating from reality and misleading protected area management; Second, fire source type identification relies on manual judgment or simple spectral thresholds, resulting in a high misjudgment rate (e.g., mistaking agricultural burning for natural fire), leading to unnecessary emergency responses and interfering with animal activities; Third, facility risk assessment uses fixed geographic fences, which cannot dynamically adapt to changes in migration routes, often resulting in missed or false alarms.
[0003] The existing technology, disclosed in CN117351413A, presents a method for predicting animal migration routes based on infrared camera monitoring data. This method includes: dividing the target monitoring area into multiple monitoring grids and configuring an infrared camera in each grid; acquiring image data captured by the infrared cameras; obtaining the animal species and their corresponding numbers monitored by the grid based on the image data; identifying high-density grids for the target animal species based on the animal species and their corresponding numbers monitored by the grids; merging adjacent high-density grids into multiple high-density regions for the target animal species; and finally obtaining the migration route of the target animal. While this method can predict animal migration routes, it relies solely on the distribution of animal numbers monitored by infrared cameras, without integrating dynamic environmental parameters such as fire conditions and topography. This makes it unable to cope with sudden disturbances such as fires. Furthermore, it generates fixed migration routes by clustering historical data, lacking a dynamic correction mechanism that couples animal behavior rules with the environment. It also lacks a risk warning mechanism, leading to a disconnect between ecological protection and disaster emergency response.
[0004] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention
[0005] The purpose of this invention is to provide a method and apparatus for predicting animal migration routes based on infrared camera monitoring data, so as to solve the problems mentioned in the background art.
[0006] To achieve the above objectives, the present invention provides the following technical solution: The method for predicting animal migration routes based on infrared camera monitoring data includes the following steps: S1: Set up monitoring points based on the historical migration routes of animals, and collect data on animal activity, fire sources and geomorphological features at the monitoring points; S2: Based on the fire source information, the fire source type is identified using a preset algorithm. If the fire source type is a natural fire, proceed to step S3. If the fire source type is a man-made fire, generate a type of alarm signal and confirm it on-site. Then, based on the confirmation result, attach different levels of risk labels to the type of alarm signal. S3: For natural fires, based on the topographic features at the monitoring point, a preset fire spread model is used to predict the fire-affected area, and a second-class alarm signal is generated when the area of the fire-affected area exceeds the threshold. S4: Based on the animal species behavior strategy database and the fire-affected area, the historical migration routes are corrected to generate predicted migration routes; S5: Visualize the predicted migration route by combining it with a digital map, identify the affected facilities based on the predicted migration route and the digital map, and send three types of alarm signals to the affected facilities.
[0007] Preferably, in step S1: Animal activity data includes animal location, animal category, and direction of movement; Fire source information includes the coordinates of the fire source center, characteristic spectrum, and smoke concentration; Landform features include slope and vegetation type.
[0008] Preferably, in step S2, the algorithm for identifying the type of fire source is a support vector machine algorithm, whose input features are the difference in CO2 absorption peak intensity and ash reflectance in the characteristic spectrum of the fire source information, and the output result is natural fire or man-made fire. The logic for generating and sending the aforementioned alarm signal is as follows: When the fire source is man-made, an alarm signal is generated, and the motion unit is driven to the center coordinates of the fire source and real-time images of the scene are collected. The system performs target recognition based on a pre-set YOLOv5 model, collects feature information of personnel in the scene images, and detects whether weapons are present in the scene images. If no weapons are found in the scene image, the characteristic information of the personnel in the scene image and a type of alarm signal are sent to the receiving end together, and the risk label attached to the type of alarm signal is low risk. If weapons are present in the scene images, the characteristic information of the personnel in the scene images, the weapon type, and a type of alarm signal will be sent to the receiving end together, and the risk label attached to the type of alarm signal will be high risk.
[0009] Preferably, the fire spread model is a cellular automata algorithm, and its working logic is as follows: Based on the coordinates of the fire source center and the preset monitoring radius, the monitoring area is marked on the digital map; The monitoring area is divided into equidistant two-dimensional grid units, each of which is called a cell. Its state is marked as burned or unburned. In the initial state, the cell where the fire source is located is activated as a burning state, while the other cells are in an unburned state. After a preset time step, all cells in the state of burning are traversed. The burning cell is used as the center point, and the fire spread rate of other cells in the 8 neighboring directions is calculated. If the fire spread rate of other cells in the 8 directions exceeds the preset threshold, the cell is re-marked as burning from the unburned state. Real-time statistics are collected on the actual areas corresponding to all cells in the "burned" state on the digital map, and these areas are divided into fire-affected zones.
[0010] Preferably, the fire spread rate is calculated as follows: In the formula Indicates grid number as The rate at which the fire spreads within the cells, subscript These represent the grid indices of the cell in the horizontal and vertical directions, respectively. Indicates the initial spread rate. , Indicates grid number as The slope of the cell, Indicates the flammability coefficient of vegetation. , Indicates the wind speed influence coefficient. Indicates the maximum spread rate, when hour, , This represents the steep slope attenuation coefficient.
[0011] Preferably, the wind speed influence coefficient is calculated as follows: In the formula Indicates the volatility coefficient. , Indicates real-time wind speed; The steep slope attenuation coefficient is calculated as follows: In the formula Indicates the initial attenuation coefficient. , , These represent the grid numbers respectively. The vegetation cover area and total area of the cells.
[0012] Preferably, in step S4, the logic for correcting the historical migration route is as follows: Non-parametric probability density estimation is performed on historical migration routes to generate a basic path heatmap. The calculation method is as follows: In the formula This indicates the coordinates on the digital map. The probability density at that location, Represents the kernel function. Indicates the first on the historical migration route The coordinates of the actual observed animal activity points, with subscripts. Indices representing the indexes of actually observed animal activity points. , This represents the total number of animal activity sites actually observed. Indicates bandwidth parameter, Indicates the norm within the parentheses; The probability density is corrected using a path weight fusion model, calculated as follows: In the formula This indicates the coordinates on the digital map. The corrected probability density at that point Indicates the fire impact coefficient. Indicates the risk aversion direction coefficient; The modified probability density is discretized into grid cells. Using the real-time animal location as the starting point and the destination along the historical migration route as the ending point, the A* algorithm is used to search for the minimum-cost path as the predicted migration route, with the cost function being... .
[0013] Preferably, the method for calculating the risk avoidance direction coefficient is as follows: In the formula This indicates the coordinates on the digital map. The shortest distance between the point and the boundary of the cell in the burned state. This indicates the coordinates on the digital map. The smoke concentration at that location Indicates the species tolerance threshold; The calculation method for the fire impact coefficient is as follows: In the formula This indicates taking the minimum value within the parentheses.
[0014] An animal migration route prediction device based on infrared camera monitoring data, characterized in that: the prediction device is used to execute the above-mentioned prediction method, specifically including: The data acquisition module is used to collect animal activity data, fire source information, and geomorphological features at the monitoring point; The fire analysis module is equipped with a support vector machine algorithm to identify the type of fire source and generate an alarm signal based on the identification result, and attach risk labels of different levels. The fire prediction module is equipped with a cellular automata algorithm to predict the fire-affected area and generate a second-class alarm signal when the area of the fire-affected area exceeds a threshold. The path correction module contains historical migration data from previous years, which is used to correct historical migration routes and generate predicted migration routes. The visualization module is used to combine the predicted migration route with the digital map for visualization, identify the affected facilities based on the predicted migration route and the digital map, and send three types of alarm signals to the affected facilities. An early warning communication module is used to send various alarm signals.
[0015] Compared with the prior art, the beneficial effects of the present invention are: This invention improves the accuracy of animal migration route prediction and emergency response efficiency through multi-source data fusion and dynamic model collaboration. First, it integrates infrared monitoring, fire spread models, and a species behavior strategy database to quantitatively model animal avoidance behaviors in fire environments, ensuring predicted paths conform to historical patterns while responding to real-time threats. Second, based on spectral characteristics and on-site verification of fire source classification mechanisms, it accurately distinguishes between natural and man-made fires, reducing ecological disturbances caused by misjudgments. Finally, through the dynamic overlay of migration paths and digital maps, it automatically identifies affected facilities and triggers tiered alarms, optimizing resource allocation. This overcomes the bottlenecks of traditional methods in dynamic environmental adaptability, species behavior modeling, and human-wildlife conflict mitigation, achieving efficient synergy between ecological protection and public safety. Attached Figure Description
[0016] Figure 1 This is a schematic diagram of the overall method flow of the present invention; Figure 2 This is a schematic diagram of the module structure of the present invention; Figure 3 This is the basic path heatmap in this invention. Detailed Implementation
[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.
[0018] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.
[0019] Example: Please see Figures 1-3 The present invention provides a technical solution: The method for predicting animal migration routes based on infrared camera monitoring data includes the following steps: S1: Set up monitoring points based on the historical migration routes of animals, and collect data on animal activity, fire sources, and geomorphological features at the monitoring points.
[0020] In step S1: Animal activity data includes animal location, animal category, and direction of movement; Fire source information includes the coordinates of the fire source center, characteristic spectrum, and smoke concentration; Landform features include slope, vegetation type, and further, the distribution of water systems.
[0021] It is understandable that animal migrations are typically seasonal and periodic, such as the classic southward migration of geese and the migration of salmon, which are predictable. Specifically, historical migration routes of animals can be obtained by collecting data from previous years' measurements. The method for setting up monitoring points based on historical migration routes can be determined through expert experience or by using clustering algorithms (such as DBSCAN) to identify key path nodes, ensuring coverage of the core areas of animal activity. Data on animal activity, fire source information, and geomorphological features can be collected in the following ways: Animal activity data was extracted from infrared images using a YOLOv5 model for target detection. The model input was a sequence of infrared video frames, and the output included animal location, category, and direction of movement. Fire source information was acquired using a multispectral sensor, which collected the characteristic spectrum of flames in the 600-1000nm band and combined it with temperature gradient distribution to locate the coordinates of the fire source center. Topographic features were extracted using a digital elevation model (DEM) and remote sensing imagery. Slope data was calculated from the elevation difference of the DEM, and vegetation type was matched with a pre-trained classification model based on the NDVI index. The infrared camera can directly acquire infrared images. By integrating a multispectral sensor, remote sensing camera, and smoke sensor, fire source information, remote sensing imagery, and smoke concentration can be collected. Since these algorithms and devices are mature existing technologies, they will not be elaborated upon here.
[0022] In this step, by integrating three types of heterogeneous data—animal activity data, fire source information, and geomorphological features—a comprehensive monitoring network covering "animal behavior, fire dynamics, and environmental characteristics" was constructed. This multi-source data fusion strategy overcomes the limitations of traditional methods that rely solely on a single type of data (such as animal trajectories or fire source monitoring), providing richer input features for subsequent models.
[0023] S2: Based on the fire source information, the fire source type is identified using a preset algorithm. If the fire source type is a natural fire, proceed to step S3. If the fire source type is a man-made fire, generate a type of alarm signal and confirm it on-site. Then, based on the confirmation result, attach different levels of risk labels to the type of alarm signal.
[0024] In step S2, the algorithm for identifying the fire source type uses a Support Vector Machine (SVM) algorithm. Its input features are the CO2 absorption peak intensity and ash reflectance difference in the characteristic spectrum of the fire source information, and the output is either a natural fire or an arson fire. SVM performs well in scenarios with small samples and high-dimensional data, making it suitable for scenarios with limited data in field monitoring. Here, the CO2 absorption peak intensity reflects the scale of combustion. For example, natural fires are usually caused by lightning strikes or volcanic activity, resulting in a large burning area and long duration, with a higher CO2 peak and more stable duration. The ash reflectance difference reflects the difference in ash between natural and arson fires. Specifically, natural fire ash contains more minerals, and its reflectance in the 2.1-2.3 μm band is lower than that of arson fires. Therefore, the fire source type can be identified based on these two sets of parameters: CO2 absorption peak intensity and ash reflectance difference, thus outputting a binary classification label, namely, natural fire / arson fire.
[0025] The logic for generating and sending one type of alarm signal is as follows: When the fire source is man-made, an alarm signal is generated, and the motion unit is driven to the center coordinates of the fire source and real-time images of the scene are collected. The system performs target recognition based on a pre-set YOLOv5 model, collects feature information of personnel in the scene images, and detects whether weapons are present in the scene images. If no weapons are found in the scene image, the characteristic information of the personnel in the scene image and a type of alarm signal are sent to the receiving end together, and the risk label attached to the type of alarm signal is low risk. If weapons are present in the scene images, the characteristic information of the personnel in the scene images, the weapon type, and a type of alarm signal will be sent to the receiving end together, and the risk label attached to the type of alarm signal will be high risk.
[0026] Specifically, the driving unit here could be a drone equipped with a camera, and the personnel's characteristics could include clothing color, body type, and mode of transportation; weapons could include knives and firearms. Understandably, research on animal migration processes needs to follow natural laws and minimize the impact of human factors. By classifying fire source types, one type of alarm signal can be used to reflect human factors in animal migration routes, thus requiring separate monitoring to reduce the impact of human factors on natural laws. This also reduces misjudgment of fire source types, which could interfere with animal behavior. For example, during the wildebeest migration in the African savanna, natural fires caused by lightning strikes during the rainy season are often misjudged as arson, leading to frequent patrols by protected area teams, wasting manpower and interfering with animal behavior.
[0027] In this step, the triple mechanism of spectral feature fusion classification, on-site image dynamic verification, and risk-level response solves the problems of high misjudgment rate of fire source type and lagging response to human interference in traditional technologies. It provides highly reliable input conditions for predicting animal migration routes. At the same time, closed-loop control realizes the synergistic optimization of ecological protection and public safety, integrating "fire identification - human interference control - path correction" into a unified logical chain, which significantly improves the scientificity and practicality of the system.
[0028] S3: Based on the topographic features and fire source type at the monitoring point, the fire spread model is used to predict the fire-affected area, and a second-class alarm signal is generated when the area of the fire-affected area exceeds the threshold.
[0029] The fire spread model uses a cellular automata algorithm, and its working logic is as follows: Based on the coordinates of the fire source center and the preset monitoring radius, the monitoring area is marked on the digital map; The monitoring area is divided into equidistant two-dimensional grid units, each of which is called a cell. Its state is marked as burned or unburned. In the initial state, the cell where the fire source is located is activated as a burning state, while the other cells are in an unburned state. After a preset time step, all cells in the state of burning are traversed. The burning cell is used as the center point, and the fire spread rate of other cells in the 8 neighboring directions is calculated. If the fire spread rate of other cells in the 8 directions exceeds the preset threshold, the cell is re-marked as burning from the unburned state. Real-time statistics are collected on the actual areas corresponding to all cells in the "burned" state on the digital map, and these areas are divided into fire-affected zones.
[0030] The method for calculating the spread rate of a fire source is as follows: In the formula Indicates grid number as The rate at which the fire spreads within the cells, subscript These represent the grid indices of the cell in the horizontal and vertical directions, respectively. Indicates the initial spread rate. , Indicates grid number as The slope of the cell, Indicates the flammability coefficient of vegetation. , Indicates the wind speed influence coefficient. Indicates the maximum spread rate, when hour, That is, when the slope is 45°, the fire spreads faster, reaching its maximum spread rate. ; This represents the slope attenuation coefficient. The initial spread rate is determined based on the geographical environment; for example, it is set to a lower 0.5 for grasslands and a higher 1.2 for forests. The specific value can be determined and adjusted based on expert experience. The vegetation flammability coefficient is determined based on the vegetation type within different cells. For example, pine forests, rich in oils, have a flammability coefficient of 1.5; grasslands have a coefficient of 1.0; and shrubs, with higher water content, have a coefficient reduced to 0.8. The wind speed effect is used to simulate the promoting effect of wind on the combustion process; that is, high-speed airflow accelerates oxygen transport and expands the spark splash range, causing the fire boundary to continuously expand.
[0031] Classical fire spread theories, such as the Rothermel model, indicate that in areas with steep slopes, the fire spread rate is significantly increased due to accelerated natural airflow and the rolling effect of burning material. Low-lying or flat areas follow a basic spread pattern. Between 0° and 45° slopes, the fire spread rate is positively correlated with the slope, exhibiting an approximately linear relationship, with each degree of slope contributing about 1 / 45 to the rate. Slopes exceeding 45° may cause burning material to roll off or restrict oxygen supply, resulting in a slower or even decreased fire spread rate. Therefore, the calculation of fire spread rate uses a piecewise function to more accurately match different terrains and reflect the impact of terrain on fire development.
[0032] The wind speed influence coefficient is calculated as follows: In the formula Indicates the volatility coefficient. , Indicates real-time wind speed; The steep slope attenuation coefficient is calculated as follows: In the formula Indicates the initial attenuation coefficient. , , These represent the grid numbers respectively. The vegetation cover area and total area of the cells. It can be seen that the higher the vegetation cover rate on steep slopes, the more combustible material there is, and the slower the decay.
[0033] The generation logic of the second-class alarm signal shows that when the area affected by a natural fire does not exceed the threshold, no human intervention is required, and the natural laws are followed; however, when the area affected by a natural fire is too large, human intervention is required to prevent it from developing further and causing severe damage to the natural environment.
[0034] In this step, by introducing a physical model that couples terrain, vegetation, and wind speed, the impact of terrain, vegetation, weather, and human intervention on fire can be accurately quantified. This not only improves the accuracy of fire prediction but also provides real-time fire threat range for the dynamic correction of animal migration routes. At the same time, it triggers a graded emergency response, taking into account both ecological protection and public safety needs.
[0035] S4: Based on the animal species behavior strategy database and the fire-affected area, historical migration routes are corrected to generate predicted migration routes.
[0036] The animal species behavior strategy database here is actually a structured database used to store the behavioral patterns, risk avoidance responses, and migration preferences of different species under specific environmental conditions. Its core function is to provide behavioral parameters for migration route correction, making predicted paths more consistent with the actual decision-making logic of animals. Specifically, it can be implemented using knowledge graphs, constructing a species behavior ontology, defining entity relationships such as "fire avoidance behavior" and "slope preference," and using SPARQL queries to infer path preferences. This can include basic species profiles: species classification (e.g., herbivores, birds of prey), migration cycles, habitat preferences (water-dependent, mountain-adapted, etc.), physiological parameters (maximum movement speed, group activity size); environmental response rules: risk avoidance distance thresholds (e.g., elephants avoid fires ≥2km, wolves ≥500m), smoke sensitivity, path selection tendencies (e.g., deer prefer gentle slopes, birds avoid strong wind areas); and historical behavioral data: records of past migration routes where animals avoided fires, predators, or extreme weather. This data can be obtained through various methods such as expert experience, research literature, and historical measured data. An example combining these is shown below: African elephants: Fire avoidance radius of 2.5km, moving speed of 5-6km / h, smoke concentration sensitivity threshold C0=200ppm, and prefer paths close to water.
[0037] In step S4, the logic for correcting the historical migration route is as follows: Non-parametric probability density estimation is performed on historical migration routes to generate a basic path heatmap. The calculation method is as follows: In the formula This indicates the coordinates on the digital map. The probability density at that location, The kernel function can be either the Gaussian kernel or the Epanechnikov kernel, and the choice can be made based on expert experience. Indicates the first on the historical migration route The coordinates of the actual observed animal activity points, with subscripts. Indices representing the indexes of actually observed animal activity points. , This represents the total number of animal activity sites actually observed. The bandwidth parameter can be calculated using an empirical formula: first calculate the standard deviation of the animal's activity point coordinates, then multiply by 1.06 and... Alternatively, you can obtain a fixed value directly based on expert experience. The parentheses represent the norm. In the basic path heatmap, high-probability areas correspond to migration paths frequently used by animals, while low-probability areas represent less active regions.
[0038] In this embodiment, 19 sets of actually observed animal activity points along the historical migration route were collected, and the corresponding coordinates were marked on a digital map. The coordinate information is shown in Table 1 below: Table 1: Coordinates of Animal Activity Points Non-parametric probability density estimation was performed on these 19 groups of actually observed animal activity points, and the resulting basic path heatmap is shown below. Figure 3 As shown in the figure, the horizontal and vertical axes of this graph represent the two-dimensional planar coordinates of the actual protected area in the digital map. The heatmap visually displays the spatial "heat" of animal activity points, facilitating dynamic adjustments and comparisons for subsequent migration path predictions. It's understandable that the actual protected area is divided into grids when generating the heatmap; the finer the grid, the more accurate the heatmap. Therefore, in the heatmap, the probability density of some areas may show a value exceeding 1. This is because multiple prediction points may overlap within that grid, causing a localized overestimation of the peak value. Increasing the grid density and resolution will eliminate this phenomenon.
[0039] The probability density is corrected using a path weight fusion model, calculated as follows: In the formula This indicates the coordinates on the digital map. The corrected probability density at that point Indicates the fire impact coefficient. Indicates the risk aversion direction coefficient; The modified probability density is discretized into grid cells. Using the real-time animal location as the starting point and the destination along the historical migration route as the ending point, the A* algorithm is used to search for the minimum-cost path as the predicted migration route, with the cost function being... .
[0040] The calculation method for the hedging direction coefficient is as follows: In the formula This indicates the coordinates on the digital map. The shortest distance between the point and the boundary of the cell in the burned state. This indicates the coordinates on the digital map. The smoke concentration at that location The tolerance threshold for a species can be obtained from a database of animal species behavioral strategies.
[0041] The calculation method for the fire impact coefficient is as follows: In the formula This indicates taking the minimum value within the parentheses.
[0042] Furthermore, the path weight fusion model can be trained and optimized based on historical migration routes containing fire events. It generates predicted migration routes based on data detected in these historical routes, then compares these predicted routes with the actual historical migration routes to calculate the intersection-union ratio (IUU). The calculation method is as follows: here Indicates intersection, union, and ratio. , These represent the trajectory areas of the predicted migration route and the historical migration route, respectively. The model is considered effective when the intersection-union ratio (IU) is ≥70%. For cases with an IU <70%, root cause analysis is performed. If the IU is caused by unlabeled human activity areas, the geomorphological database is updated and the model is retrained. If the IU is caused by animal behavioral biases, the avoidance rules in the animal species behavior strategy database are expanded and updated.
[0043] In this step, through multi-source data fusion, dynamic path correction, and closed-loop optimization mechanisms, the pain points of traditional animal migration prediction, such as poor environmental adaptability, lack of species rules, and delayed verification, are solved. Ecological laws are transformed into computable engineering models, providing high-precision and high-response decision support for animal protection and facility protection in emergency scenarios such as fires. This reduces prediction bias caused by emergencies such as fires, ensures that migration routes are consistent with actual animal behavior, and avoids misleading reserve management decisions.
[0044] S5: Visualize the predicted migration route by combining it with a digital map, identify the affected facilities based on the predicted migration route and the digital map, and send three types of alarm signals to the affected facilities.
[0045] These three types of warning signals reflect the impact of the predicted migration route on man-made facilities, such as villages and railways. It is necessary to mark the areas where the predicted migration route intersects with man-made facilities as affected facilities, and issue warning signals to these facilities in advance to reduce property damage.
[0046] In this step, by introducing dynamic facility calibration, a tiered alarm mechanism, and visual decision support, complex forecast data is transformed into actionable protection instructions, ensuring efficient coordination between ecological protection and human facility safety. This solves the problems of "high false alarm rate, delayed response, and inefficient resource allocation" in traditional animal migration early warning systems, thereby improving system usability and user satisfaction.
[0047] An animal migration route prediction device based on infrared camera monitoring data, the prediction device is used to execute the above prediction method, specifically including: The data acquisition module is used to collect animal activity data, fire source information, and geomorphological features at the monitoring point; The fire analysis module is equipped with a support vector machine algorithm to identify the type of fire source and generate an alarm signal based on the identification result, and attach risk labels of different levels. The fire prediction module is equipped with a cellular automata algorithm to predict the fire-affected area and generate a second-class alarm signal when the area of the fire-affected area exceeds a threshold. The path correction module contains historical migration data from previous years, which is used to correct historical migration routes and generate predicted migration routes. The visualization module is used to combine the predicted migration route with the digital map for visualization, identify the affected facilities based on the predicted migration route and the digital map, and send three types of alarm signals to the affected facilities. An early warning communication module is used to send various alarm signals.
[0048] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.
[0049] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.
[0050] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.
[0051] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.
Claims
1. A method for predicting the migration route of an animal based on infrared camera monitoring data, characterized by, The specific steps include: S1: Set up monitoring points based on the historical migration routes of animals, and collect data on animal activity, fire sources and geomorphological features at the monitoring points; S2: Based on the fire source information, the fire source type is identified using a preset algorithm. If the fire source type is a natural fire, proceed to step S3. If the fire source type is a man-made fire, generate a type of alarm signal and confirm it on-site. Then, based on the confirmation result, attach different levels of risk labels to the type of alarm signal. S3: For natural fires, based on the topographic features at the monitoring point, a preset fire spread model is used to predict the fire-affected area, and a second-class alarm signal is generated when the area of the fire-affected area exceeds the threshold. S4: Based on the animal species behavior strategy database and the fire-affected area, the historical migration routes are corrected to generate predicted migration routes; S5: Visualize the predicted migration route by combining it with a digital map, identify the affected facilities based on the predicted migration route and the digital map, and send three types of alarm signals to the affected facilities.
2. The animal migration route prediction method based on infrared camera monitoring data according to claim 1, characterized in that: In step S1: Animal activity data includes animal location, animal category, and direction of movement; Fire source information includes the coordinates of the fire source center, characteristic spectrum, and smoke concentration; Landform features include slope and vegetation type.
3. The animal migration route prediction method based on infrared camera monitoring data according to claim 2, characterized in that: In step S2, the algorithm for identifying the type of fire source is a support vector machine algorithm. Its input features are the difference in CO2 absorption peak intensity and ash reflectance in the characteristic spectrum of the fire source information, and the output result is natural fire or man-made fire. The logic for generating and sending the aforementioned alarm signal is as follows: When the fire source is man-made, an alarm signal is generated, and the motion unit is driven to the center coordinates of the fire source and real-time images of the scene are collected. The system performs target recognition based on a pre-set YOLOv5 model, collects feature information of personnel in the scene images, and detects whether weapons are present in the scene images. If no weapons are found in the scene image, the characteristic information of the personnel in the scene image and a type of alarm signal are sent to the receiving end together, and the risk label attached to the type of alarm signal is low risk. If weapons are present in the scene images, the characteristic information of the personnel in the scene images, the weapon type, and a type of alarm signal will be sent to the receiving end together, and the risk label attached to the type of alarm signal will be high risk.
4. The method for predicting animal migration routes based on infrared camera monitoring data according to claim 2, characterized in that: The fire spread model is a cellular automata algorithm, and its working logic is as follows: Based on the coordinates of the fire source center and the preset monitoring radius, the monitoring area is marked on the digital map; The monitoring area is divided into equidistant two-dimensional grid units, each of which is called a cell. Its state is marked as burned or unburned. In the initial state, the cell where the fire source is located is activated as a burning state, while the other cells are in an unburned state. After a preset time step, all cells in the state of burning are traversed. The burning cell is used as the center point, and the fire spread rate of other cells in the 8 neighboring directions is calculated. If the fire spread rate of other cells in the 8 directions exceeds the preset threshold, the cell is re-marked as burning from the unburned state. Real-time statistics are collected on the actual areas corresponding to all cells in the "burned" state on the digital map, and these areas are divided into fire-affected zones.
5. The method for predicting animal migration routes based on infrared camera monitoring data according to claim 4, characterized in that: The method for calculating the fire spread rate is as follows: In the formula Indicates grid number as The rate at which the fire spreads within the cells, subscript These represent the grid indices of the cell in the horizontal and vertical directions, respectively. Indicates the initial spread rate. , Indicates grid number as The slope of the cell, Indicates the flammability coefficient of vegetation. , Indicates the wind speed influence coefficient. Indicates the maximum spread rate, when hour, , This represents the steep slope attenuation coefficient.
6. The method for predicting animal migration routes based on infrared camera monitoring data according to claim 5, characterized in that: The wind speed influence coefficient is calculated as follows: In the formula Indicates the volatility coefficient. , Indicates real-time wind speed; The steep slope attenuation coefficient is calculated as follows: In the formula This represents the initial attenuation coefficient. , , These represent the grid numbers respectively. The vegetation cover area and total area of the cells.
7. The method for predicting animal migration routes based on infrared camera monitoring data according to claim 5, characterized in that: In step S4, the logic for correcting the historical migration route is as follows: Non-parametric probability density estimation is performed on historical migration routes to generate a basic path heatmap. The calculation method is as follows: In the formula This indicates the coordinates on the digital map. The probability density at that location, Represents the kernel function. Indicates the first on the historical migration route The coordinates of the actual observed animal activity points, with subscripts. Indices representing the indexes of actually observed animal activity points. , This represents the total number of animal activity sites actually observed. Indicates bandwidth parameter, Indicates the norm within the parentheses; The probability density is corrected using a path weight fusion model, calculated as follows: In the formula This indicates the coordinates on the digital map. The corrected probability density at that point Indicates the fire impact coefficient. Indicates the risk aversion direction coefficient; The modified probability density is discretized into grid cells. Using the real-time animal location as the starting point and the destination along the historical migration route as the ending point, the A* algorithm is used to search for the minimum-cost path as the predicted migration route, with the cost function being... .
8. The method for predicting animal migration routes based on infrared camera monitoring data according to claim 7, characterized in that: The calculation method for the risk avoidance direction coefficient is as follows: In the formula This indicates the coordinates on the digital map. The shortest distance between the point and the boundary of the cell in the burned state. This indicates the coordinates on the digital map. The smoke concentration at that location Indicates the species tolerance threshold; The calculation method for the fire impact coefficient is as follows: In the formula This indicates taking the minimum value within the parentheses.
9. An animal migration route prediction device based on infrared camera monitoring data, characterized in that: The prediction device is used to perform the prediction method according to any one of claims 1-8, specifically including: The data acquisition module is used to collect animal activity data, fire source information, and geomorphological features at the monitoring point; The fire analysis module is equipped with a support vector machine algorithm to identify the type of fire source and generate an alarm signal based on the identification result, and attach risk labels of different levels. The fire prediction module is equipped with a cellular automata algorithm to predict the fire-affected area and generate a second-class alarm signal when the area of the fire-affected area exceeds a threshold. The path correction module contains historical migration data from previous years, which is used to correct historical migration routes and generate predicted migration routes. The visualization module is used to combine the predicted migration route with the digital map for visualization, identify the affected facilities based on the predicted migration route and the digital map, and send three types of alarm signals to the affected facilities. An early warning communication module is used to send various alarm signals.