Forest fire control unmanned aerial vehicle inspection image processing system

Abstract

The invention relates to the technical field of fire-fighting early warning, in particular to a forest fire-fighting unmanned aerial vehicle inspection image processing system. According to the system, for an initial connected domain in an infrared image, whether fire areas in different states exist in the initial connected domain or not is judged through comparison between high-temperature points, and then the initial connected domain is segmented to obtain an accurate fire connected domain. Based on an accurate fire disaster connected domain, pixel points with more significant features are found out through classification and comparison modes, the pixel points are used as pixel points needing saliency amplification, the saliency is adjusted, and then a saliency image with errors eliminated is obtained. According to the method, the initial connected domain needing to be segmented is segmented according to the matching condition inside the initial connected domain and the edge shape outside the initial connected domain, the accurate fire area is determined, then the pixel points needing to be subjected to saliency adjustment are determined through a classification and comparison method, and the saliency image with high information reference is obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of fire warning, and in particular to a forest firefighting unmanned aerial vehicle inspection image processing system. Background Art

[0002] Forest fires are a global environmental problem, posing a significant threat to ecosystems, economies, and human safety. Preventing and responding to forest fires is crucial to mitigating their impact. However, the vast expanses and complex terrain of forests limit traditional ground-based monitoring and early warning methods. Existing technologies use drones for aerial inspections to locate fire locations and areas.

[0003] Saliency detection algorithms play an important role in forest environment monitoring collected by drones. They can help quickly locate key areas in images, such as areas of abnormal temperature concentration, potential fire sources indicated by color or texture changes, and other significant signals. However, the CA saliency detection algorithm uses contextual information to obtain regional saliency and compares it with the nearest salient area. This method of obtaining saliency is relatively simple, and it is impossible to obtain more significant saliency for certain pixels with significant features, resulting in large errors in saliency detection results. In fire scenarios, the continuous spread of fires leads to confusion in the determination of fire areas. Areas with different fire progress states may be identified as one area due to the spread of the fire. Because the area definition is unclear, errors in saliency acquisition occur when comparing saliency, which in turn affects the effectiveness of forest fire warning data. Summary of the Invention

[0004] In order to solve the technical problem that the existing technology has errors in saliency analysis of infrared images of forest fires due to factors such as unclear fire area definition, the purpose of the present invention is to provide a forest firefighting drone inspection image processing system. The technical solution adopted is as follows: The present invention proposes a forest firefighting drone inspection image processing system, the system comprising: An infrared image acquisition module is used to obtain an infrared image of a forest fire event collected by a UAV; the infrared image includes multiple initial connected domains; The fire connected domain segmentation module is configured to obtain, for each initial connected domain, the degree of edge concavity of the initial connected domain; screen out high-temperature points in the initial connected domain and combine them in pairs to obtain high-temperature point pairs; obtain the degree of matching between the pixel value variation characteristics of the two high-temperature points in the high-temperature point pairs in different directions in the initial connected domain; determine whether the initial connected domain needs to be segmented based on the matching degree and the degree of edge concavity; and for the initial connected domain that needs to be segmented, determine a segmentation line based on the difference in variation between adjacent pixel values, starting from the edge concavity point, to obtain the segmented fire connected domain. The saliency adjustment module is used to obtain different connected domain clusters based on the area of ​​the fire connected domain. In a connected domain cluster, the relationship between the pixel value difference between each pixel point and its neighboring pixels and the preset fluctuation range of the connected domain cluster to which it belongs is determined. The pixels that need to be amplified in significance are screened and the significance is adjusted to obtain an adjusted saliency image.

[0005] Furthermore, the method for obtaining the edge concavity degree includes: The curvature of each edge pixel point on the initial connected domain is obtained, the concave positions are screened out according to the curvature, and the absolute value of the average curvature of the concave positions is normalized to obtain the edge concavity.

[0006] Furthermore, the high temperature point screening method includes: In the initial connected domain, pixel points whose pixel values ​​are greater than a preset threshold are selected as the high-temperature points.

[0007] Furthermore, the method for obtaining the matching degree includes: In one of the high-temperature point pairs, for each high-temperature point, a straight line is drawn in different directions with each high-temperature point as the starting point and the end point as the edge pixel point, and the pixel values ​​on the straight line constitute a pixel value change sequence in each direction; the pixel value change sequence between the two high-temperature points is matched by obtaining sequence similarity using the DTW algorithm to obtain all matching sequence pairs, and the average value of the sequence similarities of all matching sequence pairs is used as the matching degree.

[0008] Furthermore, judging whether the initial connected domain needs to be segmented according to the matching degree and the edge concavity degree includes: For a high-temperature point pair, the overall chaos weight of each high-temperature point pair is obtained according to the volatility of the pixel value change sequence of the high-temperature point; after the matching degree is negatively correlated and normalized, it is multiplied by the overall chaos weight to obtain the non-connectivity of each high-temperature point pair; The product of the average non-connectivity of all high-temperature point pairs and the edge concavity is used as the spread degree of the initial connected domain; if the spread degree is greater than a preset spread degree threshold, it is determined that the initial connected domain needs to be segmented.

[0009] Furthermore, the method for obtaining the overall chaos weight includes: For each high temperature point in a high temperature point pair, the average variance of all pixel value change sequences is normalized to obtain the high temperature point chaos weight; the average high temperature point chaos weight of the two high temperature points in a high temperature point pair is used as the overall chaos weight of the high temperature point pair.

[0010] Furthermore, the method for obtaining the edge depression point includes: The concave position with the smallest curvature is used as the edge concave point.

[0011] Furthermore, the method for determining the dividing line includes: The edge depression point is used as the initial growth point to grow within the eight neighborhoods; the neighborhood pixel points within the neighborhood of the initial growth point are used as the first neighborhood pixel points, and the pixel points within the preset neighborhood of the first neighborhood pixel point are used as the second pixel points; in the preset neighborhood of the first neighborhood pixel point, multiple groups of symmetry directions are obtained with the first neighborhood pixel point as the symmetry point, for each direction, the pixel value difference of adjacent pixel points in each direction is obtained with the first neighborhood pixel point as the initial point, the numerical difference between the pixel value difference and the pixel value difference of the symmetrical position in the symmetry direction is obtained, and the sum of all numerical differences in a group of symmetry directions is used as the initial segmentation degree in the symmetry direction; the average initial segmentation degree of all groups of symmetry directions is used as the growth feature of the first neighborhood pixel point; the first neighborhood pixel point with the maximum growth feature within the eight neighborhoods of the initial growth point is used as a new growth point, and growth is continued based on the new growth point until it grows to the edge pixel point of the initial connected domain, and all growth points constitute the segmentation line.

[0012] Furthermore, the step of screening out pixels whose significance needs to be amplified includes: The connected domain cluster with the largest area is taken as the standard connected domain cluster, and the average value of the pixel value differences between all adjacent pixels in the standard connected domain cluster is taken as the benchmark value; For each connected domain cluster, the average area in the connected domain cluster is normalized to obtain an area weight, and the product of the area weight and the reference value is used as the fluctuation range; For each connected domain cluster, the average pixel value difference between each pixel point and all neighboring pixels in the eight neighborhoods is obtained. If the average pixel value difference is greater than the fluctuation range, the pixel point is judged to be a pixel point that needs to be amplified in significance.

[0013] Furthermore, the method for obtaining the adjusted saliency image includes: Normalizing the difference between the average pixel value difference and the fluctuation range to obtain an increase ratio; An initial saliency image of the infrared image is obtained by using a CA saliency detection algorithm. In the initial saliency image, for pixels whose saliency needs to be amplified, the initial saliency is amplified by using the amplification ratio to obtain an adjusted saliency image.

[0014] The present invention has the following beneficial effects: The present invention first considers that the spread of fire can cause fire areas of different states to merge into one fire area, resulting in errors in saliency detection. Therefore, for the initial connected domain in the infrared image, the high-temperature points are compared to determine whether there are fire areas of different states in the initial connected domain. That is, the smaller the degree of matching and the greater the degree of edge concavity, the more it indicates that the initial connected domain is formed by the fusion of fire areas of different states. Then, the initial connected domain is segmented to obtain an accurate fire connected domain. Based on the accurate fire connected domain, pixels with more significant features are found through classification and comparison. These pixels are used as pixels to be amplified and their saliency is adjusted, thereby obtaining a saliency image with errors eliminated. The present invention segments the initial connected domain to be segmented based on the matching situation inside the initial connected domain and the edge shape outside, determines the accurate fire area, and then uses classification and comparison methods to determine the pixels to be adjusted in saliency, thereby obtaining a saliency image with strong information reference. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0016] Figure 1 A block diagram of a forest firefighting drone inspection image processing system provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0017] To further illustrate the technical means and effectiveness of the present invention in achieving its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail the specific implementation, structure, features, and effectiveness of a forest firefighting drone inspection image processing system proposed in accordance with the present invention. In the following description, references to different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0018] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0019] The following describes in detail a specific solution of a forest firefighting drone inspection image processing system provided by the present invention in conjunction with the accompanying drawings.

[0020] See also Figure 1 , which shows a block diagram of a forest firefighting drone inspection image processing system provided by an embodiment of the present invention. The system includes: an infrared image acquisition module 101, a fire connectivity domain division module 102, and a saliency adjustment module 103.

[0021] It should be noted that when the drone performs a forest inspection mission in the forest inspection area, it can send the collected infrared images to the infrared image acquisition module 101 in real time. Then, using the existing connected domain detection algorithm, multiple initial connected domains on the infrared image can be obtained. The pixel values ​​in the infrared image represent the temperature values ​​in the corresponding area, that is, the higher the pixel value, the higher the temperature at the corresponding location.

[0022] During CA saliency detection, images captured by drones contain different regions with varying fire states, representing different fire phases. This results in different distribution characteristics between regions. The CA saliency detection algorithm compares regions with the nearest salient region when determining their saliency. Due to the characteristics of fire spread, regions with different fire states are considered connected domains. Comparing these connected domains can lead to significant errors in saliency detection results. Therefore, comparisons should be made only within regions of the same type. Therefore, for the initial connected domains, it is necessary to analyze their internal Wendy features to determine whether they belong to regions with different fire states, and then segment them to obtain accurate fire connected domains in the infrared image.

[0023] In the fire connected domain segmentation module 102, the same analysis method is performed on each initial connected domain. First, it is necessary to determine whether the initial connected domain requires segmentation, that is, whether it is a fusion region under different fire conditions. This embodiment of the present invention analyzes the initial connected domain from two dimensions: the internal state and the external state. Within the initial connected domain, the temperature distribution and temperature variation vary under different fire conditions. Therefore, this embodiment uses a matching method to screen high-temperature points and pair them together. The degree of matching between the pixel value variation characteristics of two high-temperature points in a pair in different directions is determined. The smaller the degree of matching, the more likely the two high-temperature points are high-temperature points under different fire conditions. The more likely the initial connected domain is a fusion region under different fire conditions, and the more likely it is to require segmentation. Regarding the exterior of the initial connected domain, region fusion results in an irregular shape of the fusion region, with noticeable concave features on the edges. Therefore, the degree of edge concave is determined. The greater the degree of edge concave, the more likely the initial connected domain is a fusion region under different fire conditions, and the more likely it is to require segmentation. Therefore, the need for segmentation of the initial connected domain can be determined based on the degree of matching and the degree of edge concavity. For initial connected domains that require segmentation, the segmentation line is determined based on the difference in the change between adjacent pixel values, starting from the edge concavity point, and the initial connected domain is split into two connected domains. It should be noted that after the initial connected domain is segmented, further analysis is still required to determine whether the segmented connected domains need to be segmented again, until all connected domains no longer need to be segmented. At this point, the resulting set of connected domains is the set of fire connected domains, and all fire connected domains after segmentation can be obtained.

[0024] Preferably, in an embodiment of the present invention, the method for obtaining the edge concavity degree includes: The curvature of each edge pixel position on the initial connected domain is obtained, the concave position is screened out according to the curvature, the absolute value of the average curvature of the concave position is normalized, and the degree of edge concavity is obtained. The curvature on the edge can capture the local set characteristics of the edge. If the curvature is positive, it means that the corresponding position is a convex position, and if the curvature is negative, it means that the corresponding position is a concave position. For the concave position, the smaller the curvature, that is, the larger the absolute value of the curvature, the more obvious the concavity. Therefore, the absolute value of the average curvature of all concave positions is normalized to obtain the degree of edge concavity. In an embodiment of the present invention, the normalization method selects range normalization, and normalization is performed by obtaining the maximum and minimum values ​​in the data dimension. The specific content is a technical means well known to those skilled in the art and will not be elaborated here.

[0025] Preferably, in an embodiment of the present invention, in the initial connected domain, pixels whose pixel values ​​are greater than a preset threshold are selected as the high-temperature points. The threshold is set to one-fourth of the average pixel value in the initial connected domain. That is, pixels greater than the threshold are considered high-temperature points with obvious high-temperature characteristics. Subsequent analysis of the high-temperature points can reduce the amount of data analysis and increase analysis efficiency.

[0026] Preferably, in one embodiment of the present invention, considering that regions under different fire conditions primarily exhibit different temperature distribution changes within the region, for example, in the middle of a fire, when the fire is intense, the entire region is at a relatively high temperature, and the temperature distribution is relatively uniform; whereas, in the final stage of a fire, due to the reduction of flammable materials, the fire gradually subsides, with only certain locations continuing to burn, maintaining high temperatures in some local locations, and other locations remaining relatively low. Based on this, in this embodiment of the present invention, the matching of pixel value changes in various directions at high-temperature points is analyzed, and the method for obtaining the matching degree includes: For each high-temperature point pair, a straight line is drawn in different directions starting from each high-temperature point and ending at an edge pixel point. The pixel values ​​along the straight line constitute a pixel value change sequence in each direction. In this embodiment of the present invention, the direction of the straight line is the eight-neighborhood direction centered on the high-temperature point. That is, starting from the high-temperature point, the straight line extends in eight directions until it reaches an edge pixel point, resulting in eight pixel value change sequences for each high-temperature point.

[0027] The pixel value change sequences between two hot spots are matched using the DTW algorithm to obtain sequence similarity. All matching sequence pairs are obtained, and the average sequence similarity of all matching sequence pairs is used as the matching degree. The smaller the matching degree between the two hot spots, the more dissimilar their temperature characteristics in different directions, and the more likely the two hot spots are located in areas with different fire conditions.

[0028] In an embodiment of the present invention, the sequence similarity is quantified using the DTW algorithm. Specifically, the sequence similarity can be the result of negative correlation mapping and normalization of the DTW distance. KM matching can be used to obtain a one-to-one pixel value change sequence matching result between two high-temperature points, thereby obtaining the matching degree.

[0029] In an embodiment of the present invention, the method of negative correlation mapping and normalization is to use the opposite number of the DTW distance as the power of an exponential function with a natural constant as the base, and the output result of the exponential function is the data after negative correlation mapping and the value range is between 0 and 1.

[0030] Furthermore, in an embodiment of the present invention, judging whether the initial connected domain needs to be segmented according to the matching degree and the edge concavity degree includes: For a high-temperature point pair, the overall chaos weight of each high-temperature point pair is obtained based on the volatility of the pixel value change sequence of the high-temperature point. The larger the overall chaos weight, the more likely the high-temperature point in the high-temperature point pair is a chaotic change point, which may be due to temperature chaos caused by the spread of the fire. The larger the overall chaos weight, the more the high-temperature point pair reflects the initial connected domain and the more it should be segmented.

[0031] The matching degree is weighted according to the overall chaos weight to obtain the non-connectivity of each high-temperature point pair. Specifically, for each high-temperature point pair, the matching degree is negatively correlated and normalized, and then multiplied by the overall chaos weight to obtain the non-connectivity of each high-temperature point pair. The larger the overall chaos weight and the smaller the matching degree, the higher the probability that the high-temperature point in the pair is a chaos change point. Furthermore, the greater the probability that the two high-temperature points are in different fire states, the greater the non-connectivity, indicating that the two high-temperature points are less likely to be in the same fire state.

[0032] It should be noted that the method for negative correlation mapping and normalization can also adopt the above-mentioned exponential function mapping method, and the details will not be repeated here.

[0033] The product of the average disconnectivity of all high-temperature point pairs and the edge concavity is used as the spread degree of the initial connected domain. Specifically, the greater the average disconnectivity and the greater the edge concavity, the more likely the initial connected domain is a fusion region of multiple fire states caused by fire spread, and the greater the spread degree. If the spread degree exceeds a preset spread degree threshold, the initial connected domain is determined to require segmentation.

[0034] In the embodiment of the present invention, the spread degree is further normalized, and the spread degree threshold is set to 0.62. The normalization method can also adopt range normalization, which will not be described in detail.

[0035] Furthermore, in an embodiment of the present invention, the volatility of the pixel value change sequence can be represented by the variance of the elements in the sequence, so the method for obtaining the overall chaos weight includes: for each high-temperature point in a high-temperature point pair, normalizing the average variance of all pixel value change sequences to obtain the high-temperature point chaos weight; and taking the average high-temperature point chaos weight of the two high-temperature points in a high-temperature point pair as the overall chaos weight of the high-temperature point pair.

[0036] Preferably, in one embodiment of the present invention, the method for obtaining the concave position based on curvature uses the concave position with the minimum curvature as the starting point for segmenting the initial connected domain.

[0037] Preferably, in one embodiment of the present invention, the method for determining the dividing line includes: Growth is performed within the eight neighborhoods using the edge depression point as the initial growth point; the neighboring pixel points within the neighborhood of the initial growth point are used as the first neighborhood pixel points, and the pixel points within the preset neighborhood of the first neighborhood pixel points are used as the second pixel points. It should be noted that the embodiment of the present invention analyzes the neighborhood of the first neighborhood pixel point in order to analyze the change in the pixel value of the first neighborhood pixel point in the symmetric direction, and therefore requires a sufficient number of data points. Therefore, in the embodiment of the present invention, the neighborhood of the first neighborhood pixel point is set to the area within the 7×7 window.

[0038] In a preset neighborhood of the first neighborhood pixel point, multiple sets of symmetry directions are obtained with the first neighborhood pixel point as a symmetry point. In an embodiment of the present invention, eight directions, namely, up, down, left, right, upper left, upper right, lower left, and lower right, are analyzed, resulting in four sets of symmetry directions: up, down, left, right, upper left, lower right, and upper right, lower left.

[0039] The purpose of analyzing the symmetric direction in the embodiment of the present invention is to take into account that the point on the dividing line should be the intersection between two different fire status areas, and the two sides of the dividing line should show different pixel value changes. Therefore, the embodiment of the present invention analyzes the pixel value changes in the symmetric direction. For each direction, the pixel value difference of the adjacent pixel points in each direction is obtained with the first neighborhood pixel point as the initial point, and the numerical difference between the pixel value difference and the pixel value difference of the symmetrical position in the symmetric direction is obtained. The sum of all the numerical differences in a group of symmetric directions is accumulated as the initial segmentation degree in the symmetric direction. The larger the numerical difference, the more asymmetrical and uniform the pixel value changes in the group of symmetric directions are, and the more likely the first neighborhood pixel point is to be a point on the dividing line.

[0040] In the embodiment of the present invention, the numerical difference is the absolute value of the difference between two pixel values. It should be noted that because pixel values ​​vary in a direction, the pixel value difference in the embodiment of the present invention is the pixel value of the previous pixel minus the pixel value of the next pixel in the corresponding direction. That is, this pixel value difference has a positive or negative sign.

[0041] A first neighborhood pixel has multiple sets of symmetry directions, so the average initial segmentation degree of all sets of symmetry directions is used as the growth feature of the first neighborhood pixel. The larger the growth feature, the more likely the first neighborhood pixel is to be a segmentation point on the segmentation line.

[0042] The first neighboring pixel with the largest growth feature within the eight neighborhoods of the initial growing point is selected as a new growing point. Growth continues based on this new growing point until it reaches the edge pixel of the initial connected domain. All these growing points form the segmentation line. The segmentation line divides the connected domain to be segmented into two connected domains. Further segmentation analysis is performed on the segmented connected domains to ultimately obtain all connected domains for fires.

[0043] After obtaining all fire connected domains, saliency adjustment can be performed. To select pixels with more significant features, saliency adjustment module 103 first classifies fire connected domains by area, obtaining different connected domain clusters. Fire connected domains within a connected domain cluster can be considered to have the same fire severity and can be analyzed uniformly.

[0044] Within a connected domain cluster, the difference between the pixel values ​​of each pixel and its neighbors is first determined. A larger difference indicates a more prominent temperature point. The relationship between this difference and the preset fluctuation range of the connected domain cluster is then determined. Specifically, the larger the difference, the more likely it is to be a pixel requiring saliency. This allows for the selection of pixels with more significant features. These pixels are typically concentrated in areas with significant features, such as the boundaries of fire zones. This further increases their saliency, resulting in an adjusted saliency image. This adjusted saliency image, which has higher saliency values ​​for pixels with significant features, facilitates subsequent fire analysis, resulting in more accurate and efficient fire analysis results.

[0045] Preferably, in an embodiment of the present invention, screening out pixels whose significance needs to be amplified includes: The connected domain cluster with the largest area is taken as the standard connected domain cluster, and the average value of the pixel value differences between all adjacent pixels in the standard connected domain cluster is taken as the benchmark value; For each connected domain cluster, the average area within the cluster is normalized to obtain an area weight. The product of this area weight and the reference value is used as the fluctuation range. This fluctuation range can be considered the overall pixel value fluctuation within the cluster. Pixels with a value difference greater than this fluctuation range are identified as having significant features.

[0046] For each connected domain cluster, the average pixel value difference between each pixel point and all neighboring pixels in the eight neighborhoods is obtained. If the average pixel value difference is greater than the fluctuation range, the pixel point is judged to be a pixel point that needs to be amplified in significance.

[0047] Furthermore, the method for obtaining the adjusted saliency image includes: Normalizing the difference between the average pixel value difference and the fluctuation range to obtain an increase ratio; An initial saliency image of the infrared image is obtained by using a CA saliency detection algorithm. In the initial saliency image, for pixels whose saliency needs to be amplified, the initial saliency is amplified by using the amplification ratio to obtain an adjusted saliency image.

[0048] In the embodiment of the present invention, since the increase ratio is a data between 0 and 1, the positive integer 1 plus the increase ratio is used as the increase multiple. The increase multiple is directly multiplied by the initial saliency to obtain the increased saliency, and then the adjusted saliency image is obtained.

[0049] In summary, the present invention determines whether there are fire areas of different states in the initial connected domain in the infrared image by comparing high-temperature points, and then segments it to obtain an accurate fire connected domain. Based on the accurate fire connected domain, through classification and comparison, the pixels with more significant features are found, and they are used as pixels that need to be amplified and the significance is adjusted, thereby obtaining a significant image with errors eliminated. The present invention segments the initial connected domain that needs to be segmented based on the matching situation inside the initial connected domain and the edge shape outside, determines the accurate fire area, and then uses classification and comparison methods to determine the pixels that need to adjust the significance, thereby obtaining a significant image with strong information reference.

[0050] It should be noted that the order in which the embodiments of the present invention are described above is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0051] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

Claims

1. A forest firefighting drone inspection image processing system, characterized in that: The system comprises: An infrared image acquisition module is used to obtain an infrared image of a forest fire event collected by a UAV; the infrared image includes multiple initial connected domains; The fire connected domain segmentation module is configured to obtain, for each initial connected domain, the degree of edge concavity of the initial connected domain; screen out high-temperature points in the initial connected domain and combine them in pairs to obtain high-temperature point pairs; obtain the degree of matching between the pixel value variation characteristics of the two high-temperature points in the high-temperature point pairs in different directions in the initial connected domain; determine whether the initial connected domain needs to be segmented based on the matching degree and the degree of edge concavity; and for the initial connected domain that needs to be segmented, determine a segmentation line based on the difference in variation between adjacent pixel values, starting from the edge concavity point, to obtain the segmented fire connected domain. The saliency adjustment module is used to obtain different connected domain clusters based on the area of ​​the fire connected domain. In a connected domain cluster, the relationship between the pixel value difference between each pixel point and its neighboring pixels and the preset fluctuation range of the connected domain cluster to which it belongs is determined. The pixels that need to be amplified in significance are screened and the significance is adjusted to obtain an adjusted saliency image.

2. A forest firefighting drone inspection image processing system according to claim 1, characterized in that: The method for obtaining the edge depression degree includes: The curvature of each edge pixel point on the initial connected domain is obtained, the concave positions are screened out according to the curvature, and the absolute value of the average curvature of the concave positions is normalized to obtain the edge concavity.

3. The forest firefighting drone inspection image processing system according to claim 1 is characterized in that: The method for screening the high temperature points includes: In the initial connected domain, pixel points whose pixel values ​​are greater than a preset threshold are selected as the high-temperature points.

4. The forest firefighting drone inspection image processing system according to claim 1 is characterized in that: The method for obtaining the matching degree includes: In one of the high-temperature point pairs, for each high-temperature point, a straight line is drawn in different directions with each high-temperature point as the starting point and the end point as the edge pixel point, and the pixel values ​​on the straight line constitute a pixel value change sequence in each direction; the pixel value change sequence between the two high-temperature points is matched by obtaining sequence similarity using the DTW algorithm to obtain all matching sequence pairs, and the average value of the sequence similarities of all matching sequence pairs is used as the matching degree.

5. The forest firefighting drone inspection image processing system according to claim 4 is characterized in that: The determining whether the initial connected domain needs to be segmented according to the matching degree and the edge concavity degree includes: For a high-temperature point pair, the overall chaos weight of each high-temperature point pair is obtained according to the volatility of the pixel value change sequence of the high-temperature point; after the matching degree is negatively correlated and normalized, it is multiplied by the overall chaos weight to obtain the non-connectivity of each high-temperature point pair; The product of the average non-connectivity of all high-temperature point pairs and the edge concavity is used as the spread degree of the initial connected domain; if the spread degree is greater than a preset spread degree threshold, it is determined that the initial connected domain needs to be segmented.

6. The forest firefighting drone inspection image processing system according to claim 5 is characterized in that: The method for obtaining the overall chaos weight includes: For each high temperature point in a high temperature point pair, the average variance of all pixel value change sequences is normalized to obtain the high temperature point chaos weight; the average high temperature point chaos weight of the two high temperature points in a high temperature point pair is used as the overall chaos weight of the high temperature point pair.

7. The forest firefighting drone inspection image processing system according to claim 2 is characterized in that: The method for obtaining the edge depression point includes: The concave position with the smallest curvature is used as the edge concave point.

8. The forest firefighting drone inspection image processing system according to claim 1 is characterized in that: The method for determining the dividing line includes: The edge depression point is used as the initial growth point to grow within the eight neighborhoods; the neighborhood pixel points within the neighborhood of the initial growth point are used as the first neighborhood pixel points, and the pixel points within the preset neighborhood of the first neighborhood pixel point are used as the second pixel points; in the preset neighborhood of the first neighborhood pixel point, multiple groups of symmetry directions are obtained with the first neighborhood pixel point as the symmetry point, for each direction, the pixel value difference of adjacent pixel points in each direction is obtained with the first neighborhood pixel point as the initial point, the numerical difference between the pixel value difference and the pixel value difference of the symmetrical position in the symmetry direction is obtained, and the sum of all numerical differences in a group of symmetry directions is used as the initial segmentation degree in the symmetry direction; the average initial segmentation degree of all groups of symmetry directions is used as the growth feature of the first neighborhood pixel point; the first neighborhood pixel point with the maximum growth feature within the eight neighborhoods of the initial growth point is used as a new growth point, and growth is continued based on the new growth point until it grows to the edge pixel point of the initial connected domain, and all growth points constitute the segmentation line.

9. The forest firefighting drone inspection image processing system according to claim 1 is characterized in that: The step of screening out pixels whose significance needs to be amplified includes: The connected domain cluster with the largest area is taken as the standard connected domain cluster, and the average value of the pixel value differences between all adjacent pixels in the standard connected domain cluster is taken as the benchmark value; For each connected domain cluster, the average area in the connected domain cluster is normalized to obtain an area weight, and the product of the area weight and the reference value is used as the fluctuation range; For each connected domain cluster, the average pixel value difference between each pixel point and all neighboring pixels in the eight neighborhoods is obtained. If the average pixel value difference is greater than the fluctuation range, the pixel point is judged to be a pixel point that needs to be amplified in significance.

10. The forest firefighting drone inspection image processing system according to claim 9, characterized in that: The method for obtaining the adjusted saliency image includes: Normalizing the difference between the average pixel value difference and the fluctuation range to obtain an increase ratio; An initial saliency image of the infrared image is obtained by using a CA saliency detection algorithm. In the initial saliency image, for pixels whose saliency needs to be amplified, the initial saliency is amplified by using the amplification ratio to obtain an adjusted saliency image.

Images