Image processing method and system for aerial photography

Abstract

The present application relates to the field of image processing technology for aerial photography, and provides an image processing method and system for aerial photography, including identifying a slope area in each aerial photographed image; identifying a target area in the slope area where a landslide is about to occur according to preset landslide features; controlling the aerial photography equipment to hover or hover over the target area and take tracking images for a preset time; detecting the difference in width, length and area of ​​the landslide features in a plurality of consecutive tracking images, and if the difference is within a second preset range, controlling the aerial photography equipment to continue taking pictures along a preset flight trajectory; if the difference is not within the second preset range, it is determined that there is a landslide risk in the target area, and the image data of the target area and its corresponding coordinates are sent to a remote terminal. The identification and dynamic monitoring of landslides are realized to improve the timeliness of early warning of landslides.

Description

Technical Field

[0001] The present application relates to the technical field of image processing of aerial photography, and in particular, to an image processing method and system for aerial photography. Background Art

[0002] Landslides, as a natural phenomenon that cannot be ignored in the field of geology, are characterized by the sudden collapse, sliding and even rolling of rock and soil or debris on mountain slopes driven by gravity. This process is not only rapid and destructive, but also often causes serious casualties and property losses. Although scientific and technological progress has prompted humans to conduct extensive and in-depth research and application of the causes, mechanisms, monitoring methods and preventive measures of landslides, the existing technology system still shows many limitations in the face of this complex and changeable natural disaster.

[0003] Specifically, high-precision and non-contact InSAR technology and three-dimensional laser scanning technology have shown extraordinary potential in capturing subtle deformations on the mountain surface, providing important technical support for landslide early warning. However, high equipment costs, low operating efficiency and complex data processing processes have become bottlenecks for the rapid response and widespread popularization of these technologies. In addition, the resolution and timeliness of remote sensing technology in monitoring are also facing challenges, especially in areas with dense clouds or dense vegetation coverage, where its monitoring efficiency is significantly limited and it is difficult to meet the needs of refined and real-time monitoring. The natural potential method, as a prediction method based on geophysical principles, predicts landslide activities by monitoring the changes in the internal potential of the landslide body, showing a certain forward-looking prediction ability. However, the implementation of this method is highly dependent on natural factors such as geological conditions and groundwater distribution, and equipment needs to be installed at specific locations, which not only limits its scope of application, especially its popularization in remote or difficult-to-reach areas, but also increases the uncertainty of the prediction results.

[0004] On the other hand, close-up photography is used to detect landslides, but this method can only detect landslides in a local field of view or in ordinary situations. After close-up photography is damaged or an earthquake occurs, the close-up photography equipment cannot detect landslides.

[0005] Therefore, there is an urgent need for an image processing method for aerial photography to realize the identification and dynamic monitoring of landslides without being constrained by the ground environment, so as to improve the timeliness of landslide warning. Summary of the invention

[0006] In view of this, the purpose of this application is to provide an aerial photography image processing method and system, which can complete the identification and dynamic monitoring of landslides through image processing, thereby improving the timeliness of landslide warning.

[0007] The purpose of this application is achieved through the following technical solutions:

[0008] An image processing method for aerial photography, comprising:

[0009] Obtain the initial map data of the target area from the map data platform; extract the landmark features of the mountain and the corresponding height change trend based on the contour lines of the initial map data;

[0010] Controlling the aerial photography equipment with positioning to continuously photograph the above-mentioned target area along a preset flight trajectory to obtain multiple photographed images;

[0011] Matching a local image having a similarity with the landmark feature greater than a first threshold value from each of the captured images, taking the center of the local image as a reference position; and identifying a slope area in the captured image according to the reference position and the height change trend;

[0012] According to the preset landslide features, a target area in the slope area where a landslide is about to occur is identified; the preset landslide features include: the area of ​​the preset landslide features is within a first preset range, the color difference between the preset landslide features and the surrounding areas is greater than a second threshold, and the shape of the preset landslide features is a horseshoe shape, a crack shape or a landslide step shape;

[0013] Controlling the aerial photography equipment to hover or circle over the target area and capture tracking images for a preset time;

[0014] Detect the difference in width, length and area of ​​the landslide features in a plurality of consecutive tracking images. If the difference is within a second preset range, control the aerial photography equipment to continue shooting along a preset flight trajectory. If the difference is not within the second preset range, determine that there is a landslide risk in the target area, and send the image data of the target area and its corresponding coordinates to a remote terminal.

[0015] In some possible embodiments, after determining that the target area has a landslide risk, the method further includes:

[0016] From the above tracking images taken at different times, the contour information of the landslide features is continuously obtained;

[0017] Compare the contour information of the landslide features in the continuous tracking images, and fit the change trend of the above landslide features;

[0018] According to the above-mentioned change trend and the contour lines of the above-mentioned target area, the movement direction of the landslide is simulated based on fluid mechanics, and the above-mentioned landslide movement direction is marked on the risk image of the above-mentioned target area.

[0019] In some possible embodiments, after identifying the slope area in the captured image, the method further includes:

[0020] comparing similarities of the same slope region in the plurality of captured images, and if the similarity is greater than or equal to a third threshold, continuing to compare other captured images;

[0021] If the similarity is less than the third threshold, the slope area is divided into a plurality of sub-areas by using an image cutting method, the similarities of the sub-areas in different captured images are compared, and the target sub-areas whose similarity is less than the third threshold are extracted;

[0022] If the positions of the multiple target sub-areas are different in different captured images and the area difference is within the third preset range, it is determined that the shooting is abnormal and the operation is terminated; if the multiple target sub-areas are arranged in chronological order, the area of ​​the target sub-area located after the above order gradually increases and there is overlap between the positions of the two target sub-areas before and after, it is determined that there is a landslide risk, and the image data of the target sub-area and its corresponding coordinates are sent to the remote terminal.

[0023] In some possible embodiments, after the landslide risk exists, the method further includes:

[0024] The plurality of target sub-regions are arranged in chronological order, with the center of the first target sub-region as the starting point and the center of the last target sub-region as the end point, and the area change trend of the target sub-regions is fitted;

[0025] According to the above area change trend and the contour lines of the above target area, the landslide movement direction is simulated based on fluid mechanics, and the above landslide movement direction is marked on the risk image of the above target sub-area.

[0026] In some possible embodiments, after marking the landslide movement direction on the risk image, the method further includes:

[0027] According to the shape characteristics of the building, identify whether there is a building on the risk image; if not, end the operation; if so, mark the building on the risk image.

[0028] In some possible embodiments, after the remote terminal receives the image data and its corresponding coordinates sent by the aerial photography equipment, it is manually confirmed whether it will have an impact on the buildings in the target area. If it is confirmed to be no, no operation is performed; if it is confirmed to be yes, an alarm message is broadcast to the receiving terminals within the range of the communication base station through the communication base station.

[0029] In some possible embodiments, after matching the local image having a similarity with the landmark feature greater than a first threshold, the method further includes:

[0030] When a local image with a similarity greater than a first threshold value to the valley feature is matched during the matching process with the above-mentioned mark feature; using the valley contour at the corresponding position of the above-mentioned map data platform as the reference contour, identifying whether there is a concave feature on the valley contour of the above-mentioned local image; if the area of ​​the above-mentioned concave feature is less than a fourth threshold value, the operation is terminated; if the area of ​​the above-mentioned concave feature is greater than the fourth threshold value, it is determined that the position of the above-mentioned concave feature is the position where the landslide ends;

[0031] The image corresponding to the above-mentioned concave feature is marked with a danger sign, and the above-mentioned image and its coordinates are uploaded to the above-mentioned map data platform for updating.

[0032] In some possible embodiments, after uploading to the map data platform, the method further includes:

[0033] The map data platform determines whether the concave feature overlaps with the road based on the coordinates of the concave feature. If not, no operation is performed; if overlapping, a prompt message indicating road damage is sent to the navigation software.

[0034] The technical solution of the embodiment of the present application has at least the following advantages and beneficial effects:

[0035] The image processing method of aerial photography disclosed in the present application first uses the initial map data and high-line information provided by the map data platform to extract the key landmark features and height change trends of the mountain, laying the foundation for the subsequent landslide identification. Subsequently, the aerial photography equipment is controlled to shoot the target area along the preset trajectory, and the slope area is located by combining the mountain landmark features and height changes through image matching technology. The three landslide features of area range, color difference and specific shape are then used to conduct a detailed analysis of the slope area to identify potential landslide areas. When a suspected landslide area is found, the aerial photography equipment will hover or hover to continuously track the changes in landslide features. By detecting the difference changes in parameters such as the width, length and area of ​​the landslide features in the continuous tracking image, the landslide risk is evaluated in real time. If the parameter change exceeds the preset range, it is immediately judged that there is a landslide risk, and the relevant information is quickly sent to the remote terminal so that countermeasures can be taken in time. The recognition and dynamic monitoring of landslides through images are realized to improve the timeliness of landslide warnings. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 This is a flow chart of Example 1 in this application;

[0037] Figure 2 A flow chart for generating the direction of landslide movement in Example 1;

[0038] Figure 3 This is a flow chart of Example 2 in this application;

[0039] Figure 4 A flow chart for generating the direction of landslide movement in Example 2;

[0040] Figure 5 This is a flow chart of Embodiment 3 of this application;

[0041] Figure 6 Schematic diagram of the structure of the image processing system for aerial photography in this application. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solutions and advantages of the present application clearer, the technical solutions of the present application will be clearly and completely described below in conjunction with specific implementation methods. The same figure marks in the accompanying drawings represent the same components. It should be noted that the described embodiments are part of the embodiments of the present application, rather than all of the embodiments. Based on the described embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0043] Compared to the embodiments shown in the drawings, feasible implementations within the scope of protection of the present application may have fewer components, other components not shown in the drawings, different components, differently arranged components, or differently connected components, etc. In addition, two or more components in the drawings may be implemented in a single component, or a single component shown in the drawings may be implemented as multiple separate components.

[0044] Embodiment 1

[0045] The embodiment of the present application provides an image processing method for aerial photography, comprising:

[0046] S101, obtaining initial map data of the target area from a map data platform; extracting the landmark features of the mountain and the corresponding height change trend according to the contour lines of the initial map data.

[0047] Step S101 is initiated before the aerial photography task is performed. Its purpose is to pre-process the initial image and provide a reference for subsequent operations. By accessing the map data platform, the initial map data of the target area is automatically downloaded and parsed. These data form the basis for understanding the geomorphic features of the target area, including but not limited to key information such as topography, contour lines, and water system distribution.

[0048] In the data processing process, contour data is particularly important because of its intuitive reflection of terrain changes. The present invention can accurately extract the iconic features of the mountain from the contour data. Specifically, the top of the mountain is the commanding height of the terrain, and its contour lines are shown as closed rings, and the values ​​gradually decrease from the center of the ring to the outside. This feature provides an intuitive basis for identifying the location of the top of the mountain. The identification of the ridge depends on the protruding shape of the contour lines to the low-altitude area. On the contour map, the outline of the ridge shows a special trend, that is, the contour lines protrude from the area with higher values ​​to the area with lower values, forming an obvious convex curve. The valley is opposite to the ridge, and its contour line characteristics are shown as concave into the high-altitude area. On the contour map, the contour lines of the valley are concave from the lower values ​​to the higher values, forming a concave curve. The change trend is directly fitted according to the numerical changes of each point on the contour line.

[0049] S102, controlling the aerial photography equipment with positioning to continuously photograph the target area along a preset flight trajectory to obtain a plurality of photographed images.

[0050] In step S102, the aerial photography equipment realizes a high-precision real-time positioning function by virtue of the Beidou or GPS satellite positioning system it carries, ensuring that the aerial photography equipment can accurately and continuously shoot the target area according to the preset flight trajectory during the flight. Through continuous shooting along the preset flight trajectory, the aerial photography equipment can fully cover the target area and capture various geographical element information including mountains, rivers, vegetation, etc. These image data not only provide rich visual data for the detection of landslides, but also directly serve the mapping work, making the map production more accurate and detailed.

[0051] S103, from each captured image, match a local image whose similarity with the marker feature is greater than a first threshold, and use the center of the local image as a reference position; and identify the slope area in the captured image according to the reference position and the height change trend.

[0052] In step S103, when performing shooting tasks, the aerial photography equipment needs to pay attention to the shooting quality and clarity to ensure the accuracy of subsequent analysis. In order to effectively avoid the impact of cloud cover on image quality, the equipment is usually set to fly at a relatively low altitude but sufficient to cover the target area. In order to accurately determine the specific location of each part in the captured image, the system uses matching and identifying significant landmark features on the mountain, such as mountain tops, ridges and valleys, as reference points for image analysis. These landmark features have been recorded in detail in the initial map data. By comparing similar features in the image, the system can quickly locate key locations such as mountain tops, ridges and valleys.

[0053] After determining these typical reference positions, combined with the height change trend information in the initial map data, the slope area in the image is identified and its approximate boundary and slope direction are marked. For example, if a valley and a ridge are identified in any image, the area between the valley and the ridge forms a slope, which slopes downward from the ridge to the valley.

[0054] S104, identifying a target area in the slope area where a landslide is about to occur according to preset landslide features; the preset landslide features include: an area of ​​the preset landslide feature is within a first preset range, a color difference between the preset landslide feature and its surrounding area is greater than a second threshold, and a shape of the preset landslide feature is a horseshoe shape, a crack shape, or a landslide step shape.

[0055] Step S104 is mainly for detailed analysis and identification of landslide characteristics. It effectively distinguishes between completed landslides and impending landslides by setting multiple landslide characteristic parameters, including area, color difference and shape, thereby confirming the existence of potential landslide areas.

[0056] First, the limitation of the characteristic area of ​​landslides is based on the analysis of a large amount of historical data and the principles of geological disasters. Landslides that have ended, due to the completion of their sliding process, often form a large-scale sliding surface with a relatively large area. On the other hand, the impact range of an impending landslide is relatively small in the initial stage, and the area is relatively small. Therefore, by setting a reasonable area threshold, the system can preliminarily screen out those areas with smaller areas that are more likely to be landslides, thereby avoiding misjudging the landslides that have ended as current risk points.

[0057] Secondly, comparing the color difference between the landslide and its surrounding area is one of the important parameters for identifying potential landslide areas. During the sliding process, the landslide will expose the soil, rocks and other materials under the surface of the mountain. These materials are often significantly different from the surface materials in terms of type, color, humidity, etc. This difference is manifested as obvious color difference in the image. By comparing the color difference between the landslide area and the surrounding environment, the color difference caused by material changes can be identified, and then it can be determined whether the area is in a state of landslide precursor.

[0058] Finally, the shape recognition of landslide features is also an indispensable parameter. Through the summary and analysis of multiple landslide precursor images, the system can identify several typical landslide precursor shapes, such as horseshoe shape, crack shape or landslide step shape. The appearance of these shapes often indicates the redistribution of internal stress of the mountain and the initiation of the landslide process. The present invention uses image processing technology to perform a detailed shape analysis of the slope area. Once a target area that matches the preset landslide characteristic shape is found, it is regarded as a potential landslide risk point.

[0059] S105, controlling the aerial photography equipment to hover or circle over the target area, and taking tracking images for a preset time.

[0060] Step S105 aims to obtain more dynamic information about the changes in the target area and provide strong support for subsequent risk assessment. When using a helicopter or a rotary-wing drone for aerial photography, the advantage is that it can respond quickly and hover directly over the target area. This capability allows operators to flexibly choose the take-off location, especially after an emergency such as an earthquake, to quickly approach the disaster area and capture the real-time situation of the disaster area from multiple angles and heights. The hovering function of a helicopter or a rotary-wing drone not only improves the stability of shooting, but also allows operators to conduct long-term detailed observations of a specific area, thereby obtaining richer and more accurate dynamic information.

[0061] If a glider is used for aerial photography, the target area is tracked mainly by hovering. Although gliders may be subject to certain restrictions in takeoff and landing, their unique flying method can also play an important role in specific situations. During the hovering process, the glider can adjust the flight attitude and shooting angle according to the shooting needs to obtain comprehensive image data. However, due to the tilt of the shooting angle, certain geometric distortion may be introduced. Therefore, for the images taken by the aerial photography equipment in hovering, the tilt angle can be determined by edge detection and Hough transform, and then the image can be rotated to obtain the corrected image.

[0062] S106, detecting the difference in width, length and area of ​​the landslide feature in multiple consecutive tracking images. If the difference is within a second preset range, controlling the aerial photography equipment to continue shooting along the preset flight trajectory; if the difference is not within the second preset range, determining that there is a landslide risk in the target area, and sending the image data of the target area and its corresponding coordinates to the remote terminal.

[0063] Step S106 aims to conduct an in-depth analysis of the continuously captured tracking images to detect the difference in changes in key parameters such as the width, length and area of ​​the landslide features within a certain period of time (20 minutes). If these differences remain within the second preset range, it indicates that the changes in the target area are relatively stable, and the system will continue to shoot along the preset trajectory for continuous monitoring. However, once it is detected that these differences exceed the preset range, it is immediately judged that there is a significant landslide risk in the target area, and the relevant image data and its geographic coordinates are immediately sent to the remote terminal. This early warning information is of great significance for taking timely disaster prevention and mitigation measures and protecting people's lives and property.

[0064] Furthermore, if Figure 2 As shown, after determining that the target area has a landslide risk, the method further includes:

[0065] S201, continuously obtaining contour information of landslide features from tracking images taken at different times. The contour information includes key parameters such as the boundary, shape, and size of the landslide body.

[0066] S202, comparing the contour information of the landslide features in the continuous tracking images, and fitting the change trend of the landslide features. This is mainly to capture the subtle changes of the landslide features (such as the landslide body boundary, crack expansion, etc.) by conducting in-depth analysis on the continuously shot tracking images, and infer the sliding direction of the landslide accordingly.

[0067] Specifically, edge detection is used for morphological analysis to accurately extract the contour information of the landslide features from the image at each time point. Subsequently, by comparing the contour information at different time points, the migration of the landslide boundary, the expansion of cracks, or the formation of new cracks can be detected. That is, the displacement vector of the landslide boundary at different time points is calculated, and a trend line or trend surface of the sliding direction is fitted based on these vector data.

[0068] S203, simulating the movement direction of the landslide based on fluid mechanics according to the change trend and the contour lines of the target area, and marking the movement direction of the landslide on the risk image of the target area.

[0069] In step S203, the change trend of the landslide body can be regarded as the initial movement direction of the landslide body in the subsequent simulation. Subsequently, in order to simplify the analysis and capture the core dynamic characteristics of the landslide body, we regard the landslide body itself as a fluid. And use the concept of fluid mechanics to construct the behavior model of the landslide body. The model includes the following key strategies: First, the movement trajectory of the landslide body is assumed to be continuous and smooth, which ignores possible local mutations or jumps, but retains the consistency of the overall movement direction of the landslide; second, the forefront of the landslide body (i.e., the front edge of the landslide) tends to flow to the lowest position within a preset range, which reflects the natural tendency of the landslide body under the action of gravity, that is, moving to the lowland to release potential energy; third, the flow of the landslide body will continue until it reaches the bottom of the valley or encounters other obstacles and stops.

[0070] This simple fluid model can quickly simulate the movement of landslides, especially in the early stages, when the landslides have not yet been significantly affected by complex terrain or external factors. Using this model, the approximate direction of movement of landslides can be estimated, which is crucial for early warning systems, risk assessments, and emergency response plans. At the same time, the model is computationally efficient and can provide valuable prediction information in a short period of time, buying valuable time for subsequent detailed analysis and response measures.

[0071] Furthermore, based on the shape characteristics of the building, it is possible to identify whether there is a building on the risk image; if not, the operation is terminated; if so, the building is marked on the risk image. Since the first people affected by a landslide are nearby residents, after marking the building, the remote terminal can be prompted to give priority to warning the people in the building, so as to buy as much evacuation time as possible.

[0072] Embodiment 2

[0073] The second embodiment is implemented based on some steps in the first embodiment, such as Figure 3 As shown, it is used to identify the landslide phenomenon that is occurring; that is, after identifying the slope area in the captured image, it specifically includes:

[0074] S301 , comparing similarities of the same slope region in a plurality of captured images, and if the similarity is greater than or equal to a third threshold, continuing to compare other captured images.

[0075] Among them, the similarity comparison is performed on the same slope area in multiple captured images. The purpose is to preliminarily evaluate whether the appearance of the slope changes significantly at different time points or viewing angles. By calculating the similarity of the same slope area in different captured images (using a pixel-level or feature-level matching algorithm), the area with significant changes in appearance or morphology can be identified. If the similarity is greater than or equal to the preset third threshold, it means that the slope area remains relatively stable during this period of time and there is no obvious morphological change, so other captured images can be compared.

[0076] S302: if the similarity is less than a third threshold, divide the slope area into a plurality of sub-areas using an image cutting method, compare the similarities of the sub-areas in different captured images, and extract the target sub-areas whose similarity is less than the third threshold.

[0077] Specifically, when it is found that the similarity of some slope areas is lower than the third threshold, it means that these areas may have abnormal changes and need further detailed analysis. Therefore, step S302 divides these slope areas into smaller sub-areas by image cutting method, and compares the similarity of sub-areas in different captured images. In this way, the specific location where the change occurs is located, and the areas with the most significant changes, namely the target sub-areas, are identified, which have extremely high landslide risks.

[0078] S303, if the positions of multiple target sub-areas are different in different captured images and the area difference is within the third preset range, it is determined that the shooting is abnormal and the operation is terminated; if the multiple target sub-areas are arranged in chronological order, the area of ​​the target sub-area located later in the order gradually increases and the positions of the two target sub-areas overlap, it is determined that there is a landslide risk, and the image data of the target sub-area and its corresponding coordinates are sent to the remote terminal.

[0079] Based on the result of step S302, the extracted target sub-areas are further analyzed and judged. If the positions of multiple target sub-areas are inconsistent in different captured images and the area difference is within the preset third range, this may indicate that the image difference is caused by external factors such as shooting angle and lighting conditions, rather than real slope changes, so it can be determined as shooting abnormality and the current operation is terminated. If multiple target sub-areas are arranged in chronological order, and the area of ​​the target sub-area located later in the order gradually increases, and there is overlap between the positions of the two target sub-areas before and after, it means that the landslide area is increasing, that is, a landslide is in progress. Therefore, the system will automatically send the image data of these target sub-areas and their corresponding coordinates to the remote terminal so that relevant personnel can respond quickly and take necessary preventive and countermeasures to ensure the safety of people's lives and property.

[0080] Furthermore, if Figure 4 As shown in the figure, after determining whether there is a landslide risk, it also includes:

[0081] S401, multiple target sub-regions are arranged in chronological order, with the center of the first target sub-region as the starting point and the center of the last target sub-region as the end point, and the area change trend of the target sub-region is fitted.

[0082] In step S401, the identified multiple target sub-areas are arranged in chronological order, which can indicate the range of the landslide body at different time points during the landslide process. Their area changes directly reflect the speed of the landslide. Taking the center of the first identified target sub-area as the starting point and the center of the last target sub-area as the end point, the system calculates the changes in the areas of these sub-areas over time and fits an area change trend line. This trend line not only shows the expansion speed of the landslide body, but also implies the acceleration or deceleration process of the landslide, which provides an important basis for predicting the future development trend of the landslide. Specifically, the trend of area change is fitted using the trend line obtained by linear regression model fitting.

[0083] S402, simulating the movement direction of the landslide based on fluid mechanics according to the area change trend and the contour lines of the target area, and marking the movement direction of the landslide on the risk image of the target sub-area.

[0084] Specifically, based on the area change trend and combined with the contour data of the target area, the movement direction of the landslide is further simulated through a fluid mechanics model. The model includes the following key strategies: First, the movement trajectory of the landslide is assumed to be continuous and smooth, which ignores possible local mutations or jumps, but retains the consistency of the overall movement direction of the landslide; second, the forefront of the landslide (i.e., the front edge of the landslide) tends to flow to the lowest position within a preset range, which reflects the natural tendency of the landslide under the action of gravity, that is, moving to the lowlands to release potential energy; third, the flow of the landslide will continue until it reaches the bottom of the valley or encounters other obstacles and stops.

[0085] This simple fluid model can quickly simulate the movement of landslides, especially in the early stages, when the landslides have not yet been significantly affected by complex terrain or external factors. Using this model, the approximate direction of movement of landslides can be estimated, which is crucial for early warning systems, risk assessments, and emergency response plans. At the same time, the model is computationally efficient and can provide valuable prediction information in a short period of time, buying valuable time for subsequent detailed analysis and response measures.

[0086] Furthermore, based on the shape characteristics of the building, it is possible to identify whether there is a building on the risk image; if not, the operation is terminated; if so, the building is marked on the risk image. Since the first people affected by a landslide are nearby residents, after marking the building, the remote terminal can be prompted to give priority to warning the people in the building, so as to buy as much evacuation time as possible.

[0087] Embodiment 3

[0088] The third embodiment is implemented based on some steps in the first embodiment, such as Figure 5 As shown, it is used to identify the landslide phenomenon that has occurred; that is, after matching the local image with a similarity with the mark feature greater than the first threshold, it also includes:

[0089] S501, when a local image with a similarity with a valley feature greater than a first threshold is matched during the matching process with the mark feature, the valley contour at the corresponding position of the map data platform is used as the reference contour to identify whether there is a concave feature on the valley contour of the local image; if the area of ​​the concave feature is less than a fourth threshold, the operation is terminated; if the area of ​​the concave feature is greater than the fourth threshold, the position of the concave feature is determined to be the position where the landslide ends.

[0090] It is worth noting that the valley feature in this embodiment is different from the valley feature in the first embodiment (which is simply regarded as a concave curve). The valley feature in this embodiment is to add a certain height threshold to the valley contour line, thereby delineating a more clear valley area. This adjustment is intended to more accurately capture the complex terrain changes formed by soil or rock sliding down the slope when a landslide occurs.

[0091] Since landslides often cause the landslide body to form a distinct protruding landform at the edge of the mountain, it is particularly difficult to accurately locate the landslide area directly by comparing the mountain contours in view of the huge volume and complex terrain of the mountain. Therefore, this embodiment uses the valley contour as a positioning reference and indirectly locates the end point of the landslide by identifying the concave features that appear on the valley contour after the landslide. The concave features here are actually the significant landform traces left in the original valley area after the landslide body slides down.

[0092] In order to distinguish between naturally formed minor terrain changes and significant landform changes caused by landslides, this embodiment sets a fourth threshold to evaluate the area size of the concave feature. When the system detects that the area of ​​the concave feature is less than the fourth threshold, it is considered that the terrain change in the area is within the normal range and no further landslide risk assessment or processing is required. However, once the area of ​​the concave feature exceeds this preset threshold, the system immediately determines that the location is the end point of the landslide, that is, the landslide activity has stopped there.

[0093] S502, annotate the image corresponding to the concave feature with a danger sign, and upload the image and its coordinates to the map data platform for updating. The purpose of marking the danger sign is to send a clear warning signal to the staff, so that the staff can check again.

[0094] Furthermore, after uploading to the map data platform, it also includes:

[0095] The map data platform determines whether the concave feature overlaps with the road based on the coordinates of the concave feature. If not, no operation is performed; if so, a prompt message indicating road damage is sent to the navigation software.

[0096] Specifically, the map data platform cross-checks the acquired coordinate information with the existing road network data to determine whether these concave features overlap with any road in space. The core of this step is to evaluate the impact of landslide events on road traffic. If it is determined that the concave features do not overlap with the road, that is, the landslide does not directly threaten the safety and traffic capacity of the road, then the system will not take further action, because there is no need to issue a road damage warning to the public or relevant departments at this time. However, once the system confirms that the concave features overlap with the road, that is, the landslide has caused substantial damage to the road or potential safety hazards, the map data platform will immediately activate the emergency response mechanism. Specifically, the platform will send an emergency prompt message to the navigation software, clearly indicating the location of the road affected by the landslide and the possible damage. This initiative is designed to help navigation software update road conditions in a timely manner, guide drivers to avoid damaged roads, and ensure the safety of the public.

[0097] Based on the same inventive concept, Figure 6 As shown, an embodiment of the present invention provides an image processing system for aerial photography, comprising:

[0098] The preprocessing module 601 is used to obtain the initial map data of the target area from the map data platform; extract the landmark features of the mountain and the corresponding height change trend according to the contour lines of the initial map data;

[0099] The shooting module 602 is used to control the aerial photography equipment with positioning to continuously shoot the target area along a preset flight trajectory to obtain multiple shooting images;

[0100] The slope area identification module 603 is used to match a local image having a similarity with the marker feature greater than a first threshold from each of the captured images, taking the center of the local image as a reference position; and identifying the slope area in the captured image according to the reference position and the height change trend;

[0101] The landslide identification module 604 is used to identify the target area where a landslide is about to occur in the slope area according to preset landslide features; the preset landslide features include: the area of ​​the preset landslide feature is within a first preset range, the color difference between the preset landslide feature and its surrounding area is greater than a second threshold, and the shape of the preset landslide feature is a horseshoe shape, a crack shape, or a landslide step shape;

[0102] A tracking shooting module 605 is used to control the aerial photography device to hover or circle over the target area and shoot tracking images for a preset time;

[0103] The tracking detection module 606 is used to detect the difference in width, length and area of ​​the landslide features in a plurality of consecutive tracking images. If the difference is within a second preset range, the aerial photography equipment is controlled to continue shooting along a preset flight trajectory. If the difference is not within the second preset range, it is determined that there is a landslide risk in the target area, and the image data of the target area and its corresponding coordinates are sent to the remote terminal.

[0104] Based on the same inventive concept, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that when the program is executed by a processor, the steps corresponding to the above-mentioned aerial photography image processing method are implemented.

[0105] The above are only preferred embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. An image processing method for aerial photography, characterized in that: The method comprises: Acquire the initial map data of the target area from the map data platform; extract the landmark features of the mountain and the corresponding height change trend according to the contour lines of the initial map data; Controlling the aerial photography equipment with positioning to continuously photograph the target area along a preset flight trajectory to obtain a plurality of photographed images; From each of the captured images, a local image having a similarity with the landmark feature greater than a first threshold is matched to locate the positions of the mountain top, the ridge and the valley; taking the center of the local image as a reference position, and identifying the slope area in the captured image according to the reference position and the height change trend; According to preset landslide features, a target area in the slope area where a landslide is about to occur is identified; the preset landslide features include: the area of ​​the preset landslide feature is within a first preset range, the color difference between the preset landslide feature and its surrounding area is greater than a second threshold, and the shape of the preset landslide feature is a horseshoe shape, a crack shape, or a landslide step shape; Controlling the aerial photography equipment to hover or circle over the target area and capture tracking images for a preset time; Detect the difference in width, length and area of ​​the landslide feature in a plurality of consecutive tracking images; if the difference is within a second preset range, control the aerial photography equipment to continue shooting along a preset flight trajectory; if the difference is not within the second preset range, determine that there is a landslide risk in the target area, and continuously obtain the contour information of the landslide feature from the tracking images taken at different times; compare the contour information of the landslide feature in the consecutive tracking images, and fit the change trend of the landslide feature; simulate the landslide movement direction based on fluid mechanics according to the change trend and the contour lines of the target area, and mark the landslide movement direction on the risk image of the target area; send the image data of the target area and its corresponding coordinates to a remote terminal; Alternatively, after identifying the slope area in the captured image, the method further includes: comparing similarities of the same slope region in the plurality of captured images, and if the similarity is greater than or equal to a third threshold, continuing to compare other captured images; If the similarity is less than a third threshold, the slope area is divided into a plurality of sub-areas by using an image cutting method, the similarities of the sub-areas in different captured images are compared, and a target sub-area whose similarity is less than the third threshold is extracted; If the positions of the plurality of target sub-regions are different in different captured images and the area difference is within the third preset range, it is determined that the shooting is abnormal and the operation is terminated; if the plurality of target sub-regions are arranged in chronological order, the area of ​​the target sub-region located after the order gradually increases and the positions of the two target sub-regions before and after overlap, it is determined that there is a landslide risk, and the plurality of target sub-regions are arranged in chronological order, with the center of the first target sub-region as the starting point and the center of the last target sub-region as the end point, and the area change trend of the target sub-region is fitted; According to the area change trend and the contour lines of the target area, the landslide movement direction is simulated based on fluid mechanics, and the landslide movement direction is marked on the risk image of the target sub-area; the image data of the target sub-area and its corresponding coordinates are sent to a remote terminal.

2. The aerial photography image processing method according to claim 1, characterized in that: After marking the landslide movement direction on the risk image, the method further includes: According to the shape characteristics of the building, it is identified whether there is a building on the risk image; if not, the operation is terminated; if so, the building is marked on the risk image.

3. The method for processing aerial photography images according to claim 1, wherein: After the remote terminal receives the image data and its corresponding coordinates sent by the aerial photography equipment, it is manually confirmed whether it will have an impact on the buildings in the target area. If it is confirmed to be no, no operation will be performed; if it is confirmed to be yes, the alarm information will be broadcast to the receiving terminals within the range of the communication base station through the communication base station.

4. The method for processing aerial photography images according to claim 1, wherein: After matching the local image whose similarity with the landmark feature is greater than the first threshold, the method further includes: When a local image having a similarity with the valley feature greater than a first threshold value is matched during the matching process with the landmark feature, Taking the valley contour at the corresponding position of the map data platform as the reference contour, identifying whether there is a concave feature on the valley contour of the local image; if the area of ​​the concave feature is less than a fourth threshold, ending the operation; if the area of ​​the concave feature is greater than the fourth threshold, determining that the position of the concave feature is the end position of the landslide; The image corresponding to the concave feature is marked with a danger sign, and the image and its coordinates are uploaded to the map data platform for updating.

5. The method for processing aerial photography images according to claim 4, characterized in that: After uploading to the map data platform, the method further includes: The map data platform determines whether the concave feature overlaps with the road based on the coordinates of the concave feature. If not, no operation is performed; if overlapping, a prompt message of road damage is sent to the navigation software.

6. A system for implementing the aerial photography image processing method according to claim 1, characterized in that: The system comprises: A preprocessing module is used to obtain the initial map data of the target area from the map data platform; extract the landmark features of the mountain and the corresponding height change trend according to the contour lines of the initial map data; A shooting module is used to control the aerial photography equipment with positioning to continuously shoot the target area along a preset flight trajectory to obtain multiple shooting images; A slope area recognition module, used to match a local image having a similarity with the marker feature greater than a first threshold from each of the captured images, taking the center of the local image as a reference position; and identifying the slope area in the captured image according to the reference position and the height change trend; A landslide identification module is used to identify a target area in the slope area where a landslide is about to occur according to preset landslide features; the preset landslide features include: the area of ​​the preset landslide feature is within a first preset range, the color difference between the preset landslide feature and its surrounding area is greater than a second threshold, and the shape of the preset landslide feature is a horseshoe shape, a crack shape, or a landslide step shape; A tracking shooting module, used to control the aerial photography equipment to hover or circle over the target area and shoot tracking images for a preset time; The tracking detection module is used to detect the difference in width, length and area of ​​the landslide features in a plurality of consecutive tracking images. If the difference is within a second preset range, the aerial photography equipment is controlled to continue shooting along a preset flight trajectory. If the difference is not within the second preset range, it is determined that there is a landslide risk in the target area, and the image data of the target area and its corresponding coordinates are sent to a remote terminal.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps corresponding to the image processing method for aerial photography as described in any one of claims 1 to 5 are implemented.

Images