Self-adaptive remote sensing image recognition method and system

Abstract

The invention discloses a self-adaptive remote sensing image recognition method and system. The method comprises the following steps: acquiring a first geographic coordinate corresponding to a to-be-recognized target area in a remote sensing image; sending the first geographic coordinate to the 3D panorama, and obtaining a panorama coordinate set with the first geographic coordinate as the center; calculating a second geographic coordinate closest to the first geographic coordinate in the panorama coordinate set; and calculating an azimuth angle according to the first geographic coordinate and the second geographic coordinate. The to-be-recognized coordinates of the remote sensing image are obtained, the second geographic coordinates of the 3D panorama are obtained according to the coordinates, the azimuth angle is calculated, the second height value of the to-be-recognized area is calculated according to the marker post building and the shadow length, and the pitch angle is obtained. And the image information of the 3D panorama is obtained through the azimuth angle, the pitch angle and the second geographic coordinate, and self-adaptive identification is carried out. Therefore, the recognition effect can be improved by utilizing the relatively high resolution of the 3D panorama.

Description

Technical Field

[0001] The present invention relates to the technical field of remote sensing image recognition, and in particular, to an adaptive remote sensing image recognition method and system. Background Art

[0002] Remote sensing images are collected by satellites for ground images. Due to the high altitude of the satellites, the collected resolution is low. Only basic item types can be distinguished from the images, such as green circles being plants and squares being buildings, etc. Further item recognition, such as what specific plants are, cannot be achieved. That is, the resolution of existing remote sensing images is low, and the problem of high-precision recognition of remote sensing images cannot be achieved. Summary of the Invention

[0003] Therefore, an adaptive remote sensing image recognition method and system are needed to solve the problem of low recognition accuracy of remote sensing images.

[0004] To achieve the above object, the present invention provides an adaptive remote sensing image recognition method, including the following steps: Obtain the first geographical coordinates corresponding to the target area to be recognized in the remote sensing image; Send the first geographical coordinates to the 3D panoramic view to obtain a set of panoramic view coordinates centered on the first geographical coordinates; Calculate the second geographical coordinates closest to the first geographical coordinates in the set of panoramic view coordinates; Calculate the azimuth angle according to the first geographical coordinates and the second geographical coordinates; Obtain the benchmark building closest to the first geographical coordinates in the benchmark building database, and obtain the first height value and the third geographical coordinates of the benchmark building; Obtain the length of the first shadow area at the third geographical coordinate position in the remote sensing image and the length of the second shadow area at the first geographical coordinate position, and calculate the second height value of the area to be recognized according to the length of the first shadow area, the length of the second shadow area and the first height value; Calculate the distance according to the first geographical coordinates and the second geographical coordinates, and calculate the pitch angle with the second height value; Send the azimuth angle, the pitch angle and the second geographical coordinates to the 3D panoramic view to obtain the first image information in the 3D panoramic view; Send the first image information to the item recognition database to obtain the recognition result.

[0005] Further, it further includes the step of: After obtaining the first image information, obtain the second image information of the target area to be recognized in the remote sensing image; Compare the similarity between the first image information and the second image information. If the similarity is greater than or equal to the preset value, send the first image information to the item recognition database to obtain the recognition result; When the similarity is less than the preset value, calculate the coordinate in the panoramic image coordinate set that is separated from the second geographical coordinate by a preset value as the new second geographical coordinate, recalculate the azimuth and elevation angles according to the new second geographical coordinate, and obtain the first image information again. Send the first image information to the item recognition database to obtain the recognition result.

[0006] Further, it further includes the step: the similarity comparison includes color comparison.

[0007] Further, before obtaining the first geographical coordinate, it further includes the steps: Obtain the central geographical coordinate of the current remote sensing image and send it to the 3D panoramic image, and obtain the panoramic image coordinate set according to the current central geographical coordinate; Determine the recognizable area within a preset range according to the panoramic image coordinate set, overlay and display this area on the remote sensing image, and click on this area to trigger the recognition of the remote sensing image.

[0008] Further, it further includes the step: The recognition result includes multiple recognized items and is sorted according to the similarity. According to the proximity of the second height value to the multiple recognized items, change the similarity sorting of the multiple recognized items.

[0009] Further, the obtaining of the first image information in the 3D panoramic image includes: Take a screenshot of the 3D panoramic image and save it as the first image information.

[0010] Further, sending the first geographical coordinate to the 3D panoramic image includes: Send the first geographical coordinate and the preset scale to the 3D panoramic image.

[0011] Further, send the recognition result to the retrieval engine for retrieval, obtain the retrieval result and display it on the remote sensing image.

[0012] Further, the item recognition database is a plant type database.

[0013] The present invention also provides an adaptive remote sensing image recognition system, including a memory and a processor. A computer program is stored on the memory, and when the computer program is executed by the processor, it implements the steps of the method described in any one of the embodiments of the present invention.

[0014] Different from the prior art, the above technical solution obtains the coordinates to be recognized in the remote sensing image, then obtains the second geographical coordinates of the 3D panoramic view according to the coordinates and calculates the azimuth angle, then calculates the second height value of the area to be recognized according to the landmark building and the shadow length, so as to calculate the pitch angle, and then obtains the image information of the 3D panoramic view through the azimuth angle, pitch angle and second geographical coordinates and performs recognition adaptively. In this way, using the relatively high resolution of the 3D panoramic view, more accurate recognition of items can be achieved, and the recognition effect can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 is a flowchart of the method of the present invention; Figure 2 is a schematic diagram of the display interface after the system of the present invention is recognized; Figure 3 is a flowchart of the method of another embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0016] To describe in detail the technical content, structural features, achieved objectives and effects of the technical solution, the following is described in detail in conjunction with specific embodiments and with reference to the accompanying drawings.

[0017] Please refer to Figures 1 to 3 , the present invention provides an adaptive remote sensing image recognition method, including the following steps: Step S101 obtains the first geographical coordinates corresponding to the target area to be recognized in the remote sensing image; the area to be recognized can be the position clicked by the mouse as the area to be recognized. The geographical coordinates are longitude and latitude coordinates. Then step S102 sends the first geographical coordinates to the 3D panoramic view to obtain a panoramic view coordinate set centered on the first geographical coordinates; the 3D panoramic view will take the first geographical coordinates as the center and obtain a plurality of panoramic view coordinates with panoramic images within a certain range around it. This coordinate is a set. Each coordinate corresponds to a 3D panoramic view. Then enter step S103 to calculate the second geographical coordinates closest to the first geographical coordinates in the panoramic view coordinate set; the distance calculation can be performed through the Haversine formula of the longitude and latitude calculation function to find the closest panoramic view coordinate point. Then in step S104, the azimuth angle is calculated according to the first geographical coordinates and the second geographical coordinates; the azimuth angle here is the angle by which the first geographical coordinates deflect relative to the second geographical coordinates with the second geographical coordinates as the center, and this angle is the angle of deflection relative to the north direction. The 3D panoramic view of the present invention can be a 3D road panoramic view as an example. In some embodiments, it can also be a 3D panoramic view in a park, etc.

[0018] After the azimuth calculation is completed, step S105 is entered to obtain the benchmark building in the benchmark building database that is closest to the first geographical coordinate, and obtain the first height value and the third geographical coordinate of the benchmark building. It should be emphasized that the benchmark buildings are landmark buildings in each region. They are relatively tall, obvious, and prominent. These benchmark buildings include their locations and heights. These data are all public data. For example, a well-known international center in Fuzhou has a height of approximately 274 meters, and its longitude and latitude are longitude 29.0737XXX° and latitude 119.3135XXX°, which are the third geographical coordinates. Similar landmark buildings exist in different cities, and these benchmark buildings can be stored in a database. Then, in step S106, the length of the first shadow area at the position of the third geographical coordinate in the remote sensing image and the length of the second shadow area at the position of the first geographical coordinate are obtained, and the second height value of the area to be recognized is calculated based on the length of the first shadow area, the length of the second shadow area, and the first height value. Here, the shadow length is the shadow area formed by the sunlight irradiating on the building in the remote sensing image. The shadow area can be obtained in the following way: First, convert the panoramic image near the benchmark building into the HSV / grayscale space and separate the brightness information, and then perform noise reduction processing (Gaussian blur). Then, according to the set brightness threshold, the shadow area can be segmented. Search for and record the shadow contour boundary, and then calculate the position of the shadow contour that is farthest from the third geographical coordinate, which is the length of the first shadow area.

[0019] Similarly, for the position of the first geographical coordinate, the length of the second shadow area can also be obtained by the same method. Since the sunlight angle is close in the same area, the height of the shadow area to be recognized can be calculated through the shadow area lengths of the benchmark building and the area to be recognized. In some embodiments, to avoid the deviation of the shadow area length calculated by a single benchmark building, multiple benchmark buildings can be used for calculation, such as three. Then, the second height values calculated based on these three benchmark buildings are averaged, which is more accurate. Of course, the second height value of the present invention is subsequently used to calculate the pitch angle, and the actual accuracy requirement is relatively low, allowing a certain error.

[0020] Then, in step S107, the distance is calculated based on the first geographical coordinate and the second geographical coordinate (this distance can also be used to calculate the magnification ratio of the panoramic image. For example, a larger magnification ratio is used for a long distance, and a smaller magnification ratio is used for a short distance), and the pitch angle is calculated based on the second height value. In this way, the direction of the pitch angle is the top position of the target area to be recognized, realizing the adaptive position acquisition in the height direction.

[0021] Then, in step S108, the azimuth angle, elevation angle, and the second geographical coordinate are sent to the 3D panoramic view to obtain the first image information in the 3D panoramic view; in this way, the first image information is the image information of the top position of the target area to be recognized. Finally, in step S109, the first image information is sent to the item recognition database to obtain the recognition result. There are many existing open databases in the item recognition database, such as the general object and scene recognition API interface of Baidu, and there are also many specific databases, such as the plant variety database, the building recognition database, etc. Based on the relatively clear photos of the panoramic view, relatively accurate item recognition can be achieved. For example, if the bougainvillea plant on the road surface is recognized, the interface information feedback by the API is: {"status": "success", "data": {"name": "Bougainvillea spectabilis", "scientific_name": "Bougainvillea spectabilis Willd.", "description": "Bougainvillea spectabilis Willd. is an angiosperm of the genus Bougainvillea in the family Nyctaginaceae, and it is a vine-like shrub; branches and leaves are densely covered with soft hairs; thorns are axillary and curved downward. The leaves are oval or ovate, with a round base; the inflorescence is axillary or terminal; the bracts are elliptic-ovate, with a base rounded to cordate, and the color is dark red or light purple; the perianth is a green tube with a narrow cylindrical shape, and the lobes are yellow when spreading; the ovary has a stalk; the fruit is densely covered with hairs; the flowering period is between winter and spring."}}, then as Figure 2 shown, the recognition content can be displayed at the position of the mouse on the interface of the system.

[0022] In the above embodiment, by obtaining the coordinates to be recognized of the remote sensing image, then obtaining the second geographical coordinate of the 3D panoramic view according to the coordinate and calculating the azimuth angle, then calculating the second height value of the area to be recognized according to the landmark building and the shadow length, thereby calculating the elevation angle, and then obtaining the image information of the 3D panoramic view through the azimuth angle, elevation angle, and the second geographical coordinate and performing adaptive recognition. In this way, using the relatively high resolution of the 3D panoramic view, relatively accurate recognition of items can be achieved, improving the recognition effect. It should be noted that since the remote sensing image and the 3D panoramic view actually have acquisition times, it is necessary to call the remote sensing image and the 3D panoramic view with close acquisition times for recognition and judgment, which can improve the accuracy.

[0023] In some embodiments, it further includes the steps of: after obtaining the first image information, obtaining the second image information of the target area to be recognized in the remote sensing image; comparing the similarity between the first image information and the second image information, and if the similarity is greater than or equal to the preset value, sending the first image information to the item recognition database to obtain the recognition result; when the similarity is less than the preset value, calculating the coordinate in the panoramic map coordinate set that is separated from the second geographic coordinate by a preset value (the preset value can be the number of coordinates, such as the next coordinate in the interval, or the distance, such as the nearest coordinate after an interval of 5 meters) as the new second geographic coordinate, recalculating the azimuth angle and elevation angle according to the new second geographic coordinate, and obtaining the first image information again, sending the first image information to the item recognition database to obtain the recognition result. For example, if the second image information of the target area is a plant with a green top, but due to possible billboard occlusion in the panoramic map, the first image information obtained is a white billboard, then when comparing the similarity between the first image information and the second image information, the similarity difference is large, so recognition is not performed, and recognition is carried out again after changing the position. This can improve the accuracy of image recognition. The image similarity comparison can use existing interfaces and algorithms to calculate the similarity.

[0024] In some embodiments, the similarity comparison includes color comparison. The color comparison can be performed by extracting the proportion of each color component. Through simple color comparison, rapid exclusion can be carried out, the comparison efficiency is higher, and the algorithm is simple.

[0025] Further, as Figure 3 shown, before obtaining the first geographic coordinate, it further includes the steps of: Step S201 obtaining the central geographic coordinate of the current remote sensing image and sending it to the 3D panoramic map, and obtaining the panoramic map coordinate set according to the current central geographic coordinate; Step S402 determining the recognizable area within a preset range according to the panoramic map coordinate set, overlaying and displaying this area on the remote sensing image, and clicking on this area to trigger the remote sensing image recognition. This can preset the area that can support remote sensing image recognition and avoid triggering unrecognizable areas.

[0026] In some embodiments, it further includes the steps of: the recognition result includes multiple recognized items and is sorted according to similarity. According to the proximity of the second height value to the multiple recognized items, the similarity sorting of the multiple recognized items is changed. This can rank the items with higher similarity in the front and make the recognition more accurate.

[0027] Further, the obtaining of the first image information in the 3D panoramic map includes: taking a screenshot of the 3D panoramic map and saving it as the first image information. This can utilize the existing 3D panoramic map without using the interface API, and the deployment is more convenient.

[0028] Further, sending the first geographical coordinate to the 3D panoramic view includes: sending the first geographical coordinate and a preset scale to the 3D panoramic view. This realizes scale control to avoid obtaining too many or too few coordinates.

[0029] Further, sending the recognition result to the retrieval engine for retrieval, obtaining the retrieval result and displaying it on the remote sensing image. In this way, the user can see the information related to the recognition result on the remote sensing image, which is convenient for information acquisition.

[0030] The present invention also provides an adaptive remote sensing image recognition system, including a memory and a processor. A computer program is stored on the memory, and when the computer program is executed by the processor, the steps of the method according to any one of the embodiments of the present invention are realized. The recognition system of the present invention obtains the coordinates to be recognized of the remote sensing image, then obtains the second geographical coordinate of the 3D panoramic view according to the coordinates and calculates the azimuth angle, then calculates the second height value of the area to be recognized according to the landmark building and the shadow length, thereby calculating the pitch angle, and then obtains the image information of the 3D panoramic view through the azimuth angle, pitch angle and the second geographical coordinate and performs adaptive recognition. In this way, using the relatively high resolution of the 3D panoramic view, more accurate recognition of items can be realized, and the recognition effect can be improved.

[0031] It should be noted that although the above embodiments have been described in this article, the patent protection scope of the present invention is not limited thereby. Therefore, based on the innovative concept of the present invention, any changes and modifications made to the embodiments described in this article, or equivalent structural or equivalent process transformations made using the content of the specification and drawings of the present invention, directly or indirectly applying the above technical solutions to other related technical fields, are all included in the patent protection scope of the present invention.

Claims

1. An adaptive remote sensing image recognition method, characterized in that: The steps include: Acquire the first geographic coordinates corresponding to the target area to be identified in the remote sensing image; Send the first geographic coordinate to the 3D panorama to obtain a panorama coordinate set centered on the first geographic coordinate; Calculate a second geographic coordinate in the panoramic coordinate set that is closest to the first geographic coordinate; Calculating an azimuth according to the first geographic coordinate and the second geographic coordinate; Acquire a benchmark building closest to the first geographic coordinate in a benchmark building database, and acquire a first height value and a third geographic coordinate of the benchmark building; Obtaining a first shadow area length at a third geographic coordinate position and a second shadow area length at a first geographic coordinate position in the remote sensing image, and calculating a second height value of the area to be identified according to the first shadow area length, the second shadow area length and the first height value; Calculate the distance according to the first geographic coordinate and the second geographic coordinate, and calculate the pitch angle with the second altitude value; Sending the azimuth angle, the elevation angle, and the second geographic coordinates to the 3D panoramic image to obtain first image information in the 3D panoramic image; The first image information is sent to an object recognition database to obtain a recognition result.

2. The adaptive remote sensing image recognition method according to claim 1, characterized in that: Also includes the steps: After acquiring the first image information, acquiring second image information of the target area to be identified in the remote sensing image; Compare the first image information with the second image information for similarity, and if the similarity is greater than or equal to a preset value, send the first image information to an object recognition database to obtain a recognition result; When the similarity is less than a preset value, the coordinate in the panoramic coordinate set that is separated from the second geographic coordinate by a preset value is calculated as a new second geographic coordinate, the azimuth and elevation angle are recalculated according to the new second geographic coordinate, and the first image information is reacquired, and the first image information is sent to the object recognition database to obtain the recognition result.

3. The adaptive remote sensing image recognition method according to claim 2, characterized in that: The method also includes the step of: the similarity comparison includes color comparison.

4. The adaptive remote sensing image recognition method according to claim 1, characterized in that: Before obtaining the first geographic coordinates, the method further includes the following steps: Get the central geographic coordinates of the current remote sensing image and send them to the 3D panorama, and get the panorama coordinate set based on the current central geographic coordinates; According to the panoramic coordinate set, a recognizable area is determined within a preset range, and the area is superimposed on the remote sensing image and displayed. Clicking the area triggers remote sensing image recognition.

5. The adaptive remote sensing image recognition method according to claim 1, characterized in that: Also includes the steps: The recognition result includes a plurality of recognized objects and is sorted according to similarity, and the similarity sorting of the plurality of recognized objects is changed according to the proximity between the second height value and the plurality of recognized objects.

6. The adaptive remote sensing image recognition method according to claim 1, characterized in that: The acquiring of first image information in the 3D panoramic image comprises: A screenshot of the 3D panoramic image is taken and saved as the first image information.

7. The adaptive remote sensing image recognition method according to claim 1, characterized in that: Sending the first geographic coordinates to the 3D panorama includes: The first geographic coordinates and the preset scale are sent to the 3D panorama.

8. The adaptive remote sensing image recognition method according to claim 1, characterized in that: The recognition results are sent to the search engine for retrieval, and the retrieval results are obtained and displayed on the remote sensing image.

9. The adaptive remote sensing image recognition method according to claim 1, characterized in that: The object identification database is a plant type database.

10. An adaptive remote sensing image recognition system, characterized in that: The method comprises a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 9 are implemented.

Images