Multispectral remote sensing image real-time processing method

Abstract

The invention discloses a multispectral remote sensing image real-time processing method. The method comprises a data acquisition node (S1), an image splicing node (S2) and a parameter inversion node (S3). The method is characterized in that multispectral images can be read and processed in real time, and a corresponding parameter distribution diagram is generated. Wherein the data acquisition node (S1) can read multispectral data from a multispectral sensor in real time and perform preprocessing such as geometric correction, radiometric calibration, image registration and image synthesis on the multispectral data; the image splicing node (S2) can generate an orthoimage in real time; and the parameter inversion node (S3) can draw a parameter distribution diagram according to the embedded parameter inversion model. According to the multispectral remote sensing data real-time processing framework, the interpretation difficulty of the multispectral remote sensing data is reduced, and the processing efficiency of the multispectral remote sensing data is improved.

Description

Technical Field

[0001] The invention relates to the field of unmanned aerial vehicle remote sensing image interpretation and mapping, and relates to a remote sensing real-time monitoring method for agricultural information, and specifically to a multi-spectral remote sensing image real-time processing method. Background Art

[0002] In the past agricultural multispectral remote sensing data acquisition, modeling inversion and spatial mapping process, the processing and interpretation of remote sensing data still requires manual work in the later stage, which requires operators to have high professional quality, and on the other hand, it also leads to a certain lag between the data processing results and the actual growth of crops. Therefore, it is necessary to provide a multispectral remote sensing image real-time processing method to solve the above problems, so the present invention has certain practical significance. Summary of the invention

[0003] The purpose of the present invention is to provide a real-time processing method for multispectral remote sensing images, develop a real-time mapping framework that integrates real-time image reading, mapping and inversion, provide a reference for the real-time processing research of remote sensing data, and improve the efficiency of agricultural multispectral remote sensing image interpretation.

[0004] In order to solve the above technical problems, the present invention proposes the following technical solutions:

[0005] A multispectral remote sensing image real-time processing method, characterized by comprising: a data acquisition node (S1), an image stitching node (S2) and a parameter inversion node (S3).

[0006] Furthermore, the data acquisition node realizes real-time acquisition of multispectral images based on a "data buffer" mechanism, and performs preprocessing such as geometric correction, radiation calibration, image registration and image synthesis on the multispectral images.

[0007] Furthermore, the image stitching node realizes orthophoto mapping of the composite image.

[0008] Furthermore, the parameter inversion node realizes band decomposition, image secondary registration and parameter calculation of the orthophoto, and generates a target parameter distribution map.

[0009] Furthermore, the "data buffer" mechanism divides data reading and data processing into two independent stages by establishing a data container and a data log, thereby improving the real-time performance of the method.

[0010] Furthermore, the geometric correction performs radial distortion correction and tangential distortion correction on the image to eliminate deformation caused by the geometric position, shape, size, orientation and other characteristics of the objects on the original image.

[0011] Furthermore, the radiation calibration converts the target image into a reflectivity image using a pre-photographed reference gray plate image with a fixed reflectance.

[0012] Furthermore, the image registration uses the blue band image (B) as the reference image, and unifies the remaining band images and the reference image into the same coordinate system according to the homography matrix between the remaining band images (red band (R), green band (G), near infrared band (NIR), red edge band (Red Edge)) and the reference image, and performs mask processing.

[0013] Furthermore, the image synthesis synthesizes the red band, green band and reference image of the registered image into a set of pseudo-color images, called "RGB" images; and synthesizes the red edge band, near infrared band and reference image of the registered image into a set of pseudo-color images, called "RE-NB" images.

[0014] Furthermore, the orthophoto mapping utilizes VSLAM technology to mosaic the “RGB” image and the “RE-NB” image to generate a complete orthophoto.

[0015] Furthermore, the band decomposition decomposes the two groups of orthophotos into five groups of single-band images, namely, R, G, B, NIR and Red Edge.

[0016] Furthermore, the secondary image registration uses the homography matrix between the B-band images shared by the two groups of composite images after decomposition to unify the remaining bands and the B-band image into the same coordinate system.

[0017] Furthermore, the parameter calculation extracts features such as vegetation index and texture index from the spliced ​​single-band orthophotos, and generates a distribution map of the target parameters in combination with a custom parameter inversion model.

[0018] The present invention is beneficial in that:

[0019] (1) An integrated processing framework of data acquisition, real-time mapping and parameter inversion was realized, which improved the efficiency of remote sensing data interpretation;

[0020] (2) The ROS system was introduced to encapsulate S1, S2, and S3 into independent ROS nodes, which reduced the computing power requirements of the method for hardware devices;

[0021] (3) The parameter inversion model can be customized according to one’s own needs, which improves the universality of the method. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments:

[0023] Figure 1 It is an overall schematic diagram of a real-time processing method for multispectral remote sensing images.

[0024] Figure 2 It is a schematic diagram of the "data processing" mechanism of a real-time processing method for multispectral remote sensing images.

[0025] Figure 3 It is a technology roadmap for data acquisition nodes of real-time processing methods for multispectral remote sensing images.

[0026] Figure 4 It is a technology roadmap for image stitching nodes of real-time processing methods for multispectral remote sensing images.

[0027] Figure 5 It is a technical roadmap for parameter inversion nodes of multispectral remote sensing image real-time processing methods.

[0028] Figure 6 It is the result of a mapping test of a real-time processing method for multispectral remote sensing images.

[0029] Figure 7 It is the accuracy test result of the parameter inversion model of a real-time processing method for multispectral remote sensing images. DETAILED DESCRIPTION

[0030] The present invention will be further described in detail below in conjunction with the accompanying drawings.

[0031] See also Figure 1 The figure shows an overall schematic diagram of a multispectral remote sensing image real-time processing method, including a data acquisition node (S1), an image stitching node (S2) and a parameter inversion node (S3).

[0032] See also Figure 2 A schematic diagram of a "data buffer" mechanism of a real-time processing method for multispectral remote sensing images is shown. The "data buffer" mechanism establishes a data container and a data log, continuously and iteratively updates the data directory of the multispectral camera in the data log, compares the content in the current data log with the content of the previous round of data log, determines the newly generated data directory in the camera, and then generates a corresponding data reading command based on the directory, sends the request to the multispectral camera, and downloads the newly generated data to the data container; on the other hand, the edge computing device takes out the data from the data container, preprocesses the data, and saves the preprocessing results.

[0033] See also Figure 3 A data acquisition node technology roadmap for a multispectral remote sensing image real-time processing method is shown, and the technology roadmap includes the following specific steps:

[0034] S106: Determine whether the communication between the computer and the multispectral sensor is established;

[0035] S107: Start the multi-spectral data reading thread;

[0036] S108: Enter the folder in the sensor that stores the data of this task;

[0037] S109: Generate an initial data log;

[0038] S110: Continuously checking whether the data log is updated;

[0039] S111: Generate a download command for updated data;

[0040] S112: Download the updated multispectral data to the data container via the Socket network;

[0041] S113: Start multispectral data preprocessing thread;

[0042] S114: Entering the data container;

[0043] S115: Continuously checking whether the content in the data container is updated;

[0044] S116: Eliminate abnormal data;

[0045] S117: Perform geometric correction on the data;

[0046] S118: performing radiometric calibration on the data;

[0047] S119: performing image registration on the data;

[0048] S120: performing image synthesis on the data to generate an “RGB” image and an “RE-NB” image;

[0049] S121: Save the composite image.

[0050] See also Figure 4 A technical roadmap of image stitching nodes for real-time processing of multispectral remote sensing images is shown, and the technical roadmap includes the following specific steps:

[0051] S201: Read the synthetic images "RGB" and "RE-NB" in S1;

[0052] S202: Reading metadata of the synthetic image, including latitude, longitude and altitude;

[0053] S203: Perform pose estimation;

[0054] S204: performing point cloud densification;

[0055] S205: generating a digital surface;

[0056] S206: performing orthogonal correction;

[0057] S207: Perform mosaic mapping;

[0058] S208: Save the stitched orthophoto.

[0059] See also Figure 5 A technical roadmap of parameter inversion node of a multispectral remote sensing image real-time processing method is shown, and the technical roadmap includes the following specific steps:

[0060] S301: Read the stitched orthophoto;

[0061] S302: Determine whether an orthophoto is received;

[0062] S303: decomposing the image into single-band images;

[0063] S304: performing secondary image registration based on the shared B-band image;

[0064] S305: cropping the image;

[0065] S306: normalizing the image to obtain a reflectance image;

[0066] S307: Extract vegetation index and texture index features from the image;

[0067] S308: Draw a target parameter distribution map based on the embedded parameter inversion model.

[0068] To make the purpose, technical solutions and advantages of the embodiments of the present invention more clear, the following is a summary of the present invention. Figure 6 and attached Figure 7 , a complete and clear description of the technical solution in the embodiment of the present invention is given:

[0069] The present invention takes field corn as the experimental object, collects a batch of UAV multispectral remote sensing images of field corn and several corresponding field corn leaf area index (LAI) ground truth values ​​in advance, inputs the multispectral images into the multispectral remote sensing image real-time processing method, obtains the corresponding field corn LAI parameter distribution map, extracts the LAI prediction value corresponding to the LAI ground truth value position in the distribution map, and performs statistical analysis. The analysis result is: the determination coefficient R 2 The average value of the regression line is 0.8825, the RMSE is 0.4117, and the MAE is 0.3158.

[0070] Shown Figure 6 This is a distribution map of field corn LAI parameters obtained by using the multispectral remote sensing image real-time processing method.

[0071] Shown Figure 7 is a scatter plot of the LAI prediction value and the LAI ground truth value.

[0072] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be within the protection scope of the present invention.

Claims

1. A multispectral remote sensing image real-time processing method, characterized in that: The steps include: S1, real-time acquisition of multispectral images and preprocessing of them such as geometric correction, radiometric calibration, image registration and image synthesis; S2, orthophoto mapping of the image preprocessed in S1; S3, performing band decomposition, image secondary registration and parameter calculation on the image obtained in S2 to generate a target parameter distribution map.

2. The method for real-time processing of multispectral remote sensing images according to claim 1 is characterized in that the real-time acquisition of multispectral images in S1 uses a "data buffer" mechanism to realize the reading and processing of multispectral images.

3. The method for real-time processing of multispectral remote sensing images according to claim 1 is characterized in that the geometric correction in S1 performs radial distortion correction and tangential distortion correction on the image to eliminate deformation caused by the geometric position, shape, size, orientation and other characteristics of the objects on the original image.

4. The method for real-time processing of multispectral remote sensing images according to claim 1 is characterized in that the radiation calibration in S1 converts the target image into a reflectance image using a pre-shot reference gray plate image with a fixed reflectance coefficient.

5. The method for real-time processing of multispectral remote sensing images according to claim 1 is characterized in that, in S3, the image registration uses a group of single-band images as reference images, and unifies the remaining band images and the reference image into the same coordinate system according to the homography matrix between the remaining band images (red band (R), green band (G), near infrared band (NIR), red edge band (Red Edge)) and the reference image (blue band (B)), and performs mask processing.

6. The real-time processing method for multispectral remote sensing images according to claim 1 is characterized in that, in the image synthesis, the red band, green band and reference image of the registered image are synthesized into a group of pseudo RGB images, called "RGB" images; and the red edge band, near infrared band and reference image of the registered image are synthesized into a group of pseudo RGB images, called "RE-NB" images.

7. The method for real-time processing of multispectral remote sensing images according to claim 1 is characterized in that the orthophoto mapping uses VSLAM technology to mosaic the "RGB" image and the "RE-NB" image to generate a complete orthophoto.

8. The method for real-time processing of multispectral remote sensing images according to claim 1 is characterized in that the band decomposition decomposes the two groups of orthophotos into five groups of single-band images: R, G, B, NIR and Red Edge.

9. The method for real-time processing of multispectral remote sensing images according to claim 1 is characterized in that the secondary image registration utilizes the homography matrix between the B-band images shared by the two sets of composite images after decomposition to unify the remaining bands and the B-band image into the same coordinate system.

10. The method for real-time processing of multispectral remote sensing images according to claim 1 is characterized in that the parameter calculation extracts features such as vegetation index and texture index from the spliced ​​single-band orthoimages, and generates a distribution map of the target parameters in combination with a custom parameter inversion model.

11. The method for real-time processing of multispectral remote sensing images according to claim 1 is characterized in that the "data buffer" mechanism continuously iterates and updates the data directory of the multispectral camera in the data log by establishing a data container and a data log, compares the content in the current data log with the content in the previous round of data log, determines the newly generated data directory in the camera, and then generates a corresponding read request according to the directory, sends the request to the multispectral camera, and downloads the newly generated data into the data container; on the other hand, the computer takes out the data from the data container, pre-processes the data, and saves the pre-processing results.

Images