A Geometric Correction Method and System for Multispectral Remote Sensing Data

Abstract

The present invention relates to the technical field of image restoration, and specifically relates to a geometric correction method and system for multi-spectral remote sensing data. The method includes: obtaining a remote sensing image of a target area, and determining a plurality of target partitions in the remote sensing image; determining the distortion degree of the target partition, and determining the model order weight of the target partition based on the distortion degree; determining the anti-distortion model order of the target partition according to the model order weight; and using the anti-distortion model corresponding to the anti-distortion model order to perform distortion correction on the target partition. Through the geometric correction method for multi-spectral remote sensing data of the present invention, a more matching anti-distortion model order is determined for different regions of the image, greatly improving the distortion correction accuracy of the image, enabling each region of the image to be adaptively corrected more reasonably and accurately, and thus more precisely restoring the real regional scene presented by the image.

Description

Technical Field

[0001] The present invention relates to the technical field of image restoration, and particularly relates to a geometric correction method and system for multi-spectral remote sensing data. Background Art

[0002] Image geometric correction refers to correcting geometric deformations caused by factors such as camera perspective and lens distortion, so as to restore the image to a more natural and accurate geometric form. Usually, such deformations can cause problems such as straight lines in the image looking curved and object proportions being distorted. The main purpose of geometric correction is to eliminate or reduce these visual errors. The radial inverse distortion model is one of the most common models for solving lens distortion, and it is used to describe the distortion phenomenon from the center of the image outwards. This distortion phenomenon makes the straight lines in the image no longer straight but curved.

[0003] Currently, when performing geometric correction on remote sensing image data through the radial distortion model, the position correction of the pixel points of the image is generally carried out with a fixed inverse distortion model order. However, since the image distortion is generally smaller in the central area of the image and more serious in the edge area of the image, using the same order of inverse distortion model for the entire image is likely to result in an unbalanced problem where the correction effect in the center of the image is good while the edge part may have insufficient correction or overcorrection, which is not conducive to accurately restoring the real regional scene presented by the image. Summary of the Invention

[0004] In order to solve the technical problem that using a fixed inverse distortion model order for image correction easily leads to unbalanced distortion correction and poor correction effect, the purpose of the present invention is to provide a geometric correction method and system for multi-spectral remote sensing data, and the specific technical solutions adopted are as follows:

[0005] The present invention provides a geometric correction method for multi-spectral remote sensing data, and the method includes:

[0006] Obtain the remote sensing image of the target area and determine multiple target partitions in the remote sensing image;

[0007] Determine the distortion degree of the target partition, and determine the model order weight of the target partition based on the distortion degree;

[0008] Determine the inverse distortion model order of the target partition according to the model order weight;

[0009] Use the inverse distortion model corresponding to the inverse distortion model order to perform distortion correction on the target partition.

[0010] Further, the step of determining multiple target partitions in the remote sensing image includes:

[0011] Taking the intersection point between the diagonals of the remote sensing image as the center and half of the diagonal length as the maximum circle radius, the remote sensing image is divided into a preset number of concentric circular target partitions.

[0012] Further, the step of determining the distortion degree of the target partition includes:

[0013] Determining the line curvature of each line in the target partition and the similarity of the bending directions of any two lines;

[0014] Using the line curvature and the similarity of the bending directions, the distortion degree of the target partition is calculated.

[0015] Further, the step of determining the model order weight of the target partition based on the distortion degree includes:

[0016] Determining the difference in distortion degree of the target partition relative to other partitions in all partitions;

[0017] Determining the number of partitions of all partitions and the target radius of the target partition;

[0018] Using the difference in distortion degree, the number of partitions, and the target radius, the model order weight of the target partition is calculated.

[0019] Further, the step of determining the anti-distortion model order of the target partition according to the model order weight includes:

[0020] According to the arrangement order of the target partitions from the inside to the outside, determining an order weight sequence composed of each model order weight;

[0021] Determining the position of the maximum difference between any two adjacent model order weights in the order weight sequence;

[0022] Using the model order weight as the region and the maximum difference position as the division condition, The function divides the order weight sequence into multiple sub-weight sequences;

[0023] Setting each target partition corresponding to the same sub-weight sequence to the same anti-distortion model order corresponding to the arrangement order;

[0024] Wherein, the maximum difference position includes a first order weight and a second order weight.

[0025] Further, the step of using the model order weight as the region and the maximum difference position as the division condition to divide the order weight sequence into multiple sub-weight sequences includes: The function divides the order weight sequence into multiple sub-weight sequences, including:

[0026] Traverse each order model weight as a region in the order weight sequence according to the described arrangement order;

[0027] Determine each pre-order model order weight before the first order weight and each post-order model order weight after and including the first order weight;

[0028] Utilize a function to divide each pre-order model order weight and each post-order model order weight into two different sub-weight sequences;

[0029] Take the sub-weight sequences as the order weight sequence and iteratively obtain multiple sub-weight sequences according to the iteration stop parameter.

[0030] Further, in the step of iteratively obtaining multiple sub-weight sequences according to the iteration stop parameter, the step of determining the iteration stop parameter includes:

[0031] Obtain the model order weight concentration degree of each sub-weight sequence formed by the current iteration and the average model order weight difference between any two sub-weight sequences;

[0032] Calculate the current iteration stop parameter by using the model order weight concentration degree and the average model order weight difference.

[0033] Further, the method further includes:

[0034] Determine the transition regions between each adjacent target partition;

[0035] In the case where the transition region is formed by adjacent target partitions corresponding to the same sub-weight sequence, set the transition region to the same anti-distortion model order as the adjacent target partition.

[0036] Further, after the step of determining the transition regions between each adjacent target partition, the method further includes:

[0037] In the case where the transition region is formed by adjacent target partitions corresponding to different sub-weight sequences, determine the inner partition and the outer partition corresponding to the transition region;

[0038] Determine the reference anti-distortion model order of the inner partition or the outer partition;

[0039] Determine the internal model order weight of the inner partition and the number of inner partitions of the sub-weight sequence to which the inner partition belongs;

[0040] Determine the external model order weight of the outer partition and the number of outer partitions of the sub-weight sequence to which the outer partition belongs;

[0041] The anti-distortion model order of the transition region is calculated by using the reference anti-distortion model order, the internal model order weight, the number of internal partitions, the external model order weight, and the number of external partitions.

[0042] The present invention also provides a geometric correction system for multi-spectral remote sensing data, which is used to implement the geometric correction method for multi-spectral remote sensing data as described in any one of the above; the system includes:

[0043] A region division module, configured to obtain a remote sensing image of a target area and determine a plurality of target partitions in the remote sensing image;

[0044] A weight calculation module, configured to determine the distortion degree of the target partition and determine the model order weight of the target partition based on the distortion degree;

[0045] An order calculation module, configured to determine the anti-distortion model order of the target partition according to the model order weight;

[0046] A model correction module, which uses the anti-distortion model corresponding to the anti-distortion model order to perform distortion correction on the target partition.

[0047] The present invention has the following beneficial effects:

[0048] Based on the discovery rule that the distortion degrees in different regions of the image are different, the present invention overcomes the problem of insufficient or overcorrection in some image regions when using an anti-distortion model with a fixed order to perform distortion correction on the whole image. By dividing the image into multiple sub-regions and fitting the distortion parameters separately for each sub-region, a more suitable anti-distortion model order is determined for the distortion states of different regions of the image, greatly improving the distortion correction accuracy of the image, enabling each region of the image to be adaptively corrected more reasonably and accurately, and thus more precisely restoring the real regional scene presented by the image. Description of the Drawings

[0049] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0050] Figure 1 It is a step flow chart of a geometric correction method for multi-spectral remote sensing data provided by an embodiment of the present invention;

[0051] Figure 2The detailed flowchart of step S2 in a geometric correction method for multi-spectral remote sensing data provided by an embodiment of the present invention;

[0052] Figure 3 The detailed flowchart of step S3 in a geometric correction method for multi-spectral remote sensing data provided by an embodiment of the present invention;

[0053] Figure 4 The schematic diagram of area division involved in a geometric correction method for multi-spectral remote sensing data provided by an embodiment of the present invention;

[0054] Figure 5 The structural schematic diagram of the hardware operating environment of a geometric correction device for multi-spectral remote sensing data involved in the solution of an embodiment of the present invention;

[0055] Figure 6 The framework structural schematic diagram of a geometric correction system for multi-spectral remote sensing data involved in the solution of an embodiment of the present invention. Detailed implementation manners

[0056] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following, in combination with the accompanying drawings and preferred embodiments, details the specific implementation manners, structures, features, and effects of a geometric correction method and system for multi-spectral remote sensing data proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0057] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs.

[0058] The following specifically describes the specific solutions of a geometric correction method and system for multi-spectral remote sensing data provided by the present invention with reference to the accompanying drawings.

[0059] Embodiment 1:

[0060] Regarding a geometric correction method for multi-spectral remote sensing data provided by the present invention, please refer to Figure 1 , which shows the flowchart of the steps of a geometric correction method for multi-spectral remote sensing data provided by an embodiment of the present invention.

[0061] The method includes:

[0062] Step S1, obtaining a remote sensing image of a target area and determining a plurality of target partitions in the remote sensing image;

[0063] In this embodiment, the remote sensing image of the target area can be obtained through aerial photography by a drone and grayscaled. Due to reasons such as the camera perspective and lens design, radial distortion may occur, which makes a straight line appear curved in the image, especially more obvious when approaching the image edge.

[0064] When interpreting the remote sensing of the distorted image, errors will occur. Therefore, geometric correction needs to be performed on the distorted image. Currently, the radial inverse distortion model is generally used to process the image. The points in the image will deviate from the center position of the image, and the degree of deviation is related to the distance from the center of the image. The formula of the radial inverse distortion model can usually be expressed as:

[0065] ;

[0066] In the formula, represents the distance from any pixel point after adjustment to the center (point) of the image, is the distance from this any pixel point to the center of the image before adjustment; it uses the calibrated radial distortion coefficient , the more coefficients, the higher the order of the inverse distortion model, and the greater the degree of correction for distortion. Applying it to the image, the coordinates of each pixel point are adjusted to restore it to its original position. The above is the prior art and will not be elaborated here.

[0067] Since the radial distortion is usually small in the central area of the image and more serious in the edge area of the image, there are local errors when correcting through a single fixed inverse distortion model, reducing the accuracy of correction. Therefore, the technical concept of this embodiment is mainly to set different inverse distortion model parameters for adaptive correction of different image regions.

[0068] After obtaining the remote sensing image, the remote sensing image can be divided into multiple target partitions according to needs. The division method is not limited here. The rectangular remote sensing image can be divided into equal parts, or the remote sensing image can be divided according to the actual distortion degree of different regions. Preferably, based on the law that the distortion degree generally tends to increase continuously from the center of the image to the outside, the remote sensing image can be divided into multiple target partitions in a concentric circle manner.

[0069] In a preferred embodiment, the step S1 includes:

[0070] Taking the intersection point between the diagonals of the remote sensing image as the center and half of the diagonal length as the maximum circle radius, the remote sensing image is divided into a preset number of concentric circle - shaped target partitions.

[0071] Please refer to Figure 4, To determine the corresponding anti-distortion model order for different distortion degrees, it is first necessary to divide the image into regions. Since the distortion degree of the central region of the generally captured remote sensing image is relatively small, and the distortion degree increases towards the edge region, in this embodiment, the image is first divided into regions. Here, the concentric circle division method is used, with the intersection of the two diagonals of the image as the center and half of the image diagonal as the maximum circle radius to divide multiple concentric circle regions. Taking Figure 4 as an example, this embodiment sets 9 target partitions. The specific number of target partitions can be set according to actual needs, and only an example is given here.

[0072] Figure 4 The solid concentric circles in are the region division lines, dividing the image into multiple target partitions. The dotted lines outside the solid lines represent the transition regions that can further smoothly connect the correction results of different sub-regions. The width of each concentric circle is the same, and the width of each transition region is the same. The distortion correction and smooth transition of the transition regions will be described in the subsequent embodiments below.

[0073] Step S2, determine the distortion degree of the target partition, and determine the model order weight of the target partition based on the distortion degree;

[0074] Specifically, the steps to determine the distortion degree of the target partition include:

[0075] Determine the line curvature of each line in the target partition and the similarity of the bending directions of any two lines;

[0076] Using the line curvature and the similarity of the bending directions, calculate the distortion degree of the target partition.

[0077] To determine the corresponding anti-distortion model order for different image regions, it is first necessary to analyze the distortion degree of the images in different regions. For regions with higher distortion degrees, a relatively higher anti-distortion model order is adopted, while for regions with lower distortion degrees, a relatively lower anti-distortion model order is adopted.

[0078] The distortion degree of an image is generally measured by the bending degree of the line features existing therein. If the bending directions of the lines in the image are the same and the bending degree is greater, it indicates that the distortion degree of the image in this region is greater. Therefore, based on the above principle, for any target partition, the possible lines are detected by the Sobel operator, and then for each target partition, there are:

[0079] ;

[0080] In the formula, represents the distortion degree of the target partition image. and respectively represent the lines in this partition with the line PCA (Principal Component Analysis) maximum principal direction vector of, representing the included angle between two direction vectors, being the number of lines in this partition, being the average value of the curvatures of all lines in this partition, being the maximum value of the curvatures of all lines in this partition.

[0081] Therefore, represents the similarity of the bending directions of any two lines in this partition. Since PCA can decompose the main trend of the lines, that is, it represents the bending direction of the lines, then the larger the value of, the more lines in this partition have similar bending directions, indicating a greater degree of image distortion in this partition. represents the degree of bending of the lines in this partition. Since more obvious distortion will produce more curved curves, therefore the larger the value of, the greater the average curvature of the lines in this partition, then indicating a greater degree of image distortion in this partition. In summary, the corresponding distortion degrees are determined for all concentric circle partitions for subsequent analysis.

[0082] Specifically, please refer to Figure 2 for the steps of determining the model order weight of the target partition based on the distortion degree, including:

[0083] Step S20, determining the distortion degree difference of the target partition relative to other partitions in all partitions;

[0084] Step S21, determining the number of partitions of all partitions and the target radius of the target partition;

[0085] Step S22, using the distortion degree difference, the number of partitions, and the target radius to calculate the model order weight of the target partition.

[0086] Since different target partitions correspond to different distortion degrees, in order to correct the distortion of the image, it is necessary to determine the model order of the corresponding image partition based on the inverse distortion model and through the difference in distortion degree. It is known that the higher the model order, the greater the distortion correction ability, and the greater the distortion degree of a certain partition of the image, the greater the distortion correction ability is required to adjust, and the smaller the distortion degree, the smaller the distortion correction ability is required. Therefore, the corresponding inverse distortion model orders can be determined for different partitions of the image for adaptive correction according to the above principle.

[0087] Since the distortion degree corresponding to the partitions closer to the image edge is greater, the model order weights corresponding to different partitions can be determined by combining the positions of different partitions and the corresponding distortion degrees, as shown below:

[0088] ;

[0089] In the formula, is the model order weight of the th partition of this image. is the distortion degree of the th partition, is the distortion degree of the remaining th partition, is the number of partitions of this image, and norm represents the linear normalization function. is the radius size (target radius) of the th partition. Then represents the distortion degree difference of the th partition relative to other partitions. The larger its value, the greater the distortion degree of this partition. Then a higher anti-distortion model order is required, and the corresponding model order weight is larger. Then, combined with the relative position of this partition in the image, that is, , The larger it is, the more the corresponding partition tends to the outer edge, and then the corresponding distortion degree is also larger, and the corresponding model order weight is larger. Finally, the model order weight corresponding to any partition is determined and the result is normalized to the range of .

[0090] Step S3, determine the anti-distortion model order of the target partition according to the model order weight;

[0091] Specifically, please refer to Figure 3 , the step S3 includes:

[0092] Step S30, determine the order weight sequence composed of each model order weight according to the arrangement order of the target partitions from the inside to the outside;

[0093] Step S31, determine the position of the maximum difference between any two adjacent model order weights in the order weight sequence;

[0094] Step S32, using the model order weight as the region and the maximum difference position as the division condition, divide the order weight sequence into multiple sub-weight sequences by using the function;

[0095] Further, step S32 includes:

[0096] Traverse each order model weight in the order weight sequence as the region according to the arrangement order;

[0097] The respective pre-order model order weights before determining the first-order weight and the post-order model order weights after and including the first-order weight;

[0098] Using a function to divide the respective pre-order model order weights and the respective post-order model order weights into two different sub-weight sequences;

[0099] Taking the sub-weight sequence as the order weight sequence, iteratively obtaining multiple sub-weight sequences according to the iteration stop parameter.

[0100] Due to the gradual change of the image distortion degree, the model order weights corresponding to the above different partitions show a gradually increasing trend from the inner circle to the outer circle. If the order of different anti-distortion models is allocated one by one according to the relative positions of different partitions, and if there is only a slight difference in the distortion degree between adjacent partitions, then a serious tearing feeling will occur at this position of the corrected image, and the purpose of accurate correction cannot be achieved. Therefore, it is necessary to determine the same or different orders based on the model order weights for different partitions, that is, to classify or group different target partitions, and the target partitions in the same category or group (the same sequence) are adapted to the order of the same anti-distortion model.

[0101] Next, it is necessary to analyze the difference degree of the model order weights between different partitions. If the difference degree of the model order weights between two adjacent partitions is larger, different model orders need to be allocated for correction. If the difference degree of the model order weights between two adjacent partitions is smaller, only the same model order needs to be allocated for correction. Therefore, a sequence (order weight sequence) is set according to the arrangement order of the model order weights of different partitions from the inside to the outside , and then iterative classification is carried out with the help of a function (function in excel), as follows:

[0102] ;

[0103] In the formula, represents the number of model order weights obtained by a certain classification and the corresponding sub-sequence, The parameter indicates that the algorithm starts to execute from , that is, through the algorithm traverses the sequence from the initial position of the sequence, The parameter indicates the position of the maximum difference between any two adjacent model order weights in the sequence (the maximum difference position includes the first-order weight and the second-order weight, the first-order weight is the larger order weight, and the second-order weight is the smaller order weight), that is, the differences between the values of any two adjacent values (the values of the model order weights) are calculated respectively, and then the two corresponding to the maximum difference are counted The number of elements before the first-order weight in the value (the number of elements, the number of model order weights, and the number of target partitions all represent the same quantity) is , and the elements before the first-order weight and including the first-order weight and those after it are respectively placed into two different sequences, and this method is continued to be used for classification, and the iteration is carried out in turn.

[0104] Among them, the steps of determining the iteration stop parameter include:

[0105] Obtain the model order weight concentration degree of each sub-weight sequence formed in the current iteration and the average model order weight difference between any two sub-weight sequences;

[0106] Use the model order weight concentration degree and the average model order weight difference to calculate the current iteration stop parameter.

[0107] During the iteration process, the sequence is continuously divided into multiple sub-sequences, and the model order weights between the multiple sub-sequences are relatively large. However, if this method is iterated infinitely, it will cause the model order weight in each to be divided into an independent sequence, thus having the same effect as allocating the order one by one according to the relative position of the partition. Therefore, in order to avoid the above situation, an iteration stop condition needs to be defined.

[0108] The purpose of this embodiment is to divide the model order weights with similar values into one sequence, and divide the model order weights with relatively large differences into different sequences, so as to specifically determine the corresponding model order. Therefore, the weight differences inside and outside the sequence generated in each iteration round can be analyzed to determine the round of iteration that needs to stop, as follows:

[0109] ;

[0110] In the formula, is the iteration stop parameter corresponding to the current round of iteration. and are respectively the average model order weights of the th sequence and the th sequence formed after this round of iteration, is the number of sequences formed in the th round of iteration. represents the average value of the standard deviations of the model order weights within all sequences in the th round of iteration. Then represents the average model order weight difference between any two sequences in this round of iteration. The larger its value, the better the division effect of the sequence. Then the iteration needs to be stopped in time, and the corresponding stop parameter is larger. Measure the concentration of the weights of each model order within each sequence. The smaller the value, the more similar the weights of each model order within the sequence, and the better the partitioning effect. Correspondingly, the larger the stopping parameter. Finally, use function to normalize the result to range and select the for the first time when it is greater than the preset stopping threshold during the iterative process to stop the iterative calculation. The preset stopping threshold can be set as needed, for example, it can be 0.88.

[0111] Step S33: Set each target partition corresponding to the same sub-weight sequence to the same anti-distortion model order corresponding to the arrangement order;

[0112] So far, the weights of the model orders in different partitions are classified through the above steps. Then, the weights of the model orders in each classification are relatively similar, and the differences in the weights of the model orders between different classifications are relatively large. For multiple model order weights in the same classification, the same anti-distortion model order can be set for the corresponding partitions. At the same time, according to the increasing trend of the distortion degree from the inside to the outside of the partition and the corresponding arrangement order, different anti-distortion model orders are set for different classifications. For example: If the classification result is , , , then for each partition image in , set the first-order anti-distortion model, that is . For each partition image in , set the second-order anti-distortion model, that is , and so on.

[0113] Step S4: Use the anti-distortion model corresponding to the anti-distortion model order to perform distortion correction on the target partition.

[0114] So far, through the anti-distortion models corresponding to different partitions in the remote sensing image, the target partition images are processed, and the correct positions of the corresponding pixel points in the image are calculated to correct the distortion of the image, so as to fully consider the different distortion characteristics of each region of the image and ensure the geometric accuracy of the overall image.

[0115] In addition, in a preferred embodiment, the method further includes:

[0116] Determine the transition zones between adjacent target partitions;

[0117] When the transition zone is formed by adjacent target partitions corresponding to the same sub-weight sequence, set the transition zone to the same anti-distortion model order as the adjacent target partition.

[0118] After determining the corresponding anti-distortion model for different partitioned images, since the anti-distortion models of different orders have different distortion correction capabilities for images, there is likely to be an overly unnatural problem at the junction of the partitioned images. Therefore, a transition area is also set between any two adjacent partitions during the partitioning process. The purpose is to set an appropriate anti-distortion model in the transition area to connect the corrected effects of the two adjacent target partitions, making the transition smoother.

[0119] Since the above embodiments classify the model order weights of different partitions, and the model orders corresponding to all partitions within the same classification sequence are the same, the transition areas between all partitions within the same classification sequence can be filled with the same model order as that of the sequence.

[0120] In another preferred embodiment, after the step of determining the transition area between each adjacent target partition, the method further includes:

[0121] When the transition area is formed by the adjacency of target partitions corresponding to different sub-weight sequences, determining the inner partition and the outer partition corresponding to the transition area;

[0122] Determining the reference anti-distortion model order of the inner partition or the outer partition;

[0123] Determining the inner model order weight of the inner partition and the number of inner partitions of the sub-weight sequence to which the inner partition belongs;

[0124] Determining the outer model order weight of the outer partition and the number of outer partitions of the sub-weight sequence to which the outer partition belongs;

[0125] Using the reference anti-distortion model order, the inner model order weight, the number of inner partitions, the outer model order weight, and the number of outer partitions, calculating the anti-distortion model order of the transition area.

[0126] For the transition areas at the junctions of partitions corresponding to different classification sequences, the anti-distortion model order of the corresponding transition area needs to be determined according to the number of partitions and the difference in model order weights. If one side of the inner and outer sides of the transition area has more partitions and a greater model order weight in the corresponding classification sequence, then the possibility of setting the anti-distortion model order of this transition area to the model order corresponding to this partition is greater, as shown below:

[0127] ;

[0128] In the formula, is the anti-distortion model order of the transition area formed by the partitions corresponding to a certain different classification sequence. is the anti-distortion model order of the inner partition corresponding to this transition area, and can also be replaced by , that is, the order of the anti-distortion model of the outer partition, and one of the two is selected as the reference anti-distortion model order; This is the number of partitions of the inner partition corresponding to this transition region, This is the model order weight of the inner partition corresponding to this transition region. This is the number of partitions of the outer partition corresponding to this transition region, This is the model order weight of the outer partition corresponding to this transition region. represents the rounding function. Then represents the difference in the number of partitions and the model order weight between the inner and outer partitions corresponding to this transition region. Taking it as a weighting factor, the larger its value indicates that the weight of the outer partition (referenced by or the inner partition (referenced by is larger. Then the anti-distortion model order of the transition region should be biased towards the order of the outer partition or the inner partition respectively, and vice versa, it should be biased towards the order of the inner partition or the outer partition respectively.

[0129] By the above method, the corresponding anti-distortion model order is determined for the transition region at the junction of the partitions corresponding to different classifications for connection correction, thereby realizing a smooth transition at the junction of the partitions.

[0130] Embodiment 2:

[0131] The embodiment of the present invention also proposes a geometric correction device for multi-spectral remote sensing data. The geometric correction device for multi-spectral remote sensing data can be a data processing terminal such as a drone, a computer, a server, etc.

[0132] As Figure 5 shown, Figure 5 is a schematic structural diagram of the hardware operating environment of the geometric correction device for multi-spectral remote sensing data involved in the embodiment solution of the present invention.

[0133] As Figure 5As shown in the figure, the geometric correction device for multi-spectral remote sensing data may include: a processor 1001, such as a CPU, a network interface 1004, a user interface 1003, a memory 1005, and a communication bus 1002. Among them, the communication bus 1002 is used to realize the connection and communication between these components. The user interface 1003 may include a display (Display) and an input unit such as a control panel. Optionally, the user interface 1003 may further include a standard wired interface and a wireless interface. The network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a WIFI interface). The memory 1005 may be a high-speed RAM memory or a stable memory (non-volatile memory), such as a disk memory. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001. As a computer storage medium, the memory 1005 may include a geometric correction program for multi-spectral remote sensing data.

[0134] Those skilled in the art can understand that Figure 5 the hardware structure shown in the figure does not constitute a limitation on the device, and it may include more or fewer components than shown in the figure, or combine certain components, or have a different component layout.

[0135] Continuing to refer to Figure 5 , Figure 5 the memory 1005 as a computer-readable storage medium in the figure may include an operating system, a user interface module, a network communication module, and a geometric correction program for multi-spectral remote sensing data.

[0136] In Figure 5 , the network communication module is mainly used to connect to the server and can communicate with the server for data. And the processor 1001 can call the geometric correction program for multi-spectral remote sensing data stored in the memory 1005 and execute the steps in each of the above embodiments.

[0137] Based on the above hardware structure of the geometric correction device for multi-spectral remote sensing data, each embodiment for realizing the geometric correction of multi-spectral remote sensing data of the present invention is implemented.

[0138] In addition, the present invention also provides a geometric correction system for multi-spectral remote sensing data. Please refer to Figure 6 , the geometric correction system for multi-spectral remote sensing data includes:

[0139] A region division module A10, configured to obtain a remote sensing image of a target area and determine a plurality of target partitions in the remote sensing image;

[0140] A weight calculation module A20, configured to determine the distortion degree of the target partition and determine the model order weight of the target partition based on the distortion degree;

[0141] An order calculation module A30 is configured to determine the anti-distortion model order of the target partition according to the model order weight.

[0142] A model correction module A40 corrects the distortion of the target partition by using the anti-distortion model corresponding to the anti-distortion model order.

[0143] Further, the region division module A10 is further configured to:

[0144] Taking the intersection point between the diagonals of the remote sensing image as the center and half of the diagonal length as the maximum circle radius, the remote sensing image is divided into a preset number of concentric circular target partitions.

[0145] Further, the weight calculation module A20 is further configured to:

[0146] Determine the line curvature of each line in the target partition and the similarity of the bending directions of any two lines;

[0147] Using the line curvature and the bending direction similarity, calculate the distortion degree of the target partition.

[0148] Further, the weight calculation module A20 is further configured to:

[0149] Determine the difference in distortion degree of the target partition relative to other partitions in all partitions;

[0150] Determine the number of partitions of all partitions and the target radius of the target partition;

[0151] Using the distortion degree difference, the number of partitions, and the target radius, calculate the model order weight of the target partition.

[0152] Further, the order calculation module A30 is further configured to:

[0153] According to the arrangement order of the target partitions from the inside to the outside, determine the order weight sequence composed of each model order weight;

[0154] Determine the position of the maximum difference between any two adjacent model order weights in the order weight sequence;

[0155] Taking the model order weight as the region and the maximum difference position as the division condition, use The function divides the order weight sequence into multiple sub-weight sequences;

[0156] Set each target partition corresponding to the same sub-weight sequence to the same anti-distortion model order corresponding to the arrangement order;

[0157] Among them, the maximum difference includes a first-order weight and a second-order weight.

[0158] Further, the order calculation module A30 is further configured to:

[0159] Traverse each order model weight in the order weight sequence as a region according to the arrangement order;

[0160] Determine each pre-order model order weight before the first-order weight and the post-order model order weight after and including the first-order weight;

[0161] Using a function, divide each pre-order model order weight and each post-order model order weight into two different sub-weight sequences;

[0162] Take the sub-weight sequence as the order weight sequence, and iteratively obtain multiple sub-weight sequences according to the iteration stop parameter.

[0163] Further, the order calculation module A30 is further configured to:

[0164] Obtain the model order weight concentration degree of each sub-weight sequence formed by the current iteration and the average model order weight difference between any two sub-weight sequences;

[0165] Use the model order weight concentration degree and the average model order weight difference to calculate the current iteration stop parameter.

[0166] Further, the order calculation module A30 is further configured to:

[0167] Determine the transition area between each adjacent target partition;

[0168] When the transition area is formed by the adjacency of target partitions corresponding to the same sub-weight sequence, set the transition area to the same anti-distortion model order as the adjacent target partition.

[0169] Further, the order calculation module A30 is further configured to:

[0170] When the transition area is formed by the adjacency of target partitions corresponding to different sub-weight sequences, determine the inner partition and the outer partition corresponding to the transition area;

[0171] Determine the reference anti-distortion model order of the inner partition or the outer partition;

[0172] Determine the internal model order weight of the inner partition and the number of inner partitions of the sub-weight sequence to which the inner partition belongs;

[0173] Determine the external model order weight of the external partition and the number of external partitions of the sub-weight sequence to which the external partition belongs;

[0174] Calculate the anti-distortion model order of the transition zone by using the reference anti-distortion model order, the internal model order weight, the number of internal partitions, the external model order weight, and the number of external partitions.

[0175] The specific implementation manner of the geometric correction system for multi-spectral remote sensing data according to the present invention is basically the same as each embodiment of the above-mentioned geometric correction method for multi-spectral remote sensing data, and will not be elaborated here.

[0176] In addition, the present invention also provides a computer-readable storage medium. A geometric correction program for multi-spectral remote sensing data is stored on the computer-readable storage medium of the present invention. When the geometric correction program for multi-spectral remote sensing data is executed by a processor, the steps of the geometric correction method for multi-spectral remote sensing data as described above are implemented.

[0177] Among them, the method implemented when the geometric correction program for multi-spectral remote sensing data is executed can refer to each embodiment of the geometric correction method for multi-spectral remote sensing data of the present invention, and will not be elaborated here.

[0178] It should be noted that the above sequence of embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In some embodiments, multi-tasking and parallel processing are also possible or may be advantageous.

[0179] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between each embodiment can be referred to each other. Each embodiment focuses on the differences from other embodiments.

[0180] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-usable program codes.

[0181] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and combinations of flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to the processors of general purpose computers, special purpose computers, embedded processors, or other programmable data processing devices to produce a machine, such that the instructions executed by the processors of the computer or other programmable data processing devices produce means for implementing the functions specified in the Figure 1 one or more flows and / or blocks Figure 1 one or more blocks

[0182] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including instruction means for implementing the functions specified in the Figure 1 one or more flows and / or blocks Figure 1 one or more blocks

[0183] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operational steps are performed on the computer or other programmable device to produce a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in the Figure 1 one or more flows and / or blocks Figure 1 one or more blocks

[0184] Although the preferred embodiments of the present invention have been described, additional changes and modifications can be made by those skilled in the art once they learn the basic creative concept.

[0185] The above are only the preferred embodiments of the present invention, and do not limit the scope of the present invention. Any equivalent structural transformation made using the description and drawings of the present invention under the inventive concept of the present invention, or direct / indirect application in other related technical fields, is included in the protection scope of the present invention.

Claims

1. A geometric correction method for multispectral remote sensing data, characterized in that: The method comprises: Acquire a remote sensing image of a target area, and determine a plurality of target subareas in the remote sensing image; Determining a degree of distortion of the target partition, and determining a model order weight of the target partition based on the degree of distortion; Determining the anti-distortion model order of the target partition according to the model order weight; Using an anti-distortion model corresponding to the order of the anti-distortion model, performing distortion correction on the target partition; The step of determining the model order weight of the target partition based on the degree of distortion comprises: Determining a difference in degree of distortion of the target partition relative to other partitions among all the partitions; Determining the number of partitions of all partitions and the target radius of the target partition; Calculate the model order weight of the target partition by using the distortion degree difference, the number of partitions and the target radius; The step of determining the anti-distortion model order of the target partition according to the model order weight comprises: Determining an order weight sequence composed of the order weights of each model according to the order of the target partitions from inside to outside; Determine the maximum difference between any two adjacent model order weights in the order weight sequence; The model order weight is used as the region and the maximum difference is used as the division condition. The function divides the order weight sequence into multiple sub-weight sequences; Setting each target partition corresponding to the same sub-weight sequence to the same anti-distortion model order corresponding to the arrangement order; The maximum difference includes a first-order weight and a second-order weight.

2. The geometric correction method for multispectral remote sensing data according to claim 1, characterized in that: The step of determining a plurality of target subareas in the remote sensing image comprises: The remote sensing image is divided into a preset number of concentric circle target partitions with the intersection of the diagonals of the remote sensing image as the center of the circle and half the length of the diagonals as the maximum circle radius.

3. The geometric correction method for multispectral remote sensing data according to claim 1, characterized in that: The step of determining the degree of distortion of the target partition comprises: Determining the line curvature of each line in the target partition and the similarity of the curvature directions of any two lines; The distortion degree of the target partition is calculated by using the line curvature and the bending direction similarity.

4. The geometric correction method for multispectral remote sensing data according to claim 1, characterized in that: The model order weight is used as the region and the maximum difference is used as the division condition. The step of dividing the function order weight sequence into multiple sub-weight sequences includes: According to the arrangement order, traverse the weights of each order model as a region in the order weight sequence; Determine the order weights of each preceding model before the first order weight and the order weights of the subsequent models after the first order weight and including the first order weight; use Function, which divides each preceding model order weight and each succeeding model order weight into two different sub-weight sequences; The sub-weight sequence is used as the order weight sequence, and multiple sub-weight sequences are iterated according to the iteration stop parameter.

5. The geometric correction method for multispectral remote sensing data according to claim 4, characterized in that: In the step of iteratively obtaining a plurality of sub-weight sequences according to the iteration stop parameter, the step of determining the iteration stop parameter comprises: Obtain the model order weight concentration of each sub-weight sequence formed in the current iteration and the average model order weight difference between any two sub-weight sequences; The current iteration stop parameter is calculated using the model order weight concentration and the average model order weight difference.

6. The geometric correction method for multispectral remote sensing data according to claim 4, characterized in that: The method further comprises: Determine the transition area between each adjacent target partition; In the case where the transition region is formed adjacent to target partitions corresponding to the same sub-weight sequence, the transition region is set to have the same anti-distortion model order as that of the adjacent target partition.

7. The geometric correction method for multispectral remote sensing data according to claim 6, characterized in that: After the step of determining the transition area between each adjacent target partition, the method further comprises: In the case where the transition zone is formed by adjacent target partitions corresponding to different sub-weight sequences, determining an inner partition and an outer partition corresponding to the transition zone; Determining a reference anti-distortion model order of the inner partition or the outer partition; Determine the internal model order weight of the internal partition and the number of internal partitions of the sub-weight sequence to which the internal partition belongs; Determine the external model order weight of the external partition and the number of external partitions of the sub-weight sequence to which the external partition belongs; The anti-distortion model order of the transition zone is calculated using the reference anti-distortion model order, the internal model order weight, the number of internal partitions, the external model order weight, and the number of external partitions.

8. A geometric correction system for multispectral remote sensing data, characterized in that: The system is used to implement the geometric correction method for multispectral remote sensing data as described in any one of claims 1 to 7; the system comprises: A region division module is used to obtain a remote sensing image of a target area and determine a plurality of target regions in the remote sensing image; A weight calculation module, used to determine the degree of distortion of the target partition, and determine the model order weight of the target partition based on the degree of distortion; An order calculation module, used to determine the anti-distortion model order of the target partition according to the model order weight; The model correction module uses the anti-distortion model corresponding to the order of the anti-distortion model to perform distortion correction on the target partition.

Images