Image processing measurement method and system based on unmanned aerial vehicle radar and oblique photography

Abstract

The application discloses an image processing measurement method and system based on a UAV radar and oblique photography, relates to the technical field of oblique photography image measurement, and is used for solving the problem of reduced accuracy of region division and abnormal detection results, obtaining high-resolution forest images through UAV oblique photography, generating vegetation distribution characteristics by combining radar echo amplitude to calculate vegetation change gradient, extracting the ground under the forest and generating uniform extrapolation points, laying out an inclinometer to obtain elevation and coordinates, calculating local slope and tree crown fluctuation amplitude, constructing terrain survey features and dividing regions, counting the proportion of the ground under each region, evaluating the vegetation coverage state of the region and screening abnormal regions, marking and outputting the extrapolated ground points of the abnormal regions, realizing high-precision survey of forest areas, accurately reflecting the complexity of the terrain and the vegetation distribution characteristics, providing reliable data support for forest management and terrain analysis, and improving survey efficiency and precision.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of oblique photography image measurement, and more particularly, to an image processing measurement method and system based on unmanned aerial vehicle radar and oblique photography. BACKGROUND

[0002] The terrain of forest areas is complex, and the vegetation is densely distributed. Traditional surveying and mapping methods rely on manual reconnaissance or single remote sensing images for data collection. However, manual reconnaissance is not only time-consuming and labor-intensive but also limited by terrain, making it difficult to obtain accurate data on the ground under the forest. Single optical remote sensing images are severely obstructed by tree canopies, and cannot directly reflect the elevation information and spatial distribution of the ground under the forest, which can easily lead to a decrease in surveying and mapping accuracy.

[0003] The prior art has the following disadvantages:

[0004] Currently, although the existing measurement methods based on unmanned aerial vehicle aerial photography or laser radar can improve data collection efficiency, oblique photography images only provide surface image gray scale information, cannot effectively distinguish between tree canopies and ground under the forest, lack image gray scale feature cooperative processing, leading to inaccurate ground point extraction, insufficient expression of vegetation distribution characteristics, decreased accuracy of region division and anomaly detection results, and affecting subsequent terrain correction and key surveying and mapping point placement work. Therefore, an image processing measurement method and system based on unmanned aerial vehicle radar and oblique photography are proposed.

[0005] The above information disclosed in the background section is only intended to enhance the understanding of the background of the present disclosure, and therefore it can include information that does not constitute the prior art known to those of ordinary skill in the art. SUMMARY

[0006] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present application provide an image processing measurement method and system based on unmanned aerial vehicle radar and oblique photography, which uses image gray scale value and radar echo amplitude fusion to construct vegetation distribution characteristics and ground under the forest extrapolation point generation technology to solve the problems raised in the above background.

[0007] To achieve the above-mentioned purpose, the present application provides the following technical scheme, an image processing measurement method based on unmanned aerial vehicle radar and oblique photography, comprising the following steps:

[0008] Step S1: When surveying a forest area, image gray scale values of each pixel point of the forest area are collected by oblique photography, radar echo amplitude values of the pixel points are obtained by radar detection, and vegetation change gradient is calculated according to the image gray scale values;

[0009] Step S2: generating a vegetation distribution feature of each pixel point by integrating the vegetation change gradient and the radar echo amplitude value, identifying and marking each pixel point according to the vegetation distribution feature to obtain the understory ground of the forest area, and generating a uniform extrapolated ground point based on the understory ground;

[0010] Step S3: detecting the forest land slope and the tree crown fluctuation amplitude of the forest area, calculating the steep slope proportion of the forest area according to the forest land slope, evaluating the topographic survey feature of the forest area in combination with the tree crown fluctuation amplitude, and regionally dividing the forest area according to the topographic survey feature;

[0011] Step S4: integrating the understory ground proportion in the divided region, evaluating the vegetation coverage state of the divided region based on the understory ground proportion, screening the divided region according to the vegetation coverage state, and marking to obtain an abnormal region and outputting the extrapolated ground point of the abnormal region.

[0012] In a preferred embodiment, in step S1, when the forest area is surveyed, the unmanned aerial vehicle carries a tilt camera to image different angles of the forest area to obtain forest images;

[0013] The forest images are processed by gray scale to obtain gray scale images, and each pixel point in the gray scale image corresponds to a gray scale value;

[0014] The pixel points of the forest images at different angles are one-to-one corresponding, and the image gray scale values of the pixel points at the same position are combined into a pixel gray scale set;

[0015] The median of the pixel gray scale set is taken as the gray scale median value, and the mean of the pixel gray scale value set is calculated as the gray scale mean value;

[0016] The average of the gray scale median value and the gray scale mean value is calculated as the image gray scale value of the pixel point.

[0017] In a preferred embodiment, in step S1, the forest area is scanned by a radar detection device carried by the unmanned aerial vehicle, electromagnetic waves are emitted to the forest area, and echo signals are received;

[0018] According to the echo signal, the signal strength value of each pixel point is extracted as the radar echo amplitude value;

[0019] Taking each pixel point as a center pixel point, the adjacent pixel points of the center pixel point are combined into a neighborhood set, and the image gray scale values of the center pixel point and the pixel points in the neighborhood set are subtracted to obtain each neighborhood gray scale difference value;

[0020] The standard deviation of each neighborhood gray scale difference value is calculated as the vegetation change gradient.

[0021] In a preferred embodiment, in step S2, the product of the normalized vegetation change gradient and the normalized radar echo amplitude of the pixel point is taken as the vegetation distribution feature of the pixel point.

[0022] The vegetation distribution features of all pixel points in the forest image are combined into a vegetation feature set.

[0023] The median of the vegetation feature set is taken as the distribution feature median, and the absolute value of the difference between the distribution feature median and each vegetation distribution feature is taken to obtain each absolute deviation. The median of each absolute deviation is multiplied by a preset adjustment coefficient to obtain the distribution feature amplitude.

[0024] In a preferred embodiment, in step S2, the difference between the distribution feature median and the distribution feature amplitude is taken to obtain the distribution feature threshold.

[0025] If the vegetation distribution feature of the pixel point is less than the distribution feature threshold, the pixel point is marked as an understory ground pixel point; otherwise, the pixel point is not marked.

[0026] Four understory ground pixel points that form a regular rectangular structure in spatial position are selected as corner points, and an understory ground grid cell is formed based on the corner points.

[0027] A uniform extrapolated ground point is generated by fitting the pixel points of the understory ground grid cell using a bilinear interpolation method.

[0028] In a preferred embodiment, in step S3, the center of a grid region formed by a predetermined grid spacing is taken as a measurement point to arrange a ground inclinometer, the elevation and horizontal coordinates of each measurement point are collected, and a local slope value is calculated.

[0029] When the local slope value is greater than or equal to a preset slope threshold, the grid region where the measurement point is located is defined as a steep slope region.

[0030] Otherwise, it is not defined.

[0031] The ratio of the number of measurement points in the grid region defined as the steep slope region to the total number of measurement points is taken as the steep slope proportion.

[0032] In a preferred embodiment, in step S3, a three-dimensional displacement sensor is arranged at the crown position of the measurement point to monitor the instantaneous displacement of the crown.

[0033] The difference between the maximum value and the minimum value of the instantaneous displacement of the crown in a preset observation period is taken as the crown fluctuation amplitude.

[0034] The steep slope proportion and the crown fluctuation amplitude are normalized to form a forest terrain feature vector, and the terrain survey feature of each measurement point is output through a logistic regression algorithm.

[0035] In a preferred embodiment, in step S3, if the topographic survey feature is greater than or equal to the preset survey feature threshold, the survey point is defined as a target area survey point;

[0036] Conversely, the survey point is not defined.

[0037] Based on the connectivity analysis, the target area survey points with a distance less than the preset distance are connected to form a continuous area as a division area of the forest area.

[0038] In a preferred embodiment, in step S4, the number of pixel points of the forest floor and the division area are obtained by tilt photography of the unmanned aerial vehicle, the number of pixel points is converted into a spatial area, and the area of the forest floor of each division area and the total area of each division area are obtained.

[0039] The ratio of the area of the forest floor of each division area to the area of each division area is taken as the forest floor coverage ratio of each division area.

[0040] When the forest floor coverage ratio is lower than the preset coverage ratio threshold, it is determined that the vegetation coverage state of the division area is abnormal coverage.

[0041] Conversely, the vegetation coverage state of the division area is normal coverage.

[0042] The division area with abnormal vegetation coverage is screened and marked as an abnormal area, and the extrapolated ground points of the abnormal area are output for subsequent survey and correction.

[0043] The image processing measurement system based on unmanned aerial vehicle radar and tilt photography includes a data acquisition module, a vegetation recognition module, a topographic analysis module, and a region evaluation module, and the functions of each module are as follows:

[0044] The data acquisition module is used to acquire the image gray value of each pixel point of the forest area by tilt photography when the forest area is surveyed, and to acquire the radar echo amplitude value of the pixel point by radar detection, and to calculate the vegetation change gradient according to the image gray value.

[0045] The vegetation recognition module is used to generate the vegetation distribution characteristics of each pixel point by comprehensively considering the vegetation change gradient and the radar echo amplitude value, to identify and mark each pixel point according to the vegetation distribution characteristics to obtain the forest floor of the forest area, and to generate uniform extrapolated ground points based on the forest floor.

[0046] The topographic analysis module is used to detect the forest land slope and the tree crown fluctuation amplitude of the forest area, to calculate the steep slope ratio of the forest area according to the forest land slope, to evaluate the topographic survey feature of the forest area in combination with the tree crown fluctuation amplitude, and to divide the forest area according to the topographic survey feature.

[0047] The regional evaluation module is used for evaluating the under-forest ground area ratio in each division region, evaluating the vegetation coverage state of each division region based on the under-forest ground area ratio, screening the division regions according to the vegetation coverage state, marking the abnormal regions and outputting the extrapolated ground points of the abnormal regions.

[0048] Technical effects and advantages of the present application:

[0049] The present application obtains high-resolution image data of the forest area through the tilt photography of the unmanned aerial vehicle, simultaneously obtains the echo amplitude of each pixel point through the radar detection, calculates the vegetation change gradient based on the image gray value, forms the vegetation distribution feature, identifies and marks each pixel point and extracts the under-forest ground by using the vegetation distribution feature, generates the uniform extrapolated ground point on the basis, obtains the forest height and horizontal coordinate through the layout of the ground inclinometer, calculates the local slope value, records the instantaneous displacement of the tree crown in the preset observation period, further calculates the tree crown fluctuation amplitude, comprehensively constructs the topographic survey feature by combining the steep slope ratio and the tree crown fluctuation amplitude, divides the forest area into regions, forms the continuous and quantifiable division regions, counts the under-forest ground area ratio of each division region through the unmanned aerial vehicle image, evaluates the vegetation coverage state of the region and screens the abnormal regions, marks and outputs the extrapolated ground points of the abnormal regions, provides the key reference for the subsequent survey and correction, realizes the high-precision survey of the forest area, accurately reflects the topographic complexity and the vegetation distribution feature, provides reliable data support for the forest management and topographic analysis, and improves the survey efficiency and precision. BRIEF DESCRIPTION OF DRAWINGS

[0050] Figure 1 The method flowchart of the image processing and measurement method based on the unmanned aerial vehicle radar and tilt photography of the present application.

[0051] Figure 2 The module schematic diagram of the image processing and measurement system based on the unmanned aerial vehicle radar and tilt photography of the present application. DETAILED DESCRIPTION

[0052] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the present application.

[0053] The present application obtains high-resolution image data of the forest area through unmanned aerial vehicle oblique photography, simultaneously obtains the echo amplitude of each pixel point through radar detection, and forms the vegetation distribution feature based on the image gray value calculation vegetation change gradient, identifies and marks each pixel point using the vegetation distribution feature, and extracts the ground under the forest, generates uniform extrapolated ground points on this basis, obtains the forest elevation and horizontal coordinates through the layout of the ground inclinometer, calculates the local slope value, records the instantaneous displacement of the tree crown in the preset observation period, and then calculates the tree crown fluctuation amplitude, and the steep slope ratio and the tree crown fluctuation amplitude are comprehensively constructed to form the terrain survey feature, and the forest area is divided into regions to form a continuous and quantifiable division region, and the ground under the forest in each division region is counted through unmanned aerial vehicle image to evaluate the vegetation coverage state of the region and screen the abnormal region, mark and output the extrapolated ground points of the abnormal region, provide a key reference for subsequent survey and correction, and realize high-precision survey of the forest area.

[0054] Embodiment 1

[0055] Please refer to Figure 1 , the image processing measurement method based on unmanned aerial vehicle radar and oblique photography, comprising the following steps:

[0056] Step S1: when surveying the forest area, the image gray value of each pixel point of the forest area is collected through oblique photography, the radar echo amplitude of the pixel point is obtained through radar detection, and the vegetation change gradient is calculated according to the image gray value;

[0057] Step S2: generate the vegetation distribution feature of each pixel point by comprehensively considering the vegetation change gradient and the radar echo amplitude, identify and mark each pixel point according to the vegetation distribution feature to obtain the ground under the forest in the forest area, and generate uniform extrapolated ground points based on the ground under the forest;

[0058] Step S3: detect the forest land slope and the tree crown fluctuation amplitude of the forest area, calculate the steep slope ratio of the forest area according to the forest land slope, evaluate the terrain survey feature of the forest area in combination with the tree crown fluctuation amplitude, and divide the forest area into regions according to the terrain survey feature;

[0059] Step S4: integrate the ground under the forest in each division region, evaluate the vegetation coverage state of the division region based on the ground under the forest, screen the division region according to the vegetation coverage state, mark and output the extrapolated ground points of the abnormal region.

[0060] The specific implementation is as follows:

[0061] In step S1, when surveying the forest area, the unmanned aerial vehicle carries an oblique photography camera to image different angles of the forest area to obtain a forest image.

[0062] The forest image is processed by gray scale to obtain a gray scale image, and each pixel point in the gray scale image corresponds to a gray scale value;

[0063] The pixel points of the forest images at different angles are one-to-one corresponding, the image gray scale values of the pixel points at the same position are combined as a pixel gray scale set, the median of the pixel gray scale set is taken as a gray scale median value, and the average of the pixel gray scale value set is calculated as a gray scale average value.

[0064] It should be explained that the tilt camera refers to a photographic imaging device arranged obliquely relative to the vertical direction, including multiple lenses arranged at different angles, for example, a normal lens and tilt lenses in front, back and left and right directions, used to obtain forest images at different angles in the forest area; the gray scale processing refers to an image processing operation of converting an input color image into a single-channel gray scale image.

[0065] The average of the gray scale median value and the gray scale average value is calculated as an image gray scale value of the pixel point, and the image gray scale value reflects the optical reflection intensity at the corresponding pixel point;

[0066] The radar detection device carried by the unmanned aerial vehicle performs radar scanning on the forest area, emits electromagnetic waves to the forest area, the receiving module of the radar detection device receives the echo signal and performs frequency conversion processing to obtain an intermediate frequency signal, the high-frequency echo signal is input to the analog-to-digital conversion module to convert it into a digital signal, and the signal processing module processes the digital signal to extract the signal intensity value of the pixel point as the radar echo amplitude value, which refers to the intensity of the reflected or scattered electromagnetic wave signal.

[0067] It should be explained that the radar detection device is a device that uses electromagnetic wave propagation and reflection principles to perceive and measure the surface structure or object distribution of the measured area, including a transmitting module, a receiving module, a frequency conversion module, an analog-to-digital conversion module, and a signal processing module; in this embodiment, the position and attitude information of the radar detection device is obtained by the positioning unit of the unmanned aerial vehicle, the accurate reflection position coordinates of the echo signal in the geographical space are calculated by the radar geometric positioning model according to the round-trip propagation time of the electromagnetic wave combined with the position and attitude information, the reflection position coordinates of the echo signal in the geographical space are obtained, and are matched with the pixel points of the forest image.

[0068] Each pixel point is taken as a center pixel point, the adjacent pixel points of the center pixel point are combined as a neighborhood set, and the image gray scale values of the center pixel point and the pixel points in the neighborhood set are subtracted to obtain each neighborhood gray scale difference value;

[0069] The standard deviation of each neighborhood gray scale difference value is taken as a vegetation change gradient.

[0070] The vegetation change gradient reflects the vegetation change of the pixel point, and the change amplitude of the vegetation coverage state at the pixel point is represented by analyzing the difference between the image gray value of the central pixel point and the adjacent pixel point.

[0071] In step S2, the product of the vegetation change gradient and the radar echo amplitude of the pixel point after standardization processing is taken as the vegetation distribution feature of the pixel point.

[0072] When the vegetation change gradient is small, it indicates that the vegetation coverage state of the pixel point is relatively uniform, and the distribution of the ground under the forest is more continuous, and the vegetation distribution feature is smaller. When the radar echo amplitude is small, it indicates that the electromagnetic wave return energy at the ground under the forest is weak, and the ground under the forest is covered and shielded by more vegetation, and the vegetation distribution feature is smaller.

[0073] The vegetation distribution features of all pixel points in the forest image are combined into a vegetation feature set, the median of the vegetation feature set is taken as the distribution feature median, the absolute value of the difference between the distribution feature median and each vegetation distribution feature is obtained, and the median of each absolute deviation is multiplied by a preset adjustment coefficient to obtain a distribution feature amplitude.

[0074] The difference between the distribution feature median and the distribution feature amplitude is obtained to obtain a distribution feature threshold.

[0075] It should be explained that the preset adjustment coefficient is used to adjust the sensitivity of the distribution feature amplitude, and the value range is determined by professional personnel according to the scene complexity of the forest area and the recognition task accuracy requirement, for example, a plurality of forest image samples are selected, the ground under the forest area is labeled, the distribution feature median and the distribution feature amplitude are calculated, the least square method is used to optimize different preset adjustment coefficient values, the pixel point set recognized as the ground under the forest in the corresponding threshold interval is counted, and the coincidence degree of the pixel point set between the labeled results is compared, and the preset adjustment coefficient with the largest coincidence degree is selected as the determined value. The standardization processing method includes but is not limited to standard linear transformation based on interval scaling, Z-Score standardization method based on statistics, or normalization method based on nonlinear mapping function. Herein, the application method of standardization processing is not described.

[0076] If the vegetation distribution feature of the pixel point is less than the distribution feature threshold, the pixel point is marked as the ground under the forest pixel point; otherwise, the pixel point is not marked.

[0077] Four ground under the forest pixel points constituting a regular rectangular structure in the spatial position are selected as the corner points, the ground under the forest grid unit is formed based on the corner points, and the uniform extrapolated ground point is generated by fitting the pixel points of the ground under the forest grid unit by using the bilinear interpolation method.

[0078] The extrapolated ground point refers to a supplementary point generated in a region based on a pixel point on the understory ground through a bilinear interpolation method, used to fill the blank area between original measuring points, so as to form uniformly distributed understory ground data.

[0079] It should be explained that the bilinear interpolation method is an interpolation algorithm for estimating the value of an unknown point in a two-dimensional grid.

[0080] In step S3, the forest land slope of the forest area is detected by arranging the ground clinometer, the ground clinometer is arranged in the forest area according to a predetermined grid spacing, that is, the center of the grid region formed by the predetermined grid spacing is arranged as a measuring point, the ground clinometer is stably connected to the ground through a fixed support, the elevation and horizontal coordinates of each measuring point are collected, and the local slope value is calculated according to the elevation and horizontal coordinates of the measuring point, and the specific calculation formula is as follows:

[0081] ;

[0082] Among them, is the local slope value, is the elevation of the measuring point, is the horizontal coordinate of the measuring point, is the horizontal spacing of the measuring point in the horizontal direction x axis, is the horizontal spacing of the measuring point in the horizontal direction y axis, is the arctangent function, used to convert the slope ratio to an angle value.

[0083] It should be noted that the ground clinometer is a measuring device that can collect the ground elevation and the corresponding horizontal coordinates at the set measuring point position, and obtain the spatial position parameters of the ground point through laser ranging, providing input data for the calculation of the local slope value; the predetermined grid spacing refers to the spatial arrangement interval of the measuring points according to the regular grid division in the forest area, which ensures the uniform distribution of the measuring points in space, and its value is set according to the area of the forest area and the target number of measuring points, for example, when the area of the forest area is , and the target number of measuring points is 100, the predetermined grid spacing is 100m.

[0084] The local slope value of all measuring points is compared with the preset slope threshold value:

[0085] When the local slope value is greater than or equal to the preset slope threshold value, the grid region where the measuring point is located is defined as a steep slope region;

[0086] Otherwise, it is not defined.

[0087] The preset slope threshold is a judgment criterion for determining whether the grid region where the measuring point is located belongs to a steep slope, and its value is set according to the classification standard of slope in topography research. For example, the standard of steep slope in the slope classification standard of geographic information is greater than 25°.

[0088] The steep slope proportion of the forest area is calculated based on the above definition result:

[0089] ;

[0090] Wherein, is the steep slope proportion, is the measuring point defined as a steep slope region in the grid region, is the total number of measuring points, is the local slope value, is the preset slope threshold.

[0091] At the same time, a three-dimensional displacement sensor is arranged at the crown position of the measuring point to monitor the crown fluctuation amplitude. The three-dimensional displacement sensor is fixed on the crown branch by a flexible support, and records the instantaneous displacement-time curve of the crown in the horizontal and vertical directions at a fixed sampling frequency within a preset observation period. The difference between the maximum and minimum values of the instantaneous displacement of the crown within the preset observation period is calculated as the crown fluctuation amplitude.

[0092] It should be noted that the three-dimensional displacement sensor refers to a measuring device that can measure the displacement of the target in three orthogonal directions. It is fixed on the crown branch by a flexible support to ensure that it can move synchronously with the branch under the action of wind or natural disturbance. The preset observation period refers to the continuous monitoring time length set to capture the dynamic fluctuation characteristics of the crown. It is determined according to the wind speed and crown vibration period of the forest area. For example, for a forest area with a wind speed of 3-5 m / s, the crown vibration period is 5-10 seconds, and the preset observation period can be set to 5 minutes. The fixed sampling frequency refers to the time interval at which the three-dimensional displacement sensor records displacement data within the preset observation period. Its setting method satisfies the Nyquist sampling theorem, i.e. the sampling frequency is at least twice the crown vibration frequency, to avoid signal aliasing. For example, if the crown vibration period is 5 seconds, i.e. the crown vibration frequency is 0.2 Hz, the sampling frequency can be set to 1 Hz or 2 Hz to ensure that the crown displacement change curve can be accurately captured.

[0093] The steep slope proportion and the crown fluctuation amplitude are standardized to form a forest terrain feature vector, and a logic regression algorithm is used for regional evaluation to output the terrain survey features of each measuring point:

[0094] ;

[0095] Wherein, is the terrain survey feature, is a natural constant, is a steep slope ratio, is a tree crown fluctuation amplitude, is an intercept term, and is a logistic regression coefficient.

[0096] It should be noted that the logistic regression algorithm is a statistical modeling method for binary classification or multi-classification problems, which predicts the probability of a test point belonging to a specific class by modeling the relationship between the input feature vector and the class label; the intercept term and the logistic regression coefficient are obtained by training, that is, the gradient descent method is used to optimize the parameters based on historical survey data, so that the error between the predicted probability and the true class is minimized, and the algorithm training process is not described here.

[0097] Comparing the terrain survey feature with the preset survey feature threshold, if the terrain survey feature is greater than or equal to the preset survey feature threshold, the test point is defined as a target area test point; otherwise, it is not defined;

[0098] Based on the connectivity analysis of the neighborhood relationship, the target area test points that are adjacent to each other are connected to form a continuous region as the division region of the forest area;

[0099] In the connectivity analysis, a four-neighborhood determination method is used, and test points with a distance less than a preset distance are considered as part of the division region to ensure the integrity and spatial continuity of the division region;

[0100] It should be noted that the preset survey feature threshold is determined according to historical survey data, forest terrain features and task requirements, by analyzing the distribution of steep slope ratio, tree crown fluctuation amplitude and other related terrain indicators, the critical value range that can distinguish target areas and non-target areas is determined; the preset distance is a determination parameter used to determine whether the test points belong to the neighborhood relationship in spatial analysis, which is set according to the grid spacing and test point density of the forest area, for example, taking 1 to 2 times the grid spacing as the preset distance to ensure that the continuous region can be correctly identified in the connectivity analysis; the connectivity analysis refers to a method of spatial neighborhood analysis of a set of test points that meet the determination conditions in the division region or anomaly region identification, which is used to combine test points that are adjacent to each other into a continuous region; the four-neighborhood determination method is a commonly used neighborhood determination method in connectivity analysis, taking the test point to be analyzed as the center, selecting the test points in the upper, lower, left and right four adjacent grids as neighborhood candidates along the two-dimensional grid direction, if the neighborhood candidate test point meets the preset condition, it is considered as belonging to the same continuous region with the center test point.

[0101] The terrain survey feature comprehensively reflects the complexity of the local terrain in the forest area, quantitatively divides the survey complexity of the region, that is, reflects the difficulty level faced by the divided region in implementing the terrain survey operation; the higher the proportion of steep slope, the greater the proportion of steep slope terrain in the divided region, and the greater the ground survey feature; the greater the tree canopy fluctuation amplitude, the more complex the tree canopy structure, the more significant the tree canopy moving track, the more difficult the undergrowth ground observation and moving measurement, and the greater the ground survey feature.

[0102] In step S4, the number of pixel points of the undergrowth ground and the divided region is obtained by the unmanned aerial vehicle oblique photography, the number of pixel points is converted into a spatial area, and the undergrowth ground area of each divided region and the total area of each divided region are obtained;

[0103] The ratio of the undergrowth ground area of each divided region to the area of each divided region is taken as the undergrowth ground proportion of each divided region.

[0104] Based on the undergrowth ground proportion, the vegetation coverage state of each divided region is evaluated, and the vegetation coverage state reflects the vegetation density in the divided region. According to a preset proportion threshold, the vegetation coverage state is divided into normal coverage or abnormal coverage:

[0105] When the undergrowth ground proportion is lower than the preset proportion threshold, it is determined that the vegetation coverage state of the divided region is abnormal coverage;

[0106] Conversely, the vegetation coverage state of the divided region is normal coverage.

[0107] It should be noted that the preset proportion threshold is set according to the forest type, historical vegetation coverage data and survey task target, the distribution characteristics of the undergrowth ground proportion in the forest region are counted, the vegetation density corresponding to different undergrowth ground proportions is analyzed, and the critical value is adjusted in combination with the survey target, for example, in the evergreen coniferous forest region, the preset proportion threshold is set to 0.3, that is, when the undergrowth ground proportion of the divided region is lower than 30%, it is determined that the vegetation coverage state of the divided region is abnormal coverage.

[0108] The divided region with abnormal coverage state is screened and marked as an abnormal region, and the extrapolated ground point of the abnormal region is output for subsequent survey and correction, including high-resolution image review, field survey or other auxiliary data verification, to ensure that the spatial position and terrain features accurately reflect the undergrowth ground conditions.

[0109] Embodiment 2

[0110] As shown in Figure 2 , the image processing measurement system based on unmanned aerial vehicle radar and oblique photography is used to realize the image processing measurement method based on unmanned aerial vehicle radar and oblique photography, which includes a data acquisition module, a vegetation identification module, a terrain analysis module and a region evaluation module, and the functions of each module are as follows:

[0111] The data acquisition module is used for acquiring image gray values of each pixel point of the forest area by oblique photography when surveying the forest area, acquiring radar echo amplitudes of the pixel points by radar detection, and calculating vegetation change gradients according to the image gray values;

[0112] The vegetation identification module is used for generating vegetation distribution characteristics of each pixel point by comprehensively considering the vegetation change gradients and the radar echo amplitudes, identifying and marking each pixel point according to the vegetation distribution characteristics to obtain the understory ground of the forest area, and generating the uniformly extrapolated ground points based on the understory ground.

[0113] The terrain analysis module is used for detecting a forest land slope and a tree crown fluctuation amplitude of the forest area, calculating a steep slope proportion of the forest area according to the forest land slope, evaluating terrain survey characteristics of the forest area in combination with the tree crown fluctuation amplitude, and dividing the forest area according to the terrain survey characteristics.

[0114] The region evaluation module is used for calculating an understory ground proportion in each divided region, evaluating a vegetation coverage state of each divided region based on the understory ground proportion, screening each divided region according to the vegetation coverage state, marking an abnormal region, and outputting extrapolated ground points of the abnormal region.

[0115] Finally, it should be noted that the relational terms herein such as first and second and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any such actual relationship or order between such entities or actions.

[0116] Moreover, the terms "include", "have", or any other variant thereof are intended to cover non-exclusive inclusions, such that processes, methods, articles, or apparatuses that include a series of elements are not limited to those elements, but can include other elements not expressly listed, or other elements inherent to such processes, methods, articles, or apparatuses. Without more limitations, an element defined by the phrase "including a" does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes the element.

[0117] In this document, the singular forms "a", "an", and "the" can also include plural forms unless the context clearly indicates otherwise. It should also be understood that the term "comprise / contain" or "have" and the like specifies the presence of stated features, integers, steps, operations, components, parts, or combinations thereof, but does not exclude the presence or addition of one or more other features, integers, steps, operations, components, parts, or combinations thereof, the possibility of which is also present in this specification. The term "and / or" used in this specification includes any and all combinations of the related listed items.

[0118] The various embodiments described in this specification are intended to be combinable unless otherwise indicated herein. Each embodiment described in this specification is intended to be focused on the differences from other embodiments, and each embodiment can be combined with any of the other embodiments as appropriate. Where the same or similar reference numerals are used in different figures, those reference numerals are intended to refer to the same or similar parts throughout this specification.

[0119] Numerous modifications to the embodiments described herein will be apparent to those skilled in the art, and the principles defined herein can be applied to other embodiments without departing from the spirit or scope of the application. Therefore, the application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for measuring based on image processing of an unmanned aerial vehicle radar and oblique photography, characterized in that: The method comprises the following steps: Step S1: When the forest area is surveyed, the image gray value of each pixel point of the forest area is collected by oblique photography, the radar echo amplitude value of the pixel point is obtained by radar detection, and the vegetation change gradient is calculated according to the image gray value; Step S2: The vegetation distribution characteristics of each pixel point are generated by synthesizing the vegetation change gradient and the radar echo amplitude value, each pixel point is identified and marked according to the vegetation distribution characteristics to obtain the understory ground of the forest area, and the uniform extrapolation ground point is generated based on the understory ground; In step S2, the product of the standardized processing of the vegetation change gradient and the radar echo amplitude value of the pixel point is taken as the vegetation distribution characteristics of the pixel point; The vegetation distribution characteristics of all pixel points in the forest image are combined into a vegetation feature set; The median of the vegetation feature set is taken as the distribution feature median, the absolute value of the difference between the distribution feature median and each vegetation distribution feature is obtained, and the distribution feature amplitude is obtained by multiplying the median of each absolute deviation by a preset adjustment coefficient; Step S3: The forest land slope and the tree crown fluctuation amplitude of the forest area are detected, the steep slope proportion of the forest area is calculated according to the forest land slope, the topographic survey characteristics of the forest area are evaluated in combination with the tree crown fluctuation amplitude, and the forest area is regionally divided according to the topographic survey characteristics; In step S3, the center of the grid region formed by the predetermined grid spacing is taken as the measuring point to arrange the ground inclinometer, the elevation and horizontal coordinates of each measuring point are collected, and the local slope value is calculated; When the local slope value is greater than or equal to the preset slope threshold, the grid region where the measuring point is located is defined as a steep slope region; Otherwise, it is not defined; The ratio of the number of measuring points in the grid region defined as the steep slope region to the total number of measuring points is taken as the steep slope proportion; In step S3, a three-dimensional displacement sensor is arranged at the crown position of the measuring point to monitor the instantaneous displacement of the crown; The difference between the maximum value and the minimum value of the instantaneous displacement of the crown in the preset observation period is taken as the tree crown fluctuation amplitude; The steep slope proportion and the tree crown fluctuation amplitude are standardized to construct a forest topographic feature vector, and the topographic survey characteristics of each measuring point are output by a logistic regression algorithm; Step S4: The understory ground proportion in each region is determined, the vegetation coverage state of the divided region is evaluated based on the understory ground proportion, the divided region is screened according to the vegetation coverage state, the abnormal region is marked, and the extrapolation ground point of the abnormal region is output.

2. The image processing measurement method based on unmanned aerial vehicle radar and oblique photography according to claim 1, wherein: In step S1, when the forest area is surveyed, the unmanned aerial vehicle carries an oblique photography camera to image different angles of the forest area to obtain a forest image; The forest image is processed by gray scale to obtain a gray scale image, and each pixel point in the gray scale image corresponds to a gray scale value; The pixel points of the forest images at different angles are one-to-one corresponding, and the image gray values of the pixel points at the same position are combined into a pixel gray set; The median of the pixel gray set is taken as the gray median value, and the mean value of the pixel gray value set is taken as the gray mean value; The average value of the gray median value and the gray mean value is taken as the image gray value of the pixel point.

3. The image processing measurement method based on unmanned aerial vehicle radar and oblique photography according to claim 2, characterized in that: In step S1, the forest area is scanned by a radar detection device carried by the unmanned aerial vehicle, electromagnetic waves are emitted to the forest area, and echo signals are received; According to the echo signals, the signal strength value of each pixel point is extracted as the radar echo amplitude value; Each pixel point is taken as a center pixel point, and the adjacent pixel points of the center pixel point are combined into a neighborhood set, and the image gray value of the center pixel point is subtracted from the image gray values of the pixel points in the neighborhood set to obtain the neighborhood gray difference value; The standard deviation of each neighborhood gray difference value is calculated as the vegetation change gradient.

4. The image processing measurement method based on unmanned aerial vehicle radar and oblique photography according to claim 1, characterized in that: In step S2, the distribution feature threshold value is obtained by subtracting the distribution feature amplitude from the distribution feature median value; If the vegetation distribution feature of a pixel point is less than the distribution feature threshold value, the pixel point is marked as an under-forest ground pixel point; otherwise, the pixel point is not marked; Four under-forest ground pixel points constituting a regular rectangular structure in spatial position are selected as corner points, and an under-forest ground grid cell is formed based on the corner points; The bilinear interpolation method is used to fit the pixel points of the under-forest ground grid cell to generate uniform extrapolated ground points.

5. The image processing measurement method based on unmanned aerial vehicle radar and oblique photography according to claim 1, characterized in that: In step S3, if the topographic survey feature is greater than or equal to the preset survey feature threshold value, the survey point is defined as a target area survey point; Otherwise, the survey point is not defined; Based on connectivity analysis, the target area survey points with a distance less than the preset distance are connected to form a continuous region as a division region of the forest area.

6. The image processing measurement method based on unmanned aerial vehicle radar and oblique photography according to claim 1, characterized in that: In step S4, the pixel point numbers of the under-forest ground and the division region are obtained by oblique photography of the unmanned aerial vehicle, the pixel point numbers are converted into spatial areas, and the under-forest ground area of each division region and the total area of each division region are obtained; The ratio of the under-forest ground area of each division region to the area of each division region is taken as the under-forest ground proportion of each division region; When the under-forest ground proportion is lower than the preset proportion threshold value, it is determined that the vegetation coverage state of the division region is abnormal coverage; Otherwise, the vegetation coverage state of the division region is normal coverage; The division region with abnormal coverage state is screened and marked as an abnormal region, and the extrapolated ground points of the abnormal region are output for subsequent survey and correction.

7. A UAV radar and oblique photogrammetry image processing measurement system for implementing the UAV radar and oblique photogrammetry image processing measurement method according to any one of claims 1-6, characterized in that: The system includes a data acquisition module, a vegetation identification module, a topographic analysis module, and a region evaluation module, and the functions of each module are as follows: The data acquisition module is used to acquire the image gray values of each pixel point of the forest area by oblique photography during the survey of the forest area, to acquire the radar echo amplitude value of each pixel point by radar detection, and to calculate the vegetation change gradient according to the image gray value. The vegetation identification module is configured to generate a vegetation distribution feature of each pixel point by comprehensively considering the vegetation change gradient and the radar echo amplitude, to identify and mark each pixel point according to the vegetation distribution feature, and to obtain the under-forest ground in the forest area, and to generate the uniform extrapolated ground point based on the under-forest ground; The terrain analysis module is configured to detect the forest land slope and the tree crown fluctuation amplitude in the forest area, to calculate the steep slope proportion of the forest area according to the forest land slope, to evaluate the terrain survey feature of the forest area in combination with the tree crown fluctuation amplitude, and to divide the forest area into regions according to the terrain survey feature; The region evaluation module is configured to calculate the under-forest ground proportion in each divided region, to evaluate the vegetation coverage state of each divided region based on the under-forest ground proportion, to screen the divided regions according to the vegetation coverage state, and to mark and output the extrapolated ground point of the abnormal region.

Images