CN110672020A - A method for measuring the height of standing trees based on monocular vision - Google Patents

Abstract

The invention discloses a method for measuring the height of a standing tree based on monocular vision. The problem that the existing tree height measurement is complicated and time-consuming, and the standing tree height cannot be detected quickly and conveniently is solved. The method includes calibrating the camera, acquiring the internal parameters of the camera, extracting the outline of the standing tree on the input image, establishing a geometric model of the distance between the image coordinate system and the world coordinate system, acquiring the depth relational expression of the standing tree image, establishing the height pixel coordinate mapping model, and obtaining the object to be measured. The relationship between the target height and the height of the standing tree is calculated according to the shooting height and angle of the camera. The invention has low measurement cost and high work efficiency, only needs to use the existing smart phone, only needs to obtain the shooting angle and shooting height, and the operation is more convenient and fast. In addition, the measurement method of the present invention does not require complicated operations, consumes less time, and has high algorithm precision, high degree of automation and simple operation.

Description

A method for measuring the height of standing trees based on monocular vision

technical field

The invention relates to the technical scope of tree detection, in particular to a method for measuring the height of a standing tree based on monocular vision.

Background technique

The height of trees is one of the important indicators to evaluate the quality of standing trees and the growth status of trees, and it has become a necessary work in the investigation of forest resources. As a basic way to obtain basic tree measurement factors, tree measurement tools and their algorithms are extremely important in accuracy and practicability, and their modernization level also reflects the level of forestry modernization ^[5] . At present, the mainstream altimeter instruments, such as the Bruce-type altimeter ^[6] , are widely used because of their low price, simple operation, and can meet the requirements of general accuracy, but they are limited by the need to measure the horizontal distance. Forest measurements are susceptible to occlusion. 3D laser scanning technology achieves the purpose of fast and accurate automatic measurement of tree height by quickly and accurately obtaining the 3D coordinate point cloud of objects. However, lidar measurement has the disadvantages of high equipment cost, not easy to carry, and long processing time.

In recent years, machine vision measurement technology has developed rapidly, and its depth and breadth in forestry applications are also increasing. The current mainstream vision measurement systems can be divided into two categories, mainly including monocular vision measurement systems and binocular vision measurement systems. Among them, the main principle of the binocular vision measurement system is to use two cameras to capture images from different visions of the target to be measured, and to reconstruct three-dimensionally through the corresponding matching points in the two images, so as to obtain the information of the target to be measured in the three-dimensional space. Compared with binocular vision, monocular vision system has the advantages of simple camera calibration process and system structure, and avoids the disadvantage of difficult stereo matching in binocular vision measurement system, so it has become an important research trend in the fields of photogrammetry and computer vision. one. In practice, the measurement flexibility of monocular vision measurement, the quickness of image acquisition, and the universality of the solution method make it widely used in non-contact measurement in natural environment. Zhu Binglin et al. used machine vision to obtain the three-dimensional phenotype of plants in the greenhouse, reconstructed the three-dimensional corn and soybean plants grown in different periods, and evaluated the accuracy of leaf length and leaf maximum width based on manual measurements. Yang Kun et al. used high-resolution images of unmanned aerial vehicles and used _PIX 4D software to produce 3D point clouds. Based on the maximum inter-class variance method, the trees were divided into tree point clouds and ground point clouds, and the tree top height and the average ground height were extracted. . Zhou Keyu et al. used the similar triangle principle when measuring the tree height, based on the trigonometric function and the known calibration object size to compare and calculate, to a certain extent, completed the rapid measurement of the standing tree height. Guan Fangli et al. extracted the minimum external rectangle of trees through digital image processing calculation, and calibrated and obtained the internal and external parameters of the camera in the smartphone. According to the two-dimensional image information of the camera and the known parameters, the coordinate reconstruction of the three-dimensional world is carried out to measure the height of the standing tree. However, most of the domestic tree measurement tools and their algorithms have problems such as poor accuracy, low degree of automation, and complicated operations. Traditional measurement principles and methods are still used in tree height measurement, and the distance information of standing trees must be measured in advance. In addition, most of them require complex calculations, which are time-consuming and cannot measure the height of standing trees quickly and conveniently.

SUMMARY OF THE INVENTION

The invention mainly solves the problems in the prior art that the tree height measurement is complicated in operation, takes a long time, and cannot quickly and conveniently detect the height of a standing tree, and provides a method for measuring the height of a standing tree based on monocular vision.

The above-mentioned technical problems of the present invention are mainly solved by the following technical solutions: a method for measuring the height of a standing tree based on monocular vision, comprising the following steps,

S1. Calibrate the camera and obtain the internal parameters of the camera;

The calibration adopts the improved Zhang Zhengyou calibration method, introduces a camera calibration model with nonlinear distortion term for calibration, realizes the distortion correction and obtains the internal parameters of the camera. Reference for the improved Zhang Zhengyou calibration method: Liu Yan, Li Tengfei. Research on the improvement of Zhang Zhengyou's camera calibration method. Optical Technology, 2014, 40(06): 565-570. If the coordinates of the origin of the image coordinate system (x, y) in the pixel coordinate system (u, v) are (u ₀ , v ₀ ). Any point (X _c , Y _c , Z _c ) in the camera coordinate system is projected to (x, y, f) on the image coordinate system, and the distance between the image coordinate system plane and the dot is f. And the process of transforming from the object's world coordinates P _w (X _W , Y _W , Z _W ) to the camera coordinates P _c is a rigid body motion, which can be described by the translation and rotation of the object. So the conversion relationship from the world coordinate system to the camera coordinate system is:

Among them, M _int and M _ext are internal and external parameters of the camera, respectively, and the external parameters of the camera include a rotation matrix R and a translation matrix T. d _x and _dy are the physical dimensions of each on the plane (unit: mm).

S2. Extract the outline of the standing tree from the input image, and obtain the pixel difference of the height of the standing tree;

S3. Establish a geometric model of the distance between the image coordinate system and the world coordinate system, and obtain the relationship between the depth of the standing tree image

where d is the depth value from the camera to the target to be measured, h is the vertical horizontal height from the center of the lens to the ground, γ is the vertical top-down angle of the camera, f _y is the focal length in the y-axis direction of the image plane, and v is the y-axis coordinate in the image coordinate system , v ₀ is the principal point of the y-axis of the image plane;

S4. Establish a height pixel coordinate mapping model to obtain the height relationship of the target to be measured

Where H is the height of the standing tree, and yy′ is the pixel difference of the height of the standing tree;

S5. Calculate the height of the standing tree according to the shooting height and angle of the camera.

In the present invention, the height of the standing tree is measured by means of a smart phone and a machine vision technology. The GraphCut algorithm is used to segment the standing tree image to automatically obtain the height pixels of the standing tree in the standing tree image. Then, the depth information of the standing tree image is automatically obtained through the geometric similarity model, and then the angle information of the built-in gyroscope of the mobile phone and the height information of the mobile phone are extracted to solve the calculation. Measure tree height. A new set of measurement methods is formed. It only needs to use the camera of the existing smartphone to obtain the shooting angle and shooting height, without measuring the distance from the standing tree in advance, and the operation is more convenient and fast. In addition, the measurement method of the present invention does not require complicated operations, consumes less time, and has high algorithm precision, high degree of automation and simple operation. The method has low measurement cost and high work efficiency, does not require strict hardware conditions when using a smartphone device for tree height measurement, and is convenient for device integration, and only one person can complete the entire tree height measurement process.

As a preferred solution, the camera internal parameters obtained after calibration in step S1 include f _x , f _y , u ₀ , and v ₀ , where f _x is the focal length in the x-axis direction of the image plane, and u ₀ is the x-axis of the image plane. main point.

As a preferred solution, the specific process of step S2 includes:

S21. Preprocess the input image to obtain an image that highlights the characteristics of the standing tree image; the preprocessing includes transforming the object pixel set using point set properties, integral geometric sets and topology theory, and dilating and eroding the image.

S22. Perform front and background segmentation on the preprocessed image to obtain a standing tree outline image;

In this step, the GraphCut algorithm is used to segment the foreground of the image to obtain the bottom center point of the standing tree to be measured and the highest point on the standing tree image. The algorithm can effectively overcome the complex background interference in the natural environment under the influence of different light intensities and other natural environments, and accurately segment specific standing trees.

S23. Scan the standing tree contour image to obtain the coordinate distribution of the standing tree contour image;

S24. Output two maximum points in the standing tree outline image, which are the maximum and minimum values of the y-axis of the standing tree outline in the image plane coordinate system, and calculate the height pixel difference of the standing tree according to the maximum points. The two maximum points in the standing tree contour image are S _max (x, y) and S _min (x, y), and yy′ is the pixel difference in the vertical direction of the standing tree image on the image plane. This scheme scans the standing tree contour image and outputs the two maximum points in the image as the input in the subsequent model for measuring the standing tree height.

As a preferred solution, the process of obtaining the depth relational expression of the standing tree image in step S3 includes:

S31. Establish a geometric model of the distance between the image coordinate system and the world coordinate system according to the camera imaging principle and the relationship between the camera and the ground plane. The geometric model is to set the point P and the optical axis to form an intersection on the image image plane ABCD through the center of the lens (x , y), (x ₀ , y ₀ ), (x, y) is the projection point of point P on the two-dimensional plane graphics coordinate system, that is, the image plane, (x ₀ , y ₀ ) is the optical axis and the image plane , the depth value from the camera to the target to be measured, that is, the distance from point P to the center of the lens is d, the distance from the center of the lens to the image plane is f, the vertical height from the center of the lens to the ground is h, and the vertical top-down angle of the camera is γ, The angle between the camera projection and the optical axis is β.

S32. Derive the relationship between the distance between pixel coordinates and world coordinates according to the geometric model

The relationship between the pixel and the focal length of β in the three-dimensional image is expressed as

According to formulas (1) and (2), we get:

S33. According to the geometric relationship between image coordinates and pixel coordinates

Obtain y=(vv ₀ )dy, given that y ₀ =0, get

得到立木图像深度关系式S34. According to

Get the relationship between the depth of the standing tree image

As a preferred solution, the process of obtaining the height relational expression of the target to be measured in step S4 includes:

S41. According to the principle of camera pinhole imaging, establish a height pixel coordinate mapping model. The model is that A ₃ A ₄ in the actual imaging plane is imaged in the image plane through the center of the lens, A ₃ A ₄ is the actual standing height H, and A ₃ is in the image plane. _The inner imaging point is xy, the imaging point of A4 in the image plane is xy', yy' is the pixel difference of the standing tree image in the vertical direction on the image plane, Q is the center of the lens, M is the intersection of the optical axis and the actual imaging plane, QM is the position of the optical axis, the distance from the lens center P to the ideal imaging surface is d, and the vertical top-down angle of the camera is γ;

xy, xy' are the imaging points of the standing tree on the image plane, wherein the maximum value S _max (x, y) and the minimum value are S _min (x, y). The ideal imaging plane is the vertical water surface, and the top-view angle of the camera is transformed according to the measurement requirements. The actual imaging plane formed is inclined, and the M point is also the intersection of the actual imaging plane and the ideal imaging plane. QO is the vertical height of the camera from the horizontal plane. Ideally, the camera top-down angle γ is 0 degrees, the optical axis is a horizontal straight line, N is the intersection of the optical axis and the centrifugal imaging plane in the ideal case, and QN is the ideal line where the optical axis is located. When the top-down angle of the camera changes, the relative angle between the two imaging planes does not change.

S42. Obtained according to the height pixel coordinate mapping model

A₃A₄＝H，得到in

A ₃ A ₄ =H, we get

S43. Combining the relationship of the depth of the standing tree image to obtain the relationship of the height of the standing tree

This scheme obtains the relationship formula of the height of the standing tree, and only needs to obtain the height H of the standing tree by obtaining the pixel difference yy′ of the standing tree image on the imaging plane and the top-down angle γ obtained by the high-precision gyroscope inside the camera, where the physical unit of h is m/ The physical force units of m, yy' and f _y are all pixels/pixel, and the distance unit of H is meters/m. The method does not need to know the motion information of the camera, does not need to perform auxiliary operations, takes less time, is fast and convenient to operate, and has high accuracy and effectiveness.

Therefore, the advantages of the present invention are:

Just use the existing smartphone to obtain the shooting angle and shooting height, without measuring the distance to the standing tree in advance, the operation is more convenient and fast.

The measurement method does not require complicated operations, takes less time, and has high algorithm accuracy, high automation and simple operation.

The measurement cost is low and the work efficiency is high. The use of smartphone devices for tree height measurement does not require strict hardware conditions and the device is easily integrated. The entire tree height measurement process can be completed by only one person.

Description of drawings

Fig. 1 is a kind of schematic diagram of the geometric model of the distance between the image coordinate system and the world coordinate system in the present invention;

Fig. 2 is a kind of schematic diagram of image coordinate system and pixel coordinate system conversion model in the present invention;

Fig. 3 is a kind of structure schematic diagram of height pixel coordinate mapping model in the present invention;

Fig. 4a is the standing tree image to be measured input in the present invention;

Fig. 4b is the standing tree outline image after segmentation in the present invention;

FIG. 5 is a comparison diagram of relative errors in an embodiment of the present invention.

Detailed ways

The technical solutions of the present invention will be further described in detail below through embodiments and in conjunction with the accompanying drawings.

Example:

This embodiment is a method for measuring the height of a standing tree based on monocular vision, specifically a method for measuring the height of a standing tree based on a smartphone combined with machine vision technology, and the smartphone has a camera and a high-precision gyroscope. Methods The GraphCut algorithm was used to segment the standing tree image to realize the automatic acquisition of the height pixels of the standing tree in the standing tree image. Then, the depth information of the standing tree image was automatically obtained through the geometric similarity model, and then the angle information of the built-in gyroscope of the mobile phone and the height information of the mobile phone were extracted to solve the problem. The calculation and measurement of tree height provides a new set of measurement methods and technical means for forest resources survey.

The specific steps of the method of this embodiment include:

S1. Calibrate the camera and obtain the internal parameters of the camera;

S2. Extract the outline of the standing tree from the input image, and obtain the pixel difference of the height of the standing tree;

S3. Establish a geometric model of the distance between the image coordinate system and the world coordinate system, and obtain the depth of the standing tree image;

S4. Establish a height pixel coordinate mapping model, and obtain the height relationship of the target to be measured;

S5. Calculate the height of the standing tree according to the camera shooting height and the angle combined with the height relationship.

In step S1, the calibration adopts an improved Zhang Zhengyou calibration method, and a camera calibration model with nonlinear distortion terms is introduced for calibration, so as to achieve distortion correction and obtain camera internal parameters. Reference for the improved Zhang Zhengyou calibration method: Liu Yan, Li Tengfei. Research on the improvement of Zhang Zhengyou's camera calibration method. Optical Technology, 2014, 40(06): 565-570. If the coordinates of the origin of the image coordinate system (x, y) in the pixel coordinate system (u, v) are (u ₀ , v ₀ ). Any point (X _c , Y _c , Z _c ) in the camera coordinate system is projected to (x, y, f) on the image coordinate system, and the distance between the image coordinate system plane and the dot is f. And the process of transforming from the object's world coordinates P _w (X _W , Y _W , Z _W ) to the camera coordinates P _c is a rigid body motion, which can be described by the translation and rotation of the object. So the conversion relationship from the world coordinate system to the camera coordinate system is:

Among them, M _int and M _ext are internal and external parameters of the camera, respectively, and the external parameters of the camera include a rotation matrix R and a translation matrix T. d _x and _dy are the physical dimensions of each on the plane (unit: mm).

The camera internal parameters obtained through calibration in this method include f _x , f _y , u ₀ , and v ₀ . f _x is the focal length in the x-axis direction of the image plane, f _y is the focal length in the y-axis direction of the image plane, v is the y-axis coordinate in the image coordinate system, u ₀ is the principal point of the x-axis of the image image plane, v ₀ is the image image plane The principal point of the y-axis.

Step S2 is to perform preprocessing and feature extraction on the input image, and the specific process includes:

S21. Preprocess the input image to obtain an image that highlights the characteristics of the standing tree image;

Due to the complex background interference in natural scenes, when collecting standing tree images, it is necessary to preprocess the original images to highlight useful information and suppress useless information, so as to improve image quality and facilitate the extraction of standing tree image feature information.

S22. Perform front and background segmentation on the preprocessed image to obtain a standing tree outline image;

Use the Graph Cut algorithm to perform front and background segmentation on the collected standing tree images to help obtain the bottom center point of the standing tree to be tested and the highest point of the standing tree on the image. The algorithm can effectively overcome the complex background interference in the natural environment under the influence of different light intensities and other natural environments, and accurately segment specific standing trees.

S23. Scan the standing tree outline image to obtain the coordinate distribution of the standing tree outline image; since the standing tree to be measured has irregular shapes, free positions and directions in the image, the coordinate distribution of the two-dimensional image is scanned.

S24. Output two maximum points in the standing tree outline image, which are the maximum and minimum values of the Y-axis of the standing tree outline in the image plane coordinate system, and calculate the height pixel difference of the standing tree according to the maximum points. The two maximum points in the standing tree contour image are S _max (x, y) and S _min (x, y), and yy′ is the pixel difference in the vertical direction of the standing tree image on the image plane. By scanning the standing tree contour image, these two maximum points in the image are output as the input in the subsequent model for measuring the standing tree height.

In the tree height measurement model, the geometric similarity method is used to measure the shape, two-dimensional position and other information of the target to be measured. , the actual parameters of the object to be measured are obtained through the spatial geometric similarity relationship between the object to be measured and the image. Step S3 establishes a geometric model of the distance between the image coordinate system and the world coordinate system according to the imaging principle of the camera and the relationship between the camera and the ground plane, and uses the geometric similarity method to obtain image depth information.

The process of obtaining the depth relational expression of the standing tree image in step S3 includes:

S31. Establish a geometric model of the distance between the image coordinate system and the world coordinate system according to the camera imaging principle and the relationship between the camera and the ground plane.

As shown in Figure 1, the geometric model is to set point P and the optical axis to form intersection points (x, y), (x ₀ , y ₀ ) on the image plane ABCD through the center of the lens respectively, and (x, y) is the point P at The two-dimensional plane graphic coordinate system is the projection point on the image plane, (x ₀ , y ₀ ) is the intersection of the optical axis and the image plane, and the depth value from the camera to the target to be measured, that is, the distance from point P to the center of the lens is d, The distance from the center of the lens to the image plane is f, the vertical height from the center of the lens to the ground is h, the vertical top-down angle of the camera is γ, the angle between the camera’s projection and the optical axis is β, and the line where the L point is located represents the ground plane.

S32. Derive the relationship between the distance between pixel coordinates and world coordinates according to the geometric model

The relationship between the pixel and the focal length of β in the three-dimensional image is expressed as

According to formulas (1) and (2), we get:

S33. As shown in Figure 2, according to the geometric relationship between the image coordinates and the pixel coordinates

Obtain y=(vv ₀ )dy, given that y ₀ =0, get

S34. According to Get the relationship between the depth of the standing tree image

where d is the depth value from the camera to the target to be measured, h is the vertical horizontal height from the center of the lens to the ground, γ is the vertical top-down angle of the camera, f _y is the focal length in the y-axis direction of the image plane, and v is the y-axis coordinate in the image coordinate system , v ₀ is the principal point of the y-axis of the image plane. Both v ₀ and f _y are internal parameters of the camera, obtained through the camera calibration in step S1.

The process of obtaining the height relational expression of the target to be measured in step S4 includes:

S41. According to the camera pinhole imaging principle, establish a height pixel coordinate mapping model.

As shown in Figure 3, the model is that A _{3 A 4 is imaged in the image plane through the center of the lens in the actual imaging plane, A 3 A 4 is the} actual height H of the standing tree, the imaging point of A ₃ in the image plane is xy, and A ₄ is in the image plane. The imaging point in the plane is xy′, yy′ is the pixel difference of the standing tree image in the vertical direction on the image plane, Q is the center of the lens, M is the intersection of the optical axis and the actual imaging plane, QM is the position of the optical axis, and the center of the lens P to The distance of the ideal imaging plane is d, and the vertical top-down angle of the camera is γ;

xy, xy' are the imaging points of the standing tree on the image plane, wherein the maximum value S _max (x, y) and the minimum value are S _min (x, y). A ₁ A ₂ is the ideal imaging plane, and the ideal imaging plane is the vertical water surface. The top-down angle of the camera is transformed according to the measurement requirements. The actual imaging plane formed is inclined, and point M is also the actual imaging plane and the ideal imaging plane. intersection. QO is the vertical height of the camera from the horizontal plane. Ideally, the camera top-down angle γ is 0 degrees, ∠OQN=90 degrees, the optical axis is a horizontal straight line, N is the intersection of the optical axis and the centrifugal imaging plane in the ideal case, and QN is the ideal case. The straight line where the lower optical axis is located, QN⊥A ₁ A ₂ , the length represented by A ₁ A ₂ is the tree height H to be measured in the three-dimensional space. In actual measurement, the top-down angle of the camera will change with the measurement requirements. According to the height pixel coordinate mapping model, when the top-down angle of the camera changes, the relative angle between the imaging surfaces will not change. If the top view angle ∠OQG=γ, then ∠GQM=90 degrees, QN⊥A ₃ A ₄ , so ∠GQM=γ.

S42. Obtain the _A3A4 height _formula according to the height pixel coordinate mapping model

已知QN＝d，A₃A₄＝H，得到in

Given that QN=d, A ₃ A ₄ =H, we get

S43. Combining the relationship of the depth of the standing tree image to obtain the relationship of the height of the standing tree

The physical unit of h is m/m, the physical unit of yy′ and f _y is pixel/pixel, and the distance unit of H is m/m. It is only necessary to obtain the height H of the standing tree by obtaining the pixel difference yy′ of the standing tree image on the imaging plane and the top-down angle 2 obtained by the high-precision gyroscope inside the camera.

This implementation is described below using a Xiaomi mobile phone with a model of MI2S. The phone's Android version is 4.1.2. After passing through the camera Ding Bing, the internal parameters of the camera are obtained: f _x =3486.5637, u ₀ =1569.0383, f _y =3497.4652, v ₀ =2107.9898, set the vertical height of the mobile phone from the ground h=1.10cm, and the image resolution is 3120 ×4208. Taking the measured data as the true value of the standing tree height, the actual height of the standing tree was measured to be 6.2 meters. As shown in Figure 4a, input the standing tree image to be measured, and use Graph Cut to segment the standing tree image through image processing technology. As shown in Figure 4b, the standing tree outline image is output.

Scanning the coordinate distribution of the standing tree contour image, the maximum value S _max (977,365) and the minimum value S _min (1299,3463) in the Y-axis direction can be obtained, and yy′=3144.6897 can be obtained. At this time, the top view angle of the mobile phone is γ=10.0° , the vertical height of the mobile phone from the ground is h=1.10m. Through the relationship formula of the height of the standing tree, the height of the standing tree H=6.2815m can be obtained.

In order to verify the accuracy of the tree height measurement model in the sample wood interval with different heights, in this embodiment, the common sample wood height range is used as the sample wood interval, and the height of the standing wood to be measured is measured from different angles. Numbering the trees and extracting the height of standing trees based on the high-resolution influence of UAVs (Yang Kun, Zhao Yanling, Zhang Jianyong, et al. Extracting tree heights using UAV high-resolution images. Journal of Beijing Forestry University, 2017, 39( 8): 17-23. Hereinafter referred to as the comparative method) for comparative analysis. The results are shown in Table 1.

Table 1 Standing tree height measurement data table

As shown in Figure 5, after calculation, the relative error range of measurement is -0.029~0.338m, the highest relative error of standing tree height measurement is 4.22%, and the lowest relative error is 0.30%. The highest relative error of the contrasting method is 16.23%, and the lowest relative error is 0.33%. Among them, only 13.64% of the method in this embodiment has an error of more than 6%, and 72.74% of the method based on the comparison method has a relative error of more than 6%, which has a large fluctuation. Through comparative analysis, it can be concluded that the method of this embodiment has higher accuracy in measuring the height of standing trees and more stable measurement results. In addition, when using the mobile phone camera for image acquisition, the rotation angle of the mobile phone should be selected to ensure that the image of the standing tree to be measured can be completely imaged on the imaging surface of the mobile phone camera, and the pixels at the bottom of the standing tree to be measured in the image are imaged in the optical center. The main point of the plane, so that the height of the standing tree to be measured can be accurately measured. In FIG. 5 , the method in this paper is the method in this embodiment, and the method in document (15) is the comparison method.

The specific embodiments described herein are merely illustrative of the spirit of the invention. Those skilled in the art to which the present invention pertains can make various modifications or additions to the described specific embodiments or substitute in similar manners, but will not deviate from the spirit of the present invention or go beyond the definitions of the appended claims range.

Claims (5)

1. A standing tree height measuring method based on monocular vision is characterized in that: comprises the following steps of (a) carrying out,

s1, calibrating a camera to obtain internal parameters of the camera;

s2, performing standing tree contour extraction on the input image to obtain a standing tree height pixel difference;

s3, establishing a geometric model of the distance between the image coordinate system and the world coordinate system to obtain a stumpage image depth relational expression

Wherein d is the depth value from the camera to the target to be measured, h is the vertical horizontal height from the center of the lens to the ground, gamma is the vertical depression angle of the camera, and f_yIs the focal length in the y-axis direction of the image plane, v is the y-axis coordinate in the image coordinate system, v₀Is the principal point of the y axis of the image plane;

s4, establishing a height pixel coordinate mapping model to obtain a height relation formula of the target to be measured

Wherein H is the standing tree height, yy' is the standing tree height pixel difference;

and S5, calculating the height of the standing tree according to the shooting height and angle of the camera.

2. The method as claimed in claim 1, wherein the camera internal parameters obtained after calibration in step S1 include f_x、f_y、u₀、v₀Wherein f is_xIs the focal length, u, of the x-axis direction of the image plane₀Is the principal point of the x-axis of the image plane.

3. The standing tree height measuring method based on the monocular vision as claimed in claim 1, wherein the step S2 comprises the following steps:

s21, preprocessing an input image to obtain an image with outstanding standing tree image characteristics;

s22, performing foreground and background segmentation on the preprocessed image to obtain a standing tree contour image;

s23, scanning the standing tree outline image to obtain the coordinate distribution condition of the standing tree outline image;

and S24, outputting two minimum points in the standing tree contour image, wherein the two minimum points are respectively the maximum value and the minimum value of the standing tree contour on the y axis under the image plane coordinate system, and calculating the standing tree height pixel difference according to the minimum points.

4. The method for measuring the height of the standing tree based on the monocular vision as claimed in claim 1, wherein the step of obtaining the depth relation of the standing tree image in the step S3 comprises:

s31, establishing a geometric model of the distance between the image coordinate system and the world coordinate system according to the camera imaging principle and the relation between the camera and the ground plane, wherein the geometric model is that a P point and an optical axis are set to form intersection points (x, y) and (x) on the image plane ABCD through the lens center respectively₀,y₀) (x, y) is the projection point of the point P on the two-dimensional plane figure coordinate system, namely the image plane, (x)₀,y₀) The distance from the camera to the depth value of a target to be detected, namely the distance from a point P to the center of the lens is d, the distance from the center of the lens to the image plane is f, the vertical horizontal height from the center of the lens to the ground is h, the vertical depression angle of the camera is gamma, and the included angle between the projection of the camera and the optical axis is beta.

S32, deriving a relational expression of the distance between the pixel coordinate and the world coordinate according to the geometric model

Beta in the three-dimensional image, the relational expression of the pixel and the focal length is

Obtained according to equations (1) and (2):

s33, according to the geometric relation between the image coordinates and the pixel coordinates

Obtaining y ═ v-v₀) dy, known y₀Is equal to 0, to obtain

S34. according to

Obtaining the vertical tree image depth relational expression

5. The method for measuring the height of the standing tree based on the monocular vision as claimed in claim 4, wherein the step S4 of obtaining the relation of the height of the object to be measured comprises:

s41, establishing a height pixel coordinate mapping model according to the camera pinhole imaging principle, wherein the model is A in an actual imaging plane₃A₄Imaging in the image plane through the lens center, A₃A₄To the actual standing height H, A₃In the image planeThe imaging point is xy, A₄The imaging point in the image plane is xy ', yy' is the pixel difference of the standing tree image in the vertical direction on the image plane, Q is the lens center, M is the intersection point of the optical axis and the actual imaging plane, QM is the position of the optical axis, the distance from the lens center P to the ideal imaging plane is d, and the vertical depression angle of the camera is gamma;

s42, obtaining the height pixel coordinate mapping model

Wherein

A₃A₄H, to obtain

S43, combining the vertical wood image depth relational expression to obtain a vertical wood height relational expression

Images