Three-dimensional measuring and positioning method and system based on binocular stereoscopic vision and bird repelling linkage method

Abstract

The invention discloses a binocular stereo vision-based three-dimensional measurement and positioning method and system and a bird repelling linkage method, and relates to the technical field of bird repelling. Comprising the steps of collecting left-eye and right-eye images and obtaining original visual data; respectively carrying out bird target detection in the left and right eye images, and determining a bird target object to be processed; binocular correction parameters are calculated, and errors caused by unparallel optical axes of left and right eye images are corrected; on the basis of linear constraint, bird targets in the left-eye image and the right-eye image are matched, and the same bird targets are found; calculating a parallax value of the same bird target; calculating the distance data of the bird target according to the distance measurement parameter and the parallax value; and calculating the spatial three-dimensional information of the bird target according to the longitude and latitude of the system deployment place and the azimuth pitching distance of the optical system. According to the invention, the three-dimensional information of all bird targets in the field of view is determined through optical binocular calibration, stereo matching and depth calculation, and the problem that an optical bird detection system cannot acquire the spatial position information of the bird targets is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of bird repelling, and particularly to a binocular stereo vision three-dimensional measurement and positioning method, system and bird repelling linkage method. Background Art

[0002] The bird detection devices at airports include binocular observation systems and radar detection systems. Among them, the commonly used optical detection devices are visible light / infrared cameras and visible light / infrared search and tracking systems. Optical detection belongs to passive detection and does not radiate electromagnetic waves outward like radar systems, so it has a high acceptance rate at airports, but it is very difficult to measure distances. The visible light / search and tracking system can be equipped with a laser rangefinder to obtain the distance information of the target, but it needs to stably track the target and can only measure the distance of the target at the center of the field of view. Airport bird prevention work requires obtaining the accurate spatial position information of bird targets so as to carry out targeted prevention and control, so it is necessary to obtain the distance information of the target. The existing problems are as follows: The ordinary monocular optoelectronic camera cannot achieve precise ranging, and the search and tracking system with a laser rangefinder can only measure the distance of the target being tracked and cannot measure the distance of all bird targets in the field of view.

[0003] In terms of bird repelling, airport bird repelling devices are divided into auditory repelling, visual stimulation, olfactory repelling, etc. Among them, auditory repelling is the main repelling means, and the available devices include: LD bullet bird repelling guns, directional sound waves, omnidirectional sound waves, gas guns, fireworks guns, etc. Acoustic bird repelling and visual bird repelling are currently the commonly used bird repelling means at airports. The bird repelling strategy is generally to turn on regularly or play in a cycle, regardless of whether there are birds around. This strategy lacks pertinence. At the initial stage of the installation of these devices, they have a certain bird repelling effect on birds. However, after running for a period of time, the birds develop tolerance to the devices, making them gradually lose their bird repelling ability. How to combine the binocular observation system with bird repelling to solve the tolerance problem of bird repelling devices is another field that needs to be challenged. Summary of the Invention

[0004] In view of the deficiencies of the above-mentioned prior art, the present invention provides a binocular stereo vision three-dimensional measurement and positioning method, system and bird repelling linkage method. By means of optical binocular calibration, stereo matching, and depth calculation, the three-dimensional information of all bird targets in the field of view is determined, and the problem that the optical bird detection system cannot obtain the spatial position information of bird targets is solved. Further, in order to solve the tolerance problem of birds to bird repelling devices, a new bird repelling strategy is proposed: According to the provided spatial position of the bird target, the system-level spatial registration method and hierarchical clustering algorithm are used to find the most suitable bird repelling device. For acoustic bird repelling devices, the natural enemy sounds of the bird species can be selected according to the bird species recognition result to precisely disperse the bird target.

[0005] The specific technical solutions are as follows:

[0006] In the first aspect, the present invention discloses a binocular stereo vision three-dimensional measurement and positioning method, including:

[0007] Collect left and right eye images based on a binocular observation system to obtain original visual data; respectively perform bird target detection in the left and right eye images to determine the bird target objects to be processed;

[0008] Calculate binocular correction parameters to correct the error caused by the non-parallel optical axes of the left and right eye images; based on the straight-line constraint, match the bird targets in the left and right eye images to find the same bird target;

[0009] Calculate the disparity value of the same bird target; according to the ranging parameter and the disparity value, calculate the distance data of the bird target;

[0010] According to the longitude and latitude of the deployment location of the binocular observation system and the azimuth, pitch, and distance of the optical system, calculate the three-dimensional spatial information of the bird target to achieve comprehensive spatial positioning and description of the bird target.

[0011] Further, in the above solution, the steps of calculating the binocular correction parameters include:

[0012] Aim at bird targets outside the distance requirement, and make the bird targets appear at N positions in the upper left, upper right, lower left, lower right, and upper middle of the left and right eye images, where N≥5;

[0013] Suppose the coordinates of the bird target in the left and right eye images are respectively: {(x 1,k ,y 1,k )k = 1, 2,..., N}, {(x 2,k ,y 2,k )k = 1, 2,..., N}, and use affine transformation to calculate the affine transformation model parameters as: α = (α1, α2, α3, α4, α5, α6), and this set of parameters is the binocular correction parameters;

[0014] Taking the left eye image as the reference, transform the coordinates of the right eye image to the left eye to correct the problem of non-parallel optical axes of the left and right eye cameras.

[0015] Further, in the above solution, calculating the binocular correction parameters to correct the error caused by the non-parallel optical axes of the left and right eye images includes:

[0016] Calculate the ranging parameter based on the fixed-distance calibration points to obtain target images at M different distances, detect the target coordinates, and assume that the target coordinates in the left eye image are: {(x 3,k ,y 3,k )k = 1, 2,..., M}, and the target coordinates in the right eye image matched with it are: {(x 4,k ,y 4,k )k = 1, 2,..., M}, and use the binocular correction parameters to correct the target coordinates in the right eye image to: {(x5,k , y 5,k ) k = 1, 2, …, M},

[0017]

[0018] Then the calculation formula of the parallax △ is as follows:

[0019]

[0020] Assume that the following reciprocal function relationship holds between the parallax and the distance;

[0021]

[0022] where β = (β1, β2, β3) represents the ranging model parameters;

[0023] Finally, use numerical optimization techniques to solve the following unconstrained optimization problem to obtain the estimated value β of the ranging parameters;

[0024]

[0025] For the above solution, further, detect the bird coordinates in the image for the left and right eye images based on the small target detection algorithm under the aerial background.

[0026] For the above solution, further, based on the straight-line constraint, match the bird targets in the left and right eye images to find the same bird targets, including: correct the target coordinates in the right eye image using the binocular calibration parameters to obtain the corrected coordinates, then find the matching point pairs in the left and right eye images based on the straight-line constraint stereo matching algorithm, and calculate the parallax value of the target using formula (2), and obtain the distance d of the target relative to the detection system using formula (3).

[0027] For the above solution, further, combine the WGS84 longitude and latitude coordinates (lo1, la1), the azimuth angle Azi relative to the due north direction, the elevation angle Ele relative to the horizontal direction, the target distance d, and the earth radius R of the binocular observation system, and use the spherical triangle formula to obtain the WGS84 longitude and latitude coordinates (lo2, la2) and height H data of the target:

[0028] la2 = asin(sin(la1)*cos(d / R) + cos(la1)*sin(d / R)*cos(Azi);

[0029] lo2 = lo1 + atan2(sin(Azi)*sin(d / R)*cos(la1), cos(d / R) - sin(la1)*sin(la2)); (5)

[0030] H = d*sin(Ele).

[0031] In a second aspect, the present invention discloses a binocular stereo vision three-dimensional measurement and positioning system, including two cameras, a fixed bracket, an image acquisition subsystem, and an image processing subsystem;

[0032] The two cameras are used to adopt the same type of image detector and the same specification of optical lens, and maintain a certain distance and face the same scene; the cameras work with the same external trigger pulse to ensure the same image acquisition time;

[0033] The fixed bracket is used to fix the two cameras;

[0034] The image acquisition subsystem is used to collect the images of the fields of view of the two cameras and transmit them to the image processing subsystem;

[0035] The image processing subsystem is used to calibrate and correct the error caused by the non-parallel optical axes of the two cameras, pre-calculate the ranging parameters, use the ranging parameters combined with the target detection result to obtain the distance information of the target, and obtain the three-dimensional spatial coordinates of the target according to the azimuth angle, pitch angle data and distance information of the fixed bracket.

[0036] In a third aspect, the present invention discloses a bird repelling linkage method based on binocular stereo vision three-dimensional measurement and positioning, including:

[0037] Detect birds in the images according to the left and right eye images of the binocular observation system;

[0038] Combine the longitude, latitude, azimuth angle, pitch angle and bird distance of the binocular observation system to obtain the spatial position information of the bird target;

[0039] Utilize the spatial position information of the birds, combine the bird movement direction, the longitude and latitude data of each bird repelling device, and the fitting degree of the action range of the bird repelling device, and select the most suitable bird repelling device based on the system-level spatial registration algorithm and the hierarchical clustering algorithm;

[0040] For the directional sound wave device, start the directional sound wave device in advance according to the movement direction of the bird to aim at the activity area of the bird.

[0041] Further, the above solution includes setting the bird repelling system activation strategy to activate when a target is detected;

[0042] According to the spatial position information of the birds, link the bird repelling devices near the birds to work. Under the condition that the distance between the birds and the bird repelling devices is close, combine the position relationship between the birds and the runway, and preferentially activate the bird repelling devices that keep the birds away from the runway;

[0043] According to the target slice obtained by the binocular observation system, roughly judge the category of the bird by using artificial intelligence methods. For the sound wave bird repelling device, play the sound of the natural enemy of the bird to drive it away.

[0044] Further, according to the spatial position information of the birds, the bird repelling device is selected. When multiple bird repelling devices need to work jointly, the coordinate systems of the respective bird repelling devices need to be registered so that they work in a unified measurement coordinate system;

[0045] It is set that the order of magnitude of the distance between the binocular observation system and the bird repelling device is d, and the order of magnitude of the distances between the binocular observation system, the bird repelling device and the measured bird target is D;

[0046] When the difference between the order of magnitude d and the order of magnitude D is greater than two orders of magnitude, a platform-level spatial registration algorithm is adopted; a common coordinate system is constructed, and the measurement information of the bird repelling device located in its respective coordinate system is projected into the common coordinate system to achieve coordinate registration;

[0047] When the difference between the order of magnitude d and the order of magnitude D is less than two orders of magnitude, a system-level spatial registration method is used; the registration algorithm includes any one of the least squares method LS, the generalized least squares method GLS, the real-time quality control method RTQC, and the maximum likelihood method ML.

[0048] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0049] The present invention determines the three-dimensional information of all bird targets in the field of view through optical binocular calibration, stereo matching, and depth calculation, and solves the problem that the optical bird detection system cannot obtain the spatial position information of bird targets.

[0050] To solve the problem of the tolerance of birds to the bird repelling device, the present invention proposes a new bird repelling strategy: according to the provided spatial position of the bird target, the system-level spatial registration method and the hierarchical clustering algorithm are used to find the most suitable bird repelling device. For the acoustic wave bird repelling device, the natural enemy sound can be selected according to the bird species recognition result to precisely disperse the bird target. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 is a schematic diagram of the steps of the three-dimensional measurement and positioning method based on binocular stereo vision;

[0052] Figure 2 is a schematic diagram of the system framework of the present invention;

[0053] Figure 3 is a schematic diagram of the steps of the bird repelling linkage method based on binocular stereo vision three-dimensional measurement and positioning;

[0054] Figure 4 is the overall structure of the PP-HGNet backbone network. DETAILED DESCRIPTION OF THE INVENTION

[0055] The following further describes the embodiments of the invention in detail with reference to the accompanying drawings of the specification, so as to more clearly present the purpose, technical solution and technical effect of the present invention.

[0056] As Figure 1 shown, the present invention discloses a binocular stereo vision three-dimensional measurement and positioning method, including:

[0057] Collecting left and right eye images based on a binocular observation system to obtain original visual data; respectively performing bird target detection in the left and right eye images to determine the bird target objects to be processed;

[0058] Calculating binocular correction parameters to correct the error caused by the non-parallel optical axes of the left and right eye images; based on the straight-line constraint, matching the bird targets in the left and right eye images to find the same bird targets;

[0059] Calculating the parallax value of the same bird target; according to the ranging parameter and the parallax value, calculating the distance data of the bird target;

[0060] According to the longitude and latitude of the deployment location of the binocular observation system and the azimuth, pitch and distance of the optical system, calculating the three-dimensional spatial information of the bird target to realize the comprehensive spatial positioning and description of the bird target.

[0061] The calculation method for the binocular observation system to detect the target and obtain the target spatial position information includes the following steps:

[0062] 1. Calculate the correction parameters

[0063] The correction parameters are mainly used to correct the non-parallel problem between the two optical axes by using the affine transformation model. The method is to aim at a target at infinity or more than 3 km away, and make the target appear at N positions in the upper left, upper right, lower left, lower right, middle, etc. of the left and right eye images, N≥5; let its coordinates in the left and right eyes be: {(x 1,k , y 1,k ) k = 1, 2,..., N}, {(x 2,k , y 2,k ) k = 1, 2,..., N}, and use the affine transformation to calculate the affine transformation model parameters as: α = (α1, α2, α3, α4, α5, α6). This set of parameters is the correction parameters. In practical applications, with the left eye as the reference, the coordinates of the right eye are transformed to the left eye to correct the problem of non-parallel optical axes of the left and right eye cameras.

[0064] 2. Calculate the ranging parameters

[0065] Before the device leaves the factory, calculate the ranging parameters based on the fixed distance calibration points. For example, use the method of hanging a small ball by a drone to obtain the target images at M different distances, where it is assumed that the adjacent distance interval of M is 100 meters; detect the target coordinates. Assume that the target coordinates in the left eye are: {(x 3,k, y 3,k ) k = 1, 2, …, M}, and the target coordinates in the right eye are: {(x 4,k , y 4,k ) k = 1, 2, …, M}. Using the calibration parameters obtained above, the target coordinates in the right eye are calibrated to: {(x 5,k , y 5,k ) k = 1, 2, …, M}.

[0066]

[0067] Then the calculation formula for the parallax △ is as follows:

[0068]

[0069] Assume that the following reciprocal function relationship holds between the parallax and the distance:

[0070]

[0071] where β = (β1, β2, β3) represents the ranging model parameters.

[0072] Finally, numerical optimization techniques are used to solve the following unconstrained optimization problem to obtain the estimated value β of the ranging parameters.

[0073]

[0074] Collect images and detect the targets in the left and right eyes;

[0075] Collect the left and right eye images and detect the coordinates in the images based on the small target detection algorithm under the aerial background.

[0076] 3. Stereo matching

[0077] Use the pre-computed calibration parameters to calibrate the target coordinates in the right eye image to obtain the calibrated coordinates, then find the matching point pairs in the left and right eye images based on the stereo matching algorithm with line constraints, and calculate the parallax value of the target using formula (2), and obtain the distance d of the target relative to the detection system using formula (3).

[0078] 4. Calculate the spatial position information of the target

[0079] Combining the WGS84 longitude and latitude coordinates (lo1, la1) of the binocular observation system, the azimuth angle Azi relative to the true north direction, the pitch angle Ele relative to the horizontal direction, the target distance d, and the earth radius R, use the spherical trigonometry formula to obtain the WGS84 longitude and latitude coordinates (lo2, la2) and altitude H data of the target.

[0080] la2 = a * sin(sin(la1) * cos(d / R) + cos(la1) * sin(d / R) * cos(Azi));

[0081] lo2 = lo1 + atan2(sin(Azi) * sin(d / R) * cos(la1), cos(d / R) - sin(la1) * sin(la2)); (5)

[0082] H = d * sin(Ele).

[0083] As Figure 2 shown, the present invention discloses a binocular stereo vision three - dimensional measurement and positioning system, including two cameras, a fixed bracket, an image acquisition subsystem, and an image processing subsystem.

[0084] The two cameras are used to adopt the same type of image detector and the same specification of optical lens, and maintain a certain distance and face the same scene; the cameras work with the same external trigger pulse to ensure the same image acquisition time.

[0085] The fixed bracket is used to fix the two cameras.

[0086] The image acquisition subsystem is used to acquire the images of the fields of view of the two cameras and transmit them to the image processing subsystem.

[0087] The image processing subsystem is used to calibrate and correct the error caused by the non - parallel optical axes of the two cameras, pre - calculate the ranging parameters, use the ranging parameters combined with the target detection results to obtain the distance information of the target, and obtain the three - dimensional spatial coordinates of the target according to the azimuth angle, pitch angle data of the fixed bracket and the distance information.

[0088] The cameras can be symmetrically arranged relative to the fixed bracket. For example, the two cameras are placed horizontally. Of course, they can also be placed vertically; during the installation process, ensure that the optical axes of the two cameras are parallel. The cameras work with the same external trigger pulse to ensure the same image acquisition time. The image acquisition subsystem simultaneously acquires the images of the fields of view of the two cameras and transmits them to the image processing subsystem. Before formal use, it is necessary to calibrate and correct the error caused by the non - parallel optical axes of the system, pre - calculate the ranging parameters, use the ranging parameters combined with the target detection results to obtain the distance information of the target, and obtain the three - dimensional spatial coordinates of the target according to the azimuth angle, pitch angle data of the fixed bracket and the distance information.

[0089] As Figure 3 shown, the present invention discloses a bird - repelling linkage method based on binocular stereo vision three - dimensional measurement and positioning, including:

[0090] Detect birds in the image according to the left and right eye images of the binocular observation system;

[0091] Combine the longitude, latitude, azimuth, elevation angle of the binocular observation system and the distance of the bird to obtain the spatial position information of the bird target;

[0092] Utilize the spatial position information of the bird, combine the movement direction of the bird, the longitude and latitude data of each bird repellent device, and the fitting degree of the effective range of the bird repellent device, and select the most suitable bird repellent device based on the system-level spatial registration algorithm and the hierarchical clustering algorithm;

[0093] For the directional sound wave device, start the directional sound wave device in advance according to the movement direction of the bird and aim it at the activity area of the bird.

[0094] In the prior art, people simulate the sounds of the natural enemies of birds, such as the calls of raptors, which can quickly attract the attention of birds and make them stay away. For example, imitating the calls of raptors such as eagles and falcons, in addition to the sounds of natural enemies, certain specific alarm sounds can also play a role in driving away birds. These sounds usually have the characteristics of high decibels, suddenness or persistence, which can break the normal living environment of birds, make them feel uneasy and choose to leave. For example, some bird repellent devices emit high-frequency sound waves, which may be difficult for humans to detect, but are obvious interference signals for birds and can effectively drive them away.

[0095] Based on this, the steps to improve the bird repellent effect are as follows:

[0096] 1. It is necessary to clarify the types of birds to be driven away. Different birds have different reactions to different frequencies of sound, so choosing the appropriate sound is the key.

[0097] 2. Select the corresponding sound according to the natural enemies of the target birds. For the natural enemies of birds, the research results of biology can be used as a reference to determine their natural enemies. For example, if the target is a pigeon, the sound of an eagle can be selected; if it is a sparrow, the sound of an owl can be selected.

[0098] 3. Sound effect test: After selecting the sound, it can be tested in a small area to observe the reaction of birds to the sound. If the effect is not good, it may be necessary to adjust the frequency, volume or playback method of the sound.

[0099] 4. The way of playing the sound: Change from turning on regularly to turning on only when a bird target is detected, so that the sound wave has suddenness and simulates the dangerous signals in nature to achieve the purpose of driving away birds.

[0100] In order to reduce the tolerance of birds to bird repellent devices, the present invention intends to solve it by the following technical means:

[0101] 1. Change the opening strategy of the bird repellent device from the original regular opening to opening when a target is detected.

[0102] 2. Based on the spatial position information of the birds, the software links the bird repelling devices near the flying birds to work. Under the condition that the distance between the birds and the bird repelling devices is close, combined with the positional relationship between the birds and the runway, the bird repelling devices that keep the birds away from the runway are preferentially activated.

[0103] 3. Based on the target slices obtained by the binocular observation system, use artificial intelligence methods to roughly judge the category of the birds. For the acoustic wave bird repelling devices, play the sounds of the natural enemies of these birds to drive them away.

[0104] Among them, for the linked bird repelling devices, it is mainly to select appropriate bird repelling devices according to the spatial position information of the birds.

[0105] This involves the spatial registration problem of multiple bird repelling devices, specifically the spatial registration problem of the sensors in multiple bird repelling devices. When multiple bird repelling devices need to work jointly, it is necessary to register the coordinate systems of the sensors of each bird repelling device so that they work in a unified measurement coordinate system, and enable the sensors of each bird repelling device to identify the measurement information of a single sensor in the entire joint system.

[0106] It is set that the order of magnitude of the distance between the binocular observation system and the bird repelling device is d, and the order of magnitude of the distances between the binocular observation system, the bird repelling device and the measured bird target is D.

[0107] When the difference between the order of magnitude d and the order of magnitude D is greater than or equal to two orders of magnitude, a platform-level spatial registration algorithm is adopted; a common coordinate system is constructed, and the measurement information of the bird repelling devices located in their respective coordinate systems is projected into the common coordinate system to achieve coordinate registration.

[0108] When the difference between the order of magnitude d and the order of magnitude D is less than two orders of magnitude, a system-level spatial registration method is used; the registration algorithm includes any one of the least squares method LS, the generalized least squares method GLS, the real-time quality control method RTQC, and the maximum likelihood method ML.

[0109] For the method of bird recognition, the PP-HGNet (High Performance GPU Net) model can be adopted. The overall structure of the PP-HGNet backbone network is as Figure 4 shown.

[0110] After verification, the model can identify birds at the family level (phylum, class, order, family, genus, species), with an accuracy rate of 89%. In the actual application at the airport, it is not necessary to train and identify all bird targets in the world. Only the resident birds and migratory birds that often appear around a specific airport need to be trained and identified. Generally speaking, the number of bird species around an airport is in the range of 50 - 100, and the bird species also change in different seasons. The number and specific names of bird species can be determined by combining the ecological research report of the airport, and targeted training and identification work of bird species can be carried out to reduce the workload of training and identification while increasing the identification rate of bird species.

[0111] Test results

[0112] At a certain airport, a binocular optoelectronic detection system was used to conduct on-site tests on bird detection, bird identification, and coordinated bird repelling. The binocular detection system can effectively detect bird targets within a range of 1.5 km, and obtain the spatial position information and bird pictures of the targets. Based on the spatial position information, movement direction of the bird targets, and the longitude and latitude data of each bird repelling device, a suitable bird repelling device is selected and activated. For the acoustic bird repelling device, the sound of the natural enemies of this type of bird is selected in combination with the bird identification result to drive the bird targets. At the same time, for the safety of flight takeoff and landing, the command and control system receives ADSB information, and 1 - 2 minutes before the flight takeoff and landing, the bird repelling devices beside the runway are activated for repelling. In cooperation with the airport's strategies such as grass control, pest control, water control, and rodent control, the probability of bird strikes on flights is reduced to ensure flight safety. The above test lasted for two months. By comparing the bird activity heat maps before and after two months, it was found that the number of birds appearing in the airport had decreased significantly, and no bird strike incidents occurred.

[0113] The above are only the preferred and feasible embodiments of the present invention, and are not intended to limit the scope of the patent application of the present invention. Any equivalent changes, equivalent substitutions, or modified changes completed within the technical spirit and principles disclosed by the present invention shall be included within the scope of patent protection covered by the present invention.

Claims

1. A three-dimensional measurement and positioning method based on binocular stereo vision, characterized in that Including: Collect left and right eye images based on a binocular observation system to obtain original visual data; perform bird target detection in the left and right eye images respectively to determine the bird target objects to be processed. Calculate binocular calibration parameters to correct the error caused by the non-parallel optical axes of the left and right eye images; based on the line constraint, match the bird targets in the left and right eye images to find the same bird targets. Calculate the parallax of the same bird target; based on the ranging parameter and the parallax, calculate the distance data of the bird target. Calculate the three-dimensional spatial information of the bird target according to the longitude, latitude, azimuth, pitch, and distance of the binocular observation system deployment location, and realize the comprehensive spatial positioning and description of the bird target.

2. The three-dimensional measurement and positioning method based on binocular stereo vision according to claim 1, wherein: The steps of calculating the binocular calibration parameters include: Align the bird targets outside the distance requirement, so that the bird targets appear in N positions in the upper left, upper right, lower left, lower right, and upper middle of the left and right eye images, where N≥5. Suppose the coordinates of the bird target in the left and right eye images are respectively: {(x 1,k , y 1,k ) k = 1, 2, …, N}, {(x 2,k , y 2,k ) k = 1, 2, …, N}. The affine transformation model parameters are calculated by using affine transformation as: α = (α1, α2, α3, α4, α5, α6). This group of parameters is the binocular calibration parameters; Taking the left eye image as a reference, transform the coordinates of the right eye image to the left eye to correct the problem of non-parallel optical axes of the left and right eye cameras.

3. The three-dimensional measurement and positioning method based on binocular stereo vision according to claim 2, wherein: Calculating the binocular calibration parameters to correct the error caused by the non-parallel optical axes of the left and right eye images includes: Calculate the ranging parameters based on the fixed-distance calibration points, obtain the target images at M different distances, and detect the target coordinates. Assume that the target coordinates in the left-eye image are: {(x 3,k , y 3,k ) k = 1, 2, …, M}, and the target coordinates in the right-eye image that match them are: {(x 4,k , y 4,k ) k = 1, 2, …, M}. Using the binocular calibration parameters, correct the target coordinates in the right-eye image to: {(x 5,k , y 5,k ) k = 1, 2, …, M} Then the calculation formula of the parallax △ is as follows: Assume that the following reciprocal function relationship is satisfied between the parallax and the distance; where β=(β1,β2,β3) represents the ranging model parameters; Finally, use the numerical optimization technology to solve the following unconstrained optimization problem to obtain the estimated value β of the ranging parameter; 4. The three-dimensional measurement and positioning method based on binocular stereo vision according to claim 1, characterized in that: Detect the bird coordinates in the left and right eye images based on the small target detection algorithm under the aerial background.

5. The three-dimensional measurement and positioning method based on binocular stereo vision according to claim 3, characterized in that: Based on the line constraint, matching the bird targets in the left and right eye images to find the same bird targets includes: correcting the target coordinates in the right eye image using the binocular calibration parameters to obtain the corrected coordinates, then finding the matching point pairs in the left and right eye images based on the line constraint stereo matching algorithm, and calculating the parallax of the target using formula (2), and obtaining the distance d of the target relative to the detection system using formula (3).

6. The three-dimensional measurement and positioning method based on binocular stereo vision according to claim 5, wherein: Combining the WGS84 longitude and latitude coordinates (lo1, la1) of the binocular observation system, the azimuth angle Azi relative to the due north direction, the pitch angle Ele relative to the horizontal direction, the target distance d, and the earth radius R, use the spherical triangle formula to obtain the WGS84 longitude and latitude coordinates (lo2, la2) and height H data of the target:

7. Binocular stereo vision-based three-dimensional measurement and positioning system, characterized in that: Including two cameras, a fixed bracket, an image acquisition subsystem, and an image processing subsystem; Two cameras, which are used to adopt the same type of image detector and the same specification of optical lens, and keep a certain distance and face the same scene; the cameras work with the same external trigger pulse to ensure the same image acquisition time. The fixed bracket is used to fix the two cameras. The image acquisition subsystem is used to collect the field-of-view images of the two cameras and transmit them to the image processing subsystem. The image processing subsystem is used to calibrate and correct the error caused by the non-parallel optical axes of the two cameras, pre-calculate the ranging parameters, use the ranging parameters combined with the target detection results to obtain the distance information of the target, and obtain the three-dimensional spatial coordinates of the target according to the azimuth angle, pitch angle data, and distance information of the fixed bracket.

8. A bird repelling linkage method based on binocular stereo vision three-dimensional measurement and positioning, characterized in that The binocular stereo vision-based three-dimensional measurement and positioning method according to any one of claims 1-6 includes: Detecting birds in the images based on the left and right eye images of the binocular observation system; Obtaining the spatial position information of the bird target by combining the longitude, latitude, azimuth angle, pitch angle of the binocular observation system and the distance of the bird; Using the spatial position information of the bird, combining the bird movement direction, the longitude and latitude data of each bird repelling device, and the fitting degree of the action range of the bird repelling device, and selecting the most suitable bird repelling device based on the system-level spatial registration algorithm and the hierarchical clustering algorithm; For the directional sound wave device, start the directional sound wave device in advance according to the movement direction of the bird to aim at the activity area of the bird.

9. The bird repelling linkage method based on binocular stereo vision three-dimensional measurement and positioning according to claim 8, characterized in that: Setting the bird repelling system activation strategy to activate when a target is detected; Linking the operation of the bird repelling devices near the bird according to the spatial position information of the bird, and preferentially enabling the bird repelling devices that keep the bird away from the runway in combination with the positional relationship between the bird and the runway under the condition that the distance between the bird and the bird repelling device is close; Judging the category of the bird by using an artificial intelligence method according to the target slice obtained by the binocular observation system; for the sound wave bird repelling device, playing the sound of the natural enemy of the bird to drive it away.

10. The bird repelling linkage method based on binocular stereo vision three-dimensional measurement and positioning according to claim 9, characterized in that: Selecting a bird repelling device according to the spatial position information of the bird. When multiple bird repelling devices need to work together, it is necessary to register the coordinate systems of each bird repelling device so that they work in a unified measurement coordinate system; Setting that the order of magnitude of the distance between the binocular observation system and the bird repelling device is d, and the order of magnitude of the distances between the binocular observation system, the bird repelling device and the measured bird target is D; When the difference between the order of magnitude d and the order of magnitude D is greater than two orders of magnitude, adopt the platform-level spatial registration algorithm; construct a common coordinate system, and project the measurement information of the bird repelling device located in its respective coordinate system into the common coordinate system to achieve coordinate registration; When the difference between the order of magnitude d and the order of magnitude D is less than two orders of magnitude, use the system-level spatial registration method; the registration algorithm includes any one of the least squares method LS, the generalized least squares method GLS, the real-time quality control method RTQC, and the maximum likelihood method ML.

Images