A recognition method and device based on aircraft recognition AI scene

Abstract

The present invention proposes a recognition method and device based on aircraft recognition AI scene, taking photos of each preset scene, and classifying and storing the photos according to the preset scenes; judging whether the photos are clear; if so, playing all the photos in turn, and recognizing the preset scenes in the photos; recognizing the icons in the scenes in the photos, judging whether the icons are the icons in the corresponding preset scenes; obtaining the recognition rate of the aircraft's AI device according to the number of icons in the scenes in the photos that are successfully recognized. The present invention can improve the recognition rate of AI recognition scenes, making aircraft and recognition AI scenes more widely used, and also improving the mathematical model formula of aircraft.

Description

Technical Field

[0001] The present invention relates to the field of intelligent computing, and in particular to the technical field of an identification method and device based on aircraft identification AI scenes. Background Art

[0002] Among the existing aircraft, some are built based on aerodynamic mathematical models and are characterized by the symmetry of aerodynamic layout. They are divided into axisymmetric aircraft and mirror-symmetric aircraft. Most aircraft are equipped with artificial intelligence identifiers, and AI scene recognition technology has received widespread attention.

[0003] Application No. CN202010969611.2 proposes a test method and device, storage medium, and terminal for the recognition rate of AI scenes, the method comprising: receiving a standard recognition icon of a preset AI scene and a standard AI scene recognition atlas; controlling the playback of the standard AI scene recognition atlas, and controlling the recognition device to recognize the AI ​​scene in each picture; each time a picture is played, determining whether the recognition device determines the recognition icon of the AI ​​scene of the currently played picture after identification; if the recognition device obtains the recognition icon of the AI ​​scene of the currently played picture after identification, then comparing the recognition icon with the standard recognition icon to obtain the comparison result of the currently played picture; according to the comparison result of at least a part of the pictures in the standard AI scene recognition atlas, determining the recognition rate of the recognition device for the preset recognition AI scene. The present invention can improve the test efficiency and the accuracy of the test results. In the invention, it is limited to ordinary recognition AI devices, and does not mention the AI ​​device on the aircraft. Therefore, the present invention is required to improve the test method and device for the recognition rate of the AI ​​device on the aircraft.

[0004] Specifically, the aircraft performs AI scene recognition during flight, collects information about different scenes, and automatically adjusts the image effects for the collected scene information and improves the sharpness, exposure value, color temperature, etc. of the image, so that the image is more in line with the real scene. However, when the aircraft is flying too fast, there will be jitter, which causes the scene information to become blurred when the AI ​​scene recognizer recognizes the scene, and the specific scene information cannot be clearly obtained. The details of some scenes are not easy to display. In order to improve the recognition rate of AI scenes, it is necessary to test the AI ​​scene recognition technology. At this time, a recognition method and device based on aircraft recognition of AI scenes is particularly important. Summary of the invention

[0005] In order to solve the above problems, the present invention proposes an identification method and device based on aircraft identification AI scene, so as to more accurately solve the above-mentioned test method of the recognition rate of the AI ​​device on the aircraft.

[0006] The present invention is achieved through the following technical solutions:

[0007] The present invention proposes a recognition method based on aircraft recognition AI scene, comprising the following steps:

[0008] Take photos for each preset scene, and classify and store the photos according to the preset scenes;

[0009] Determining whether the photograph is clear;

[0010] If so, all the photographs are played in turn, and preset scenes in the photographs are identified;

[0011] Identify the icon captured in the scene in the photograph, and determine whether the icon is an icon in the corresponding preset scene;

[0012] The recognition rate of the AI ​​device of the aircraft is obtained according to the number of icons successfully recognized in the scene in the photograph.

[0013] Furthermore, the step of determining whether the photograph is clear includes:

[0014] The clarity of the photograph is detected by a preset clarity detection device; when the preset clarity detection device detects the clarity, the clarity is determined by detecting the resolution of the photograph, wherein the resolution range of the preset clarity detection equipment is 72dpi-450dpi.

[0015] Furthermore, after the step of determining whether the photograph is clear, the following steps are performed:

[0016] If not, adjustments are made to the aircraft's mathematical formulas.

[0017] Further, if not, the step of adjusting the mathematical formula of the aircraft includes:

[0018] Determining whether a mathematical formula of an aircraft is a linear function, wherein the aircraft is composed of an aerodynamic mathematical model;

[0019] If yes, the mathematical formula of the aerodynamic mathematical model is modified according to the Mach number, the altitude value and the Re number. If no, the aerodynamic coefficient based on the aerodynamic mathematical model is modified by changing the direction of the velocity vector.

[0020] Furthermore, if yes, then the step of playing all the photographs in turn and identifying the preset scenes in the photographs includes:

[0021] Playing the photographed photos in turn within a preset time, and taking screenshots of the photos during the playing process;

[0022] Send analog information to the preset identifier to keep the identifier in working state.

[0023] Furthermore, before the step of identifying the icons in the scene in the photograph, the step includes:

[0024] The icons in the scene and the photographs are processed by using a binary image processing technique, wherein the binary processing is to set the grayscale value of the points on the image to 0 or 255, so that the image appears in black and white.

[0025] Furthermore, the step of identifying the icon in the scene in the photograph and determining whether the icon is an icon in the corresponding preset scene includes:

[0026] Determine whether the grayscale value of the icon in the scene and the scene icon in the photograph is within a preset error value;

[0027] If yes, it is marked as an icon of successful recognition; if no, it is marked as an icon of unsuccessful recognition and not in the preset scene.

[0028] Further, if not, the step of marking it as an unsuccessful identification and not an icon in the preset scene includes:

[0029] Count the number of icons that failed to be recognized;

[0030] The unsuccessful recognition icon is marked as a red area and displayed on the photograph.

[0031] Furthermore, the step of obtaining the recognition rate of the AI ​​device of the aircraft according to the number of icons successfully recognized in the scene in the photograph includes:

[0032] The recognition rate of the preset scene is obtained by comparing the number of icons that failed to be recognized and the number of icons in the scene.

[0033] Obtain the recognition rate of each preset scene, and calculate the average of the recognition rates of all preset scenes;

[0034] The recognition rate of the AI ​​device of the aircraft is obtained based on the obtained average value.

[0035] The present invention provides a device for identifying AI scenes based on aircraft, comprising:

[0036] A shooting module is used to shoot each preset scene, and classify and store the taken photos according to the preset scenes;

[0037] A judging module, used to judge whether the photograph is clear;

[0038] A first recognition module, configured to, if yes, play all the photographs in turn and recognize the preset scenes in the photographs;

[0039] A second recognition module, used to recognize the icon in the scene in the photograph, and determine whether the icon is an icon in the corresponding preset scene;

[0040] The acquisition module is used to obtain the recognition rate of the AI ​​device of the aircraft according to the number of icons successfully recognized in the scene in the photograph.

[0041] Beneficial effects of the present invention:

[0042] By taking photos of each preset scene, and classifying and storing the photos according to the preset scenes, determining whether the photos are clear, playing the clear photos in turn, and identifying the preset scenes in the photos, and then identifying the icons in the photos that are taken in the scenes, the recognition rate of the AI ​​device of the aircraft is obtained according to the number of icons in the photos that are successfully recognized. The present invention can improve the recognition rate of AI recognition scenes, making the aircraft and the recognition AI scenes more widely used, and also improving the mathematical model formula of the aircraft. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a flow chart of a recognition method based on aircraft recognition AI scene of the present invention;

[0044] Figure 2 The present invention is a structural block diagram of a device for aircraft-based AI scene recognition. DETAILED DESCRIPTION

[0045] In order to more clearly and completely illustrate the technical solution of the present invention, the present invention is further described below in conjunction with the accompanying drawings.

[0046] Please refer to Figure 1 The present invention proposes a recognition method and device based on aircraft recognition AI scene, comprising the following steps:

[0047] S1, taking photos of each preset scene, and classifying and storing the photos according to the preset scenes;

[0048] S2, determining whether the photograph is clear;

[0049] S3, if yes, playing all the photographs in turn and identifying the preset scenes in the photographs;

[0050] S4, identifying an icon in the scene in the photograph, and determining whether the icon is an icon in the corresponding preset scene;

[0051] S5, obtaining the recognition rate of the AI ​​device of the aircraft according to the number of icons successfully recognized in the scene in the photograph.

[0052] In the above steps, first, it is necessary to shoot 360° for each preset scene, take multiple photos of the preset scenes, and classify the photos taken, classify and store them according to each preset scene, and then determine whether the photos taken are clear. Clear means that the resolution of the photos can reach 150dpi. If they are clear photos, all clear photos are played in turn and in a timely manner. The playback order is to play each scene one by one. During the playback process, the preset scene in the taken photo is identified to determine whether it is the preset scene. If it is determined to be the preset scene, the icon captured in the scene in the taken photo is identified. The identification method is to perform binary image processing on the corresponding taken photo and the picture of the preset scene, compare the two according to the grayscale value after the binarization processing, and determine whether it is the icon of the preset scene. Finally, the number of icons successfully identified is counted, and the ratio is made with the number of icons in the preset scene to obtain the recognition rate value, and then the average value is calculated according to the recognition rate value of each scene to obtain the recognition rate of the AI ​​device.

[0053] In a specific embodiment, the preset scenes are lawn, bedroom, kitchen, and beach.

[0054] In this embodiment, the step of determining whether the photograph is clear includes:

[0055] The clarity of the photograph is detected by a preset clarity detection device; when the preset clarity detection device detects the clarity, the clarity is determined by detecting the resolution of the photograph, wherein the resolution range of the preset clarity detection equipment is 72dpi-450dpi.

[0056] In a specific embodiment, the clarity detection equipment usually uses a spectrodensitometer, and the resolution range of the spectrodensitometer is 72dpi-450dpi. By scanning the image with the spectrodensitometer, the resolution of the image can be obtained, and it is determined whether the resolution of the image is 150dpi. Only when it reaches 150dpi, it has recognition value.

[0057] In a specific embodiment, if the image resolution is 60, the spectrophotometer cannot detect it and displays zero. If the image resolution is 72, the spectrophotometer can detect it, but the resolution is insufficient to be used as a material photo for identification. If the image resolution is greater than 450, it exceeds the detection range of the spectrophotometer, and the spectrophotometer will only display 450.

[0058] In this embodiment, after the step of determining whether the photograph is clear, the following steps are included:

[0059] If not, adjustments are made to the aircraft's mathematical formulas.

[0060] In a specific embodiment, the aircraft used is constructed based on an aerodynamic mathematical model. When the taken photos are not clear, it may be because the aircraft is not stable enough, and the mathematical formula of the aircraft needs to be adjusted.

[0061] In this embodiment, if not, the step of adjusting the mathematical formula of the aircraft includes:

[0062] Determining whether a mathematical formula of an aircraft is a linear function, wherein the aircraft is composed of an aerodynamic mathematical model;

[0063] If yes, the mathematical formula of the aerodynamic mathematical model is modified according to the Mach number, the altitude value and the Re number; if no, the aerodynamic coefficient based on the aerodynamic mathematical model is modified by changing the direction of the velocity vector.

[0064] In a specific embodiment, the linear function of the aircraft is a common function used to control the performance of the aircraft. If so, the dependent variable and independent variable of the function can be set by changing the Mach number, altitude value and Re number of the linear function. Mach number 1 is one times the speed of sound (sound waves can propagate in solid, liquid or gaseous media. The greater the density of the medium, the faster the speed of sound, so the size of Mach is not fixed): Mach number less than 1 is subsonic, and Mach number greater than 5 is hypersonic; Mach number is the ratio of the speed of flight to the speed of sound at that time. Greater than 1 means faster than the speed of sound. Similarly, less than 1 is slower than the speed of sound. Re=ρvL / μ, ρ, μ are fluid density and dynamic viscosity coefficient, v, L are characteristic speed and characteristic length of the flow field. Reynolds number physically represents the ratio of inertial force and viscosity force level. For external flow problems, v and L generally take the far front flow velocity and the main size of the object (such as the chord length of the wing or the diameter of a sphere); for internal flow problems, the average flow velocity in the channel and the channel diameter are taken. If the Reynolds numbers of two geometrically similar flow fields are equal, then the ratio of the inertial force to the viscous force of the corresponding micro-groups is equal. If not, then when the Mach number, altitude value and Re number are fixed parameters, the aerodynamic coefficient will change due to the direction of the velocity vector, the deflection angle of the control surface and the angular velocity of rotation relative to the center of gravity. The deflection angle includes pitch rudder deflection and rudder deflection, and the two have a vector and directional relationship. The aerodynamic coefficient will change accordingly according to the direction and size of the two, thereby ensuring the stability of the aircraft.

[0065] In this embodiment, if yes, then the step of playing all the photographs in turn and identifying the preset scenes in the photographs includes:

[0066] Playing the photographed photos in turn within a preset time, and taking screenshots of the photos during the playing process;

[0067] Send analog information to the preset identifier to keep the identifier in working state.

[0068] In a specific embodiment, when the taken photos are played in turn, they should be played scene by scene to ensure that one scene is fully recognized before the next scene is recognized. The rotation playback is set to 10 seconds to switch to the next picture. Screenshot recognition can better record whether the recognition results are the same after multiple recognitions. The screenshot picture with the most recognition can be used for the next icon recognition. The recognizer is a mobile phone, and it is necessary to send simulated information to the mobile phone to ensure that the mobile phone can maintain a long working state and can perform long-term and efficient image recognition.

[0069] In this embodiment, before the step of identifying the icon captured in the scene in the photograph, the method includes:

[0070] The icons in the scene and the photographs are processed by using a binary image processing technique, wherein the binary processing is to set the grayscale value of the points on the image to 0 or 255, so that the image presents an obvious black and white effect.

[0071] In a specific embodiment, the binarization image processing technology is a fixed algorithm, namely fixed threshold binarization. The algorithm itself is very simple. Many machine vision processes are 8-bit grayscale images with grayscale values ​​ranging from 0 to 255. The so-called binarization is to give a threshold, and let the grayscale value less than this threshold be 0, and the grayscale value greater than or equal to this threshold be set to 255. In this way, a black and white image is displayed on the image. After being processed into a black and white image, the difference between the two icons can be better compared.

[0072] In this embodiment, the step of identifying the icon in the scene in the photograph and determining whether the icon is an icon in the corresponding preset scene includes:

[0073] Determine whether the grayscale value of the icon in the scene and the scene icon in the photograph is within a preset error value;

[0074] If yes, it is marked as an icon of successful recognition; if no, it is marked as an icon of unsuccessful recognition and not in the preset scene.

[0075] In a specific embodiment, the preset error range is set to 0-3 when comparing at the icon 0 threshold. If it is within the error range, it can be judged as a successful icon recognition. If the error is above 3, the recognition fails. To write the actual, such as in a specific embodiment, its error value is 4, which exceeds the preset error value 3, the recognition fails

[0076] In another specific embodiment, the preset error range is set to 0-5 when compared at the icon 255 threshold. After the judgment is completed, if it is within the error range, it can be judged as a successful recognition icon. If the error is above 5, the recognition fails.

[0077] In this embodiment, if not, the step of marking as unsuccessful recognition and not being an icon in the preset scene includes:

[0078] Count the number of icons that failed to be recognized;

[0079] The unsuccessful recognition icon is marked as a red area and displayed on the photograph.

[0080] In a specific embodiment, the number of each icon that is not successfully recognized is counted and displayed as a red area on the photograph, so that it can be determined which icon cannot be recognized, so as to facilitate data adjustment of the aircraft's AI device after the recognition is completed.

[0081] In this embodiment, the step of obtaining the recognition rate of the AI ​​device of the aircraft according to the number of icons successfully recognized in the scene in the photograph includes:

[0082] The recognition rate of the preset scene is obtained by comparing the number of icons that failed to be recognized and the number of icons in the scene.

[0083] Obtain the recognition rate of each preset scene, and calculate the average of the recognition rates of all preset scenes;

[0084] The recognition rate of the AI ​​device of the aircraft is obtained based on the obtained average value.

[0085] In a specific embodiment, the recognition rate of the AI ​​device in the scene is obtained by comparing the number of icons that failed to be recognized with the number of icons in the preset scene. Then, the recognition rate of the AI ​​device of the aircraft is obtained by taking the average value based on the recognition rate of the AI ​​device in each preset scene.

[0086] In summary, the first step is to use the AI ​​device to shoot each preset scene. When shooting, it is necessary to take multiple photos at different angles. After completing the shooting of each scene, the photos taken need to be classified and stored. They are classified according to each scene, and then it is necessary to determine whether the photos are clear. The clarity detection equipment usually uses a spectrodensitometer. The resolution range of the spectrodensitometer is 72dpi-450dpi. By scanning the picture with the spectrodensitometer, the resolution of the picture can be obtained, and it is determined whether the resolution of the picture is 150dpi. If it is reached, the photos should be played scene by scene when they are played in turn to ensure that one scene is fully recognized before proceeding to the next scene. In addition, the rotation playback is set to 10s to switch to the next picture. Screen capture recognition can better record whether the recognition results are the same after multiple recognitions. The screenshot picture with the most recognition can be used for the next icon recognition. The recognizer is a mobile phone, and it is necessary to simulate the information sending to the mobile phone to ensure that the mobile phone can maintain a long working state and can perform long-term and efficient image recognition. If it cannot be achieved, the mathematical formula of the aircraft is adjusted. It is necessary to first determine whether the mathematical formula of the aircraft is a linear function, and the aircraft is composed of an aerodynamic mathematical model. If so, the dependent variable and independent variable of the function can be set by changing the Mach number, altitude value and Re number of the linear function. If not, the Mach number, altitude value and Re number are fixed. In the case of parameters, the aerodynamic coefficient will change due to the direction of the velocity vector, the deflection angle of the control surface and the angular velocity of rotation relative to the center of gravity, among which the deflection angle includes pitch rudder deflection and rudder deflection, and there is a vector and directional relationship between the two. The aerodynamic coefficient will change accordingly according to the direction and size of the two, thereby ensuring the stability of the aircraft. Then, after identifying the preset scene, the icons in the scene and the photographs taken are processed by binary image processing technology, wherein the binarization processing is to set the grayscale value of the points on the image to 0 or 255, that is, to present an obvious black and white effect to the entire image, and to determine whether the grayscale values ​​of the icons in the scene and the scene icons in the photographs are within the preset error, and the preset error range is set to the icon 0 threshold. When comparing, the error range is 0-3. When comparing at the icon 255 threshold, the error range is 0-5. After the judgment is completed, if it is within the error range, it can be judged as an icon that has been successfully recognized. If the error is above 3 when comparing at the 0 threshold, the recognition fails. When comparing at the 255 threshold, the error is above 5, the recognition fails. It is not an icon in the preset scene. It is necessary to count the number of each icon that is not successfully recognized, and it will be displayed as a red area on the photo. Finally, the ratio of the number of icons that are not successfully recognized to the number of icons in the preset scene can be used to obtain the recognition rate of the AI ​​device in the scene. Then, based on the recognition rate of the AI ​​device of each preset scene, the average value is taken to obtain the recognition rate of the AI ​​device of the aircraft.

[0087] Please refer to Figure 2 The present invention proposes an aircraft based on an aerodynamic mathematical model and a device for identifying the recognition rate of an AI scene, comprising:

[0088] The shooting module 10 is used to shoot each preset scene, and classify and store the taken photos according to the preset scene;

[0089] A judging module 20, used to judge whether the photograph is clear;

[0090] A first recognition module 30, for, if yes, playing all the photographs in turn, and recognizing the preset scenes in the photographs;

[0091] A second recognition module 40 is used to recognize the icon in the scene in the photograph, and determine whether the icon is an icon in the corresponding preset scene;

[0092] The acquisition module 50 is used to acquire the recognition rate of the AI ​​device of the aircraft according to the number of icons successfully recognized in the scene in the photograph.

[0093] In this embodiment, first, the shooting module 10 will shoot each preset scene through the AI ​​device, wherein the shooting needs to shoot the scene from multiple angles, and after shooting, the photos of each preset scene will be classified and stored, and then, the judgment module 20 will judge whether the photograph is clear, wherein the device used for judging the clarity is a spectrodensitometer, and the clarity judgment needs to check its resolution, and the resolution range that the spectrodensitometer can detect is 72dpi-450dpi. The first recognition module 30, if it is a clear photograph, sets a timed playback, and identifies the preset scene of the photograph during the playback, wherein the device used for recognition is a mobile phone, and then the second recognition module 40 will judge whether the icon in the scene in the photograph is the corresponding preset scene. Icon, wherein the method used to identify the icon is binary image processing. After processing, a black and white image can be obtained, and then the grayscale value of the black and white image is compared to see if it is within the preset error value. The preset error range is set to 0-3 when compared at the icon 0 threshold, and 0-5 when compared at the icon 255 threshold. After the judgment is completed, if it is within the error range, it can be judged as an icon that has been successfully recognized. If the error is greater than 3 when compared at the 0 threshold, the recognition fails. If the error is greater than 5 when compared at the 255 threshold, the recognition fails. Finally, the acquisition module 50 will obtain the recognition rate of the scene based on the statistical number of successful recognitions and the number of icons of the preset scene. The average recognition rate of each preset scene is calculated to obtain the recognition rate of the aircraft's AI device.

[0094] In one embodiment, the determination module 20 includes:

[0095] The detection unit is used to detect the clarity of the taken photo through a preset clarity detection equipment; when the preset clarity detection equipment detects the clarity, the clarity is determined by detecting the resolution of the taken photo, wherein the resolution range of the preset clarity detection equipment is 72dpi-450dpi.

[0096] The adjustment submodule is used to adjust the mathematical formula of the aircraft if the answer is no.

[0097] A judging subunit, used to judge whether the mathematical formula of the aircraft is a linear function, wherein the aircraft is composed of an aerodynamic mathematical model;

[0098] The judging subunit, if yes, modifies the mathematical formula of the aerodynamic mathematical model according to the Mach number, the altitude value and the Re number; if no, modifies the data of the aerodynamic coefficient based on the aerodynamic mathematical model by changing the direction of the velocity vector.

[0099] In this embodiment, the detection unit first detects the clarity of the photographed photo and detects the resolution of the photographed photo. The device used is a spectrodensitometer, and the detectable resolution is 72dpi-450dpi. The high-definition photo required is greater than or equal to 150dpi. The adjustment submodule, if not, adjusts the mathematical formula of the aircraft. The judgment subunit judges whether the mathematical formula of the aircraft is a linear function, wherein the aircraft is composed of an aerodynamic mathematical model. The judgment subunit, if so, modifies the mathematical formula of the aerodynamic mathematical model according to the Mach number, the altitude value and the Re number. If not, the aerodynamic coefficient based on the aerodynamic mathematical model is changed by changing the direction of the velocity vector. The aerodynamic coefficient will change due to the direction of the velocity vector, the deflection angle of the control surface and the rotational angular velocity relative to the center of gravity. The deflection angle includes pitch rudder deflection and rudder deflection, and the two have a vector and direction relationship. The aerodynamic coefficient will change accordingly according to the direction and size of the two, thereby ensuring the stability of the aircraft.

[0100] In one embodiment, the first identification module 30 includes:

[0101] A playback unit, used to play the photographs in turn within a preset time and take screenshots of the photographs during playback;

[0102] The transmission unit is used for the control end to transmit analog information to the preset identifier to keep the identifier in a working state.

[0103] In this embodiment, when the playback unit plays the taken photos in turn, it should play them scene by scene to ensure that one scene is fully recognized before proceeding to the next scene recognition. The rotation playback is set to 10s to switch to the next picture. Screenshot recognition can better record whether the recognition results are the same after multiple recognitions. The screenshot picture with the most recognition can be taken for the next icon recognition. The transmission unit, the identifier is a mobile phone, and it is necessary to transmit analog information to the mobile phone to ensure that the mobile phone can maintain a long working state and can perform long-term and efficient image recognition.

[0104] In one embodiment, the second identification module 40 includes:

[0105] A processing submodule, used to process the icons in the scene and the photographed photos by using a binary image processing technique, wherein the binary processing is to set the grayscale value of the points on the image to 0 or 255, so that the image presents an obvious black and white effect;

[0106] A judging unit, used to judge whether the grayscale value of the icon in the scene and the scene icon in the photograph is within a preset error value;

[0107] The judging unit, if yes, marks it as an icon of successful recognition; if no, marks it as an icon of unsuccessful recognition and not in the preset scene;

[0108] A counting subunit, used to count the number of icons that failed to be recognized;

[0109] The marking subunit is used to mark the unsuccessfully recognized icon as a red area and display it on the photograph.

[0110] In this embodiment, the processing submodule processes the icons in the scene and the captured photos through the binarization image processing technology. The so-called binarization is to give a threshold value, set the grayscale value less than the threshold value to 0, and set the grayscale value greater than or equal to the threshold value to 255, so that a black and white image is displayed on the image. The judgment unit judges whether the grayscale values ​​of the icons in the scene and the scene icons in the captured photos are within a preset error range. The preset error range is set to 0-3 when compared at the icon 0 threshold, and 0-5 when compared at the icon 255 threshold. After the judgment is completed, if it is within the error range, it can be judged as an icon that has been successfully recognized. If the error is greater than 3 when compared at the 0 threshold, the recognition fails. If the error is greater than 5 when compared at the 255 threshold, the recognition fails. The statistical subunit counts the number of icons that have not been successfully recognized. The calibration subunit marks the icons that have not been successfully recognized as red areas and displays them on the captured photos.

[0111] In one embodiment, the acquisition module 50 includes:

[0112] A ratio unit is used to compare the number of icons that are unsuccessfully recognized with the number of icons in the scene to obtain a recognition rate of the preset scene;

[0113] A calculation unit, used to obtain the recognition rate of each preset scene and calculate the average value of the recognition rates of all preset scenes;

[0114] The acquisition unit is used to acquire the recognition rate of the AI ​​device of the aircraft according to the obtained average value.

[0115] In this embodiment, a ratio unit is used to compare the number of icons that failed to be recognized with the number of icons in the preset scene, so as to obtain the recognition rate of the AI ​​device in the scene. The calculation unit calculates the average value according to the recognition rate of the AI ​​device in each preset scene, and the acquisition unit obtains the recognition rate of the AI ​​device of the aircraft.

[0116] In summary: In the specific implementation, the shooting module 10 first takes all-round photos of the preset scene, and classifies and stores the photos. Then, in the judgment module 20, the detection unit first detects the clarity of the photos, and uses a spectrodensitometer to detect the resolution of the photos. The detectable resolution is 72dpi-450dpi, and the required high-definition photos are greater than or equal to 150dpi. The adjustment submodule, if not, adjusts the mathematical formula of the aircraft. The judgment subunit determines whether the mathematical formula of the aircraft is a linear function, wherein the aircraft is composed of an aerodynamic mathematical model. The judgment subunit, if so, modifies the mathematical formula of the aerodynamic mathematical model according to the Mach number, altitude value and Re number. If not, through Changing the direction of the velocity vector changes the data of the aerodynamic coefficient based on the mathematical model of aerodynamic force, wherein the aerodynamic coefficient will change due to the direction of the velocity vector, the deflection angle of the control surface and the angular velocity of rotation relative to the center of gravity, wherein the deflection angle includes the pitch rudder deflection and the rudder deflection, and there is a vector and direction relationship between the two, and the aerodynamic coefficient will change accordingly according to the direction and size of the two, thereby ensuring the stability of the aircraft. The first recognition module 30, when the playback unit plays the taken photos in turn, it should play them scene by scene to ensure that one scene is fully recognized before the next scene is recognized. The rotation playback is set to 10s to switch to the next picture. Screen capture recognition can better record whether the recognition results are the same after multiple recognitions. The screenshot with the most recognition can be taken for the next step of icon recognition. The transmission unit, the identifier is a mobile phone, and it is necessary to transmit analog information to the mobile phone to ensure that the mobile phone can maintain a long working state and can perform long-term and efficient image recognition. Then, the second recognition module 40, the processing submodule, processes the icons in the scene and the captured photos through the binarization image processing technology. The so-called binarization is to give a threshold value, so that the gray value less than the threshold value is 0, and the gray value greater than or equal to the threshold value is set to 255, so that a black and white image is displayed on the image. The judgment unit judges whether the gray value of the icon in the scene and the scene icon in the captured photo is within a preset error range. The preset error range is set to compare at the icon 0 threshold. When the icon is compared at the 255 threshold, the error range is 0-3. When the icon is compared at the 255 threshold, the error range is 0-5. After the judgment is completed, if it is within the error range, it can be judged as an icon that has been successfully recognized. If the error is greater than 3 when compared at the 0 threshold, the recognition fails. When the error is greater than 5 when compared at the 255 threshold, the recognition fails. The statistical subunit counts the number of icons that have not been successfully recognized. The calibration subunit marks the icons that have not been successfully recognized as red areas and displays them on the photograph. Finally, the acquisition module 50 acts as a ratio unit, and the ratio of the icons that have not been successfully recognized to the number of icons in the preset scene can be obtained. The recognition rate of the AI ​​device in the scene can be obtained. The calculation unit, according to the recognition rate of the AI ​​device in each preset scene,Calculate the average value, obtain the unit, and obtain the recognition rate of the aircraft's AI device.

[0117] In summary, the present invention proposes a recognition method and device based on aircraft recognition AI scene. The shooting module 10 first uses the AI ​​device to shoot each preset scene. When shooting, it is necessary to shoot multiple photos at different angles, and after completing the shooting of each scene, it is necessary to classify and store these photos. The classification method is to classify according to each scene, and then the judgment module 20 determines whether the taken photos are clear. The detection equipment is a spectrodensitometer. The scanned resolution is 72dpi-450dpi. It is necessary to determine whether the taken photos reach 150dpi. Then the first recognition module 30, if it reaches, when the taken photos are played in turn, they should be played scene by scene to ensure that a scene After the scene is completely recognized, the next scene recognition is carried out. The rotation playback is set to 10s to switch to the next picture. Screenshot recognition can better record whether the recognition results are the same after multiple recognitions. The screenshot picture with the most recognition can be used for the next icon recognition. The identifier is a mobile phone, and it is necessary to simulate the information sending of the mobile phone to ensure that the mobile phone can maintain a long working state and can perform long-term and efficient image recognition. If it cannot be achieved, the mathematical formula of the aircraft is adjusted. It is necessary to first determine whether the mathematical formula of the aircraft is a linear function, and the aircraft is composed of an aerodynamic mathematical model. If so, the dependent variable and independent variable of the function are set by changing the Mach number, altitude value and Re number of the linear function. If not, in the Mach When the parameters of Hz, altitude and Re are fixed, the aerodynamic coefficient will change due to the direction of the velocity vector, the deflection angle of the control surface and the angular velocity of rotation relative to the center of gravity. The deflection angle includes pitch rudder deflection and rudder deflection, and there is a vector and directional relationship between the two. The aerodynamic coefficient will change accordingly according to the direction and size of the two, thereby ensuring the stability of the aircraft. The second recognition module 40 first performs binary image processing on the icons of the preset scene and the icons of the taken photos, wherein the binarization processing is to set the grayscale value of the points on the image to 0 or 255, that is, to present an obvious black and white effect on the entire image, and to judge whether the grayscale values ​​of the icons in the scene and the scene icons in the taken photos are within a preset error range, and the preset error range is set to within When the icon is compared at the 0 threshold, the error range is 0-3, and when the icon is compared at the 255 threshold, the error range is 0-5. After the judgment is completed, if it is within the error range, it can be judged as an icon that has been successfully recognized. If the error is greater than 3 when compared at the 0 threshold, the recognition fails. When compared at the 255 threshold, the error is greater than 5, the recognition fails. It is necessary to count the number of each icon that has not been successfully recognized, and it will be displayed as a red area on the photograph. Finally, the unsuccessful recognition icons obtained by the acquisition module 50 are compared with the number of icons in the preset scene, and the recognition rate of the AI ​​device in the scene can be obtained. Then, according to the recognition rate of the AI ​​device of each preset scene, the average value is taken to obtain the recognition rate of the AI ​​device of the aircraft.

[0118] Of course, the present invention may have many other implementations. Based on this implementation, other implementations obtained by ordinary technicians in this field without any creative work are all within the scope of protection of the present invention.

Claims

1. A recognition method based on aircraft recognition AI scene, characterized in that: The following steps are involved: Take photos for each preset scene, and classify and store the photos according to the preset scenes; Determining whether the photograph is clear; If so, all the photographs are played in turn, and preset scenes in the photographs are identified; Identify the icon captured in the scene in the photograph, and determine whether the icon is an icon in the corresponding preset scene; Obtaining a recognition rate of the AI ​​device of the aircraft according to the number of icons successfully recognized in the scene in the photograph; The step of determining whether the photograph is clear includes: Testing the clarity of the photographed photo by using a preset clarity testing device; when the preset clarity testing device is testing the clarity, the clarity is determined by testing the resolution of the photographed photo, wherein the resolution range of the preset clarity testing device is 72dpi-450dpi; After the step of determining whether the photograph is clear, the method further comprises: If not, then make adjustments to the aircraft’s mathematical formulas; If not, the step of adjusting the mathematical formula of the aircraft includes: Determining whether a mathematical formula of an aircraft is a linear function, wherein the aircraft is composed of an aerodynamic mathematical model; If yes, the mathematical formula of the aerodynamic mathematical model is modified according to the Mach number, the altitude value and the Re number. If no, the aerodynamic coefficient based on the aerodynamic mathematical model is modified by changing the direction of the velocity vector. If so, the step of playing all the photographs in turn and identifying the preset scenes in the photographs includes: Playing the photographed photos in turn within a preset time, and taking screenshots of the photos during the playing process; Send analog information to the preset identifier to keep the identifier in working state.

2. The recognition method based on aircraft recognition AI scene according to claim 1 is characterized in that: Before the step of identifying the icon in the scene in the photograph, the method includes: The icons in the scene and the photographs are processed by using a binary image processing technique, wherein the binary processing is to set the grayscale value of the points on the image to 0 or 255, so that the image appears in black and white.

3. The recognition method based on aircraft recognition AI scene according to claim 1 is characterized in that: The step of identifying the icon in the scene in the photograph and determining whether the icon is an icon in the corresponding preset scene includes: Determine whether the grayscale value of the icon in the scene and the scene icon in the photograph is within a preset error value; If yes, it is marked as an icon of successful recognition; if no, it is marked as an icon of unsuccessful recognition and not in the preset scene.

4. The recognition method based on aircraft recognition AI scene according to claim 3 is characterized in that: If not, the step of marking as unsuccessful recognition and not being an icon in the preset scene includes: Count the number of icons that failed to be recognized; The unsuccessful recognition icon is marked as a red area and displayed on the photograph.

5. The recognition method based on aircraft recognition AI scene according to claim 1, characterized in that: The step of obtaining the recognition rate of the AI ​​device of the aircraft according to the number of icons successfully recognized in the scene in the photograph comprises: The recognition rate of the preset scene is obtained by comparing the number of icons that failed to be recognized and the number of icons in the scene. Obtain the recognition rate of each preset scene, and calculate the average of the recognition rates of all preset scenes; The recognition rate of the AI ​​device of the aircraft is obtained based on the obtained average value.

6. A device for identifying AI scenes based on aircraft, applied to a method for identifying AI scenes based on aircraft as claimed in any one of claims 1 to 5, characterized in that: include: A shooting module is used to shoot each preset scene, and classify and store the taken photos according to the preset scenes; A judging module, used to judge whether the photograph is clear; A first recognition module, configured to, if yes, play all the photographs in turn and recognize the preset scenes in the photographs; A second recognition module, used to recognize the icon in the scene in the photograph, and determine whether the icon is an icon in the corresponding preset scene; The acquisition module is used to obtain the recognition rate of the AI ​​device of the aircraft according to the number of icons successfully recognized in the scene in the photograph.

Images