Three-super platform test method based on image recognition

Abstract

The invention discloses a three-super platform test method based on image recognition. The method comprises the following steps: constructing an image recognition model and a displacement-current-angle model; obtaining a driving current value and a displacement remote measurement value of the to-be-measured three-super platform under the current to-be-measured working condition, substituting the driving current value and the displacement remote measurement value into the displacement-current-angle model, and calculating to obtain a predicted rotation angle phi; obtaining a standing image of the to-be-detected three-super platform under the current to-be-detected working condition, and determining a plane line rotation angle phi'of the to-be-detected three-super platform maneuvering platform through the image recognition model; calculating a difference rate I according to phi and phi '; and according to the difference rate I obtained through calculation, the consistency of the to-be-tested three-super platform maneuvering component is judged. By means of the method, closed-loop correctness testing and long-term action consistency testing can be introduced into the three-hyper platform maneuvering component, action correctness and component health results of the three-hyper platform maneuvering mechanism are given, and the testing results are effective and reliable.

Description

Technical Field

[0001] The invention belongs to the technical field of spacecraft testing, and in particular relates to a three-super platform testing method based on image recognition. Background Art

[0002] The current spacecraft has increased the agility and stability of maneuvers by adding three super platforms, which has greatly improved the satellite's related indicators. Among them, the three super platforms are a computing and processing unit of the onboard control subsystem. The three super platforms are mainly composed of maneuvering platforms and maneuvering components. Through the pointing and damping of the maneuvering components, high stability and high pointing accuracy at the imaging moment are achieved, thereby realizing the satellite's ultra-precise, ultra-stable and ultra-agile maneuvering imaging.

[0003] At present, ground testing does not have relevant tests for the three-super-platform, especially for the mobile parts, which often rely on visual confirmation of the status, mainly facing the following problems:

[0004] 1) Since most motorized parts move at the millimeter level, there is a certain possibility of misinterpretation when relying on visual confirmation;

[0005] 2) It is impossible to retain relevant physical test records, which makes it difficult to trace the test data later;

[0006] 3) It is impossible to judge whether the maneuvering of the motorized parts into the closed loop is correct;

[0007] 4) The consistency of input and output of motorized parts cannot be effectively judged; Summary of the invention

[0008] The technology of the present invention solves the problem: overcomes the shortcomings of the prior art, provides a three-ultrasound platform testing method based on image recognition, can realize the introduction of closed-loop correctness test and long-term action consistency test of the three-ultrasound platform's motorized components, and gives the correct action and component health results of the three-ultrasound platform's motorized mechanism, and the test results are effective and reliable.

[0009] In order to solve the above technical problems, the present invention discloses a three-super platform testing method based on image recognition, comprising:

[0010] Construct image recognition model and displacement-current-angle model;

[0011] Obtain the driving current value and displacement telemetry value of the three-ultrasound platform under the current working condition, substitute them into the displacement-current-angle model, and calculate the predicted rotation angle Φ;

[0012] Obtain a static image of the three-ultrasound platform to be tested under the current working condition to be tested, and determine the plane line rotation angle Φ′ of the mobile platform of the three-ultrasound platform to be tested through the image recognition model;

[0013] According to Φ and Φ′, the difference rate I is calculated;

[0014] According to the calculated difference rate I, the consistency of the mobile components of the three-ultrasound platform to be tested is judged; wherein, the smaller the difference rate I is, the better the consistency of the mobile components of the three-ultrasound platform to be tested is.

[0015] In the above-mentioned three-super platform testing method based on image recognition, the construction process of the image recognition model is as follows:

[0016] The three-super-platform mobile platform installed on the satellite is imaged continuously at multiple angles using a photographing device to obtain sample images;

[0017] Based on the manual annotation method, the plane lines of the three super-platform maneuvering platforms in the sample images are annotated to obtain the dataset for YOLOv5 model migration;

[0018] Based on the obtained data set of YOLOv5 model migration, the YOLOv5 model is trained, and through parameter optimization, the YOLOv5 model accurately outputs the recognized plane lines of the three-super platform maneuvering platform to obtain an image recognition model.

[0019] In the above-mentioned three-super platform testing method based on image recognition, the construction process of the displacement-current-angle model is as follows:

[0020] Obtain a static image of the motorized components of the three-ultrasound platform when the driving current is 0, and identify the plane line of the motorized platform of the three-ultrasound platform through the image recognition model as the reference plane line, and the corresponding rotation angle is recorded as Φ0;

[0021] Within the driving current range, the driving current output is carried out according to the normal distribution to obtain n groups of current values ​​A1, A2, ..., A n and the corresponding displacement telemetry values ​​L1, L2, ..., L n ;

[0022] Get the three super platform motor parts at a driving current of A i The static image of the case is used to identify the plane line of the three-supervision platform maneuvering platform through the image recognition model, and the coincidence analysis is performed with the reference plane line to calculate the corresponding rotation angle Φ i ;where,i=1,2,...,n,A i ≠0;

[0023] Get n sets of displacement-current-angle data (L1, A1, Φ i );

[0024] For the n sets of displacement-current-angle data (L1, A1, Φ i) is fitted to obtain the displacement-current-angle model: Φ=f(L,A); where f() represents the displacement-current-angle fitting function.

[0025] In the above-mentioned three-super platform test method based on image recognition, the driving current range that can be input to the motorized components of the three-super platform is [A min ,A max ]; among them, A min Indicates the minimum driving current value that can be input to the motorized components of the three-super platform, A max Indicates the maximum driving current value that can be input to the motorized components of the three-super platform, A i ∈[A min ,A max ].

[0026] In the above three-super platform test method based on image recognition, the rotation angle Φ i , used to indicate whether the movement of the three-supervisory platform's maneuvering parts is correct.

[0027] In the above three-super platform testing method based on image recognition, the difference rate I is calculated as follows:

[0028]

[0029] In the above-mentioned three-super platform testing method based on image recognition, it also includes: when I≥5%, a ground warning is given.

[0030] In the above-mentioned three-super platform testing method based on image recognition, it also includes: obtaining historical difference rate and performing parameter correction on the image recognition model and / or the displacement-current-angle model.

[0031] The present invention has the following advantages:

[0032] The present invention discloses a three-supervision platform testing method based on image recognition, which can realize the closed-loop correctness test and long-term action consistency test of the three-supervision platform's motorized components, and provide the input-output consistency test results of the three-supervision platform's motorized mechanism, and the test results are effective and reliable. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 is a flow chart of a three-super platform testing method based on image recognition in an embodiment of the present invention;

[0034] Figure 2 It is a network architecture diagram of a YOLOv5 model in an embodiment of the present invention. DETAILED DESCRIPTION

[0035] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments disclosed in the present invention will be further described in detail below with reference to the accompanying drawings.

[0036] One of the core ideas of the present invention is to provide a three-super platform testing method based on image recognition. In order to accurately identify the state of the three-super platform's motorized components, the motorized components are imaged at high speed to obtain a time series image of the motorized component state. By annotating the data to form a data set, YOLO_V5 is used for transfer learning to identify the plane formed by the motorized rod (the motorized platform plane line). By fusing the different image recognition results for angle analysis, the corresponding state of the motorized components can be obtained. At the same time, based on the obtained angle, displacement, and current, a displacement-current-angle model is established to further determine the consistency of the motorized components.

[0037] Reference Figure 1 In this embodiment, the three-super platform testing method based on image recognition includes:

[0038] Step 1: Build an image recognition model.

[0039] In this embodiment, the construction process of the image recognition model is as follows: the three-super-platform mobile platform installed on the satellite is continuously imaged at multiple angles using a shooting device to obtain a sample image; based on a manual annotation method, the plane lines of the three-super-platform mobile platform in the sample image are annotated to obtain a data set for YOLOv5 model migration; based on the obtained data set for YOLOv5 model migration, the YOLOv5 model is trained, and through parameter optimization, the YOLOv5 model accurately outputs the recognized plane lines of the three-super-platform mobile platform to obtain an image recognition model. In the process of constructing the data set for YOLOv5 model migration, in order to reduce the possibility of overfitting of the training network due to too little image data, the initial data set can be expanded by translating, rotating and mirroring the image.

[0040] like Figure 2 As shown in the figure, the backbone feature extraction network of the YOLOv5 model is Backbone, and its structure is called the CSPD-arknet model. When processing an image, the CSPD-arknet model is first used for feature extraction; the YOLO Head obtains the effective feature layer through the CSPD-arknet model, and judges the feature points to determine whether there is an object corresponding to the feature points. In an image, the convolution kernel slides over each pixel of the image from the very beginning, from top to bottom, from left to right, in units of one pixel.

[0041] Step 2, displacement-current-angle model.

[0042] In this embodiment, the construction process of the displacement-current-angle model is as follows: a static image of the motorized components of the three-ultrasound platform is obtained when the driving current is 0, and the plane line of the motorized platform of the three-ultrasound platform is identified by the image recognition model as the reference plane line, and the corresponding rotation angle is recorded as Φ0; within the driving current range, the driving current is output according to the normal distribution to obtain n groups of current values ​​A1, A2, ..., A n and the corresponding displacement telemetry values ​​L1, L2, ..., L n ; Get the three super platform motor components when the driving current is A i The static image of the case is used to identify the plane line of the three-supervision platform maneuvering platform through the image recognition model, and the coincidence analysis is performed with the reference plane line to calculate the corresponding rotation angle Φ i (The plane line rotation angle in the polar coordinate system is used to indicate whether the three-supervision platform motorized components are moving correctly); obtain n sets of displacement-current-angle data (L1, A1, Φ i ); for the n sets of displacement-current-angle data (L1, A1, Φ i ) is fitted to obtain the displacement-current-angle model: Φ=f(L,A). Where i=1,2,...,n, A i ≠0; f() represents the displacement-current-angle fitting function; the driving current range that can be input to the motorized components of the three-super platform is [A min ,A max ], A min Indicates the minimum driving current value that can be input to the motorized components of the three-super platform, A max Indicates the maximum driving current value that can be input to the motorized components of the three-super platform, A i ∈[A min ,A max ].

[0043] Step 3, obtain the driving current value and displacement telemetry value of the three-ultrasound platform under the current working condition, substitute them into the displacement-current-angle model, and calculate the predicted rotation angle Φ.

[0044] Step 4, obtaining a static image of the three-ultrasound platform to be tested under the current working condition to be tested, and determining the plane line rotation angle Φ′ of the mobile platform of the three-ultrasound platform to be tested through the image recognition model.

[0045] Step 5: Calculate the difference rate I based on Φ and Φ′.

[0046] In this embodiment, the difference rate I is calculated as follows:

[0047]

[0048] Step 6: According to the calculated difference rate I, the consistency of the motorized components of the three-ultrasound platform to be tested is determined.

[0049] In this embodiment, the smaller the difference rate I is, the better the consistency of the mobile components of the three-supervision platform to be tested. The difference rate represents the health status of the mobile components (rods) to a certain extent. When the difference rate is >5%, a ground warning can be given to warn that the mobile components may have poor consistency. If the problem of the on-board product itself is eliminated through ground intervention, the image recognition model and / or the displacement-current-angle model can be parameterized according to the historical difference rate to improve the image recognition model and / or the displacement-current-angle model.

[0050] Although the present invention has been disclosed as above in the form of a preferred embodiment, it is not intended to limit the present invention. Any person skilled in the art may make possible changes and modifications to the technical solution of the present invention by using the methods and technical contents disclosed above without departing from the spirit and scope of the present invention. Therefore, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall fall within the protection scope of the technical solution of the present invention.

[0051] The contents not described in detail in the specification of the present invention belong to the common knowledge of the professionals in this field.

Claims

1. A three-super platform testing method based on image recognition, characterized in that: include: Construct image recognition model and displacement-current-angle model; Obtain the driving current value and displacement telemetry value of the three-ultrasound platform under the current working condition, substitute them into the displacement-current-angle model, and calculate the predicted rotation angle Φ; Obtain a static image of the three-ultrasound platform to be tested under the current working condition to be tested, and determine the plane line rotation angle Φ′ of the mobile platform of the three-ultrasound platform to be tested through the image recognition model; According to Φ and Φ′, the difference rate I is calculated; According to the calculated difference rate I, the consistency of the mobile components of the three-ultrasound platform to be tested is judged; wherein, the smaller the difference rate I is, the better the consistency of the mobile components of the three-ultrasound platform to be tested is.

2. The three-super platform testing method based on image recognition according to claim 1 is characterized in that: The construction process of the image recognition model is as follows: The three-super-platform mobile platform installed on the satellite is imaged continuously at multiple angles using a photographing device to obtain sample images; Based on the manual annotation method, the plane lines of the three super-platform maneuvering platforms in the sample images are annotated to obtain the dataset for YOLOv5 model migration; Based on the obtained data set of YOLOv5 model migration, the YOLOv5 model is trained, and through parameter optimization, the YOLOv5 model accurately outputs the recognized plane lines of the three-super platform maneuvering platform to obtain an image recognition model.

3. The three-super platform testing method based on image recognition according to claim 1 is characterized in that: The construction process of the displacement-current-angle model is as follows: Obtain a static image of the motorized components of the three-ultrasound platform when the driving current is 0, and identify the plane line of the motorized platform of the three-ultrasound platform through the image recognition model as the reference plane line, and the corresponding rotation angle is recorded as Φ0; Within the driving current range, the driving current output is carried out according to the normal distribution to obtain n groups of current values ​​A1, A2, ..., A n and the corresponding displacement telemetry values ​​L1, L2, ..., L n ; Get the three super platform motor parts at a driving current of A i The static image of the case is used to identify the plane line of the three-supervision platform maneuvering platform through the image recognition model, and the coincidence analysis is performed with the reference plane line to calculate the corresponding rotation angle Φ i ;where,i=1,2,...,n,A i ≠0; Get n sets of displacement-current-angle data (L1, A1, Φ i ); For the n sets of displacement-current-angle data (L1, A1, Φ i ) is fitted to obtain the displacement-current-angle model: Φ=f(L,A); where f() represents the displacement-current-angle fitting function.

4. The three-super platform testing method based on image recognition according to claim 3 is characterized in that: The driving current range that can be input to the motorized components of the three-super platform is [A min ,A max ]; among them, A min Indicates the minimum driving current value that can be input to the motorized components of the three-super platform, A max Indicates the maximum driving current value that can be input to the motorized components of the three-super platform, A i ∈[A min ,A max ].

5. The three-super platform testing method based on image recognition according to claim 3 is characterized in that: Rotation angle Φ i , used to indicate whether the movement of the three-supervisory platform's maneuvering parts is correct.

6. The three-super platform testing method based on image recognition according to claim 1 is characterized in that: The calculation formula of difference rate I is as follows:

7. The three-super platform testing method based on image recognition according to claim 1 is characterized in that: Also includes: When I ≥ 5%, a ground warning is given.

8. The three-super platform testing method based on image recognition according to claim 1 is characterized in that: Also includes: The historical difference rate is obtained, and parameters of the image recognition model and / or the displacement-current-angle model are corrected.

Images