Angle calculation method of AoA positioning algorithm based on convolutional neural network

Abstract

The present invention discloses an angle calculation method of an AoA positioning algorithm based on a convolutional neural network. The method utilizes the horizontal angle and elevation angle of a positioning object to form a two-dimensional array and combines the similarity of graphics. The convolutional neural network is used to perform positioning calculations, which can improve the accuracy of the results and solve the problem that traditional mathematical methods cannot improve the accuracy through learning.

Description

Technical Field

[0001] The present invention relates to the technical field of AoA positioning algorithms, and in particular to an angle calculation method of an AoA positioning algorithm based on a convolutional neural network. Background Art

[0002] In the AoA positioning algorithm, after obtaining a certain number of IQ values, these IQ values ​​can be used to obtain a large number of horizontal angles and elevation angles, and positioning is performed through the horizontal angles and elevation angles. To find the most suitable angle from a large number of elevation angles and horizontal angles, traditional mathematical methods can be used to calculate and obtain the angle value with the highest probability. Summary of the invention

[0003] According to an embodiment of the present invention, a method for calculating an angle of an AoA positioning algorithm based on a convolutional neural network is provided, comprising the following steps:

[0004] Collect a preset number of IQ values ​​in the AoA positioning algorithm to obtain a preset number of horizontal angle and elevation angle values;

[0005] In AoA positioning, set the maximum positioning range of the base station's horizontal angle and vertical angle;

[0006] When the positioning object is always under the base station, the size of the positioning range forms a graph;

[0007] The values ​​of the preset number of horizontal angles and elevation angles are combined into a two-dimensional array, and the base station calculates the position coordinates of the positioning object. The value of the position coordinates is a possibility array, and the possibility array represents the probability of each position coordinate appearing in the two-dimensional array;

[0008] The size of the probability is represented by different colors in the graph;

[0009] Collect a preset amount of test data, build a training model, put the test data into the training model for training, and obtain a trained model.

[0010] Furthermore, the maximum positioning range of the base station is: horizontal angle 0 to 360 degrees, and pitch angle 0 to 90 degrees.

[0011] Furthermore, the figure is hemispherical.

[0012] Furthermore, the hemispherical shape represents a horizontal angle of 0 to 360 degrees and a pitch angle of 0 to 90 degrees.

[0013] According to the angle calculation method of the AoA positioning algorithm based on a convolutional neural network in an embodiment of the present invention, a convolutional neural network is a traditional method used by deep learning to process graphic data. By utilizing the similarity between the two-dimensional array of positioning angles and graphics, positioning calculations are performed through a convolutional neural network, which can improve the accuracy of the results and solve the problem that traditional mathematical methods cannot improve accuracy through learning.

[0014] It is to be understood that both the foregoing general description and the following detailed description are exemplary, and are intended to provide further explanation of the technology as claimed. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 The present invention is a flowchart of an angle calculation method of an AoA positioning algorithm based on a convolutional neural network according to an embodiment of the present invention.

[0016] Figure 2 This is a diagram showing the positioning range of the angle calculation method of the AoA positioning algorithm based on a convolutional neural network according to an embodiment of the present invention.

[0017] Figure 3 This is a hemispherical diagram of the positioning range of the angle calculation method of the AoA positioning algorithm based on a convolutional neural network according to an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The preferred embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings to further illustrate the present invention.

[0019] First, combine Figures 1 to 3 The angle calculation method of the AoA positioning algorithm based on a convolutional neural network according to an embodiment of the present invention is described, and the application scenarios for improving the accuracy of angle calculation are very wide.

[0020] like Figures 1 to 3 As shown, the angle calculation method of the AoA positioning algorithm based on the convolutional neural network in the embodiment of the present invention is

[0021] like Figures 1-2 As shown,

[0022] S1: Collect a preset number of IQ values ​​in the AoA positioning algorithm to obtain a preset number of horizontal angle and elevation angle values.

[0023] like Figures 1-2 As shown,

[0024] S2: In AoA positioning, set the maximum positioning range of the base station's horizontal angle and vertical angle.

[0025] The maximum positioning range of the base station is: horizontal angle 0 to 360 degrees, and pitch angle 0 to 90 degrees (actually less than 90 degrees).

[0026] like Figures 1-2 As shown,

[0027] S3: When the positioning object is always under the base station, the size of the positioning range forms a graph.

[0028] The graph is hemispherical, which represents a horizontal angle of 0 to 360 degrees and a pitch angle of 0 to 90 degrees.

[0029] like Figures 1-2 As shown,

[0030] S4: A preset number of horizontal angles and elevation angles are combined into a two-dimensional array. The base station calculates the position coordinates of the positioning object. The value of the position coordinates is a possibility array. The possibility array represents the probability of each position coordinate appearing in the two-dimensional array.

[0031] All the values ​​of the horizontal angle and elevation angle are organized into a two-dimensional array, with one degree as one unit, as follows:

[0032] <![CDATA[ 0 ]]> <![CDATA[ 1 ]]> <![CDATA[ 2 ]]> <![CDATA[ 3 ]]> … <![CDATA[ 359 ]]> <![CDATA[ 0 ]]> <![CDATA[ 1 ]]> <![CDATA[ … ]]> <![CDATA[ 89 ]]>

[0033] like Figures 1 to 3 As shown,

[0034] S5: Indicate the size of the probability by using different colors in the graph.

[0035] like Figure 3 The red A and blue B represent different probabilities. The goal is to find the best point from the above graph. This best point is not necessarily the darkest point, because if the surrounding points of the darkest point are not dark in color, and the surrounding points of a certain second-darkest point are all very dark in color, the best point is likely to be this second-darkest point.

[0036] like Figures 1 to 3 As shown,

[0037] S6: Collect a preset amount of test data, establish a training model, put the test data into the training model for training, and obtain a trained model.

[0038] By collecting a large amount of test data, each set of data has a two-dimensional array of position probabilities and real position data, that is, each set of data is a picture, and the position of the point to be found is also marked in the picture. This is a typical CNN (convolutional neural network) problem, finding the required point on the picture.

[0039] Use TensorFlow2.0 to build the model and conduct training.

[0040] Use Google's xception pre-trained model,

[0041] xception=tf.keras.applications.Xception(weights='imagenet',

[0042] include_top=False,input_shape=(90,360,3))

[0043] inputs=tf.keras.layers.Input(shape=(90,360,3))

[0044] x=xception(inputs)#Use the existing xception model

[0045] x = tf.keras.layers.GlobalAveragePooling2D()(x) # Convert to a two-dimensional vector

[0046] x = tf.keras.layers.Dense(2048, activation = 'relu')(x) #Fully connected layer

[0047] x = tf.keras.layers.Dense(256, activation = 'relu')(x) #Fully connected layer

[0048] out1 = tf.keras.layers.Dense(1)(x) #The first output is the x coordinate

[0049] out2 = tf.keras.layers.Dense(1)(x) # The second output is the y coordinate

[0050] predictions = [out1, out2]

[0051] model=tf.keras.models.Model(inputs=inputs, outputs=predictions)

[0052] Put the test data into this model for training, and the trained model can be used to calculate the real x and y coordinate values. The accuracy of the calculation can be improved through continuous training using convolutional neural networks.

[0053] Above, refer to Figures 1 to 3The present invention describes an angle calculation method for an AoA positioning algorithm based on a convolutional neural network according to an embodiment of the present invention. A convolutional neural network is a traditional method used by deep learning to process graphic data. By utilizing the similarity between a two-dimensional array of positioning angles and graphics and performing positioning calculations through a convolutional neural network, the accuracy of the results can be improved, solving the problem that traditional mathematical methods cannot improve accuracy through learning.

[0054] It should be noted that, in this specification, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprises..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0055] Although the content of the present invention has been described in detail through the above preferred embodiments, it should be appreciated that the above description should not be considered as a limitation of the present invention. After reading the above content, it will be apparent to those skilled in the art that various modifications and substitutions of the present invention will occur. Therefore, the protection scope of the present invention should be limited by the appended claims.

Claims

1. An angle calculation method for AoA positioning algorithm based on convolutional neural network, It is characterized in that The following steps are involved: Collect a preset number of IQ values ​​in the AoA positioning algorithm to obtain a preset number of horizontal angle and elevation angle values; In AoA positioning, set the maximum positioning range of the base station's horizontal angle and vertical angle; When the positioning object is always below the base station, the size of the positioning range forms a graph; The values ​​of the preset number of horizontal angles and the elevation angles are combined into a two-dimensional array, and the base station calculates the position coordinates of the positioning object, wherein the values ​​of the position coordinates are a possibility array, and the possibility array represents the probability of each position coordinate appearing in the two-dimensional array; The size of the probability is represented by different colors in the graph; Collect a preset amount of test data, build a training model, put the test data into the training model for training, and obtain a trained model.

2. The angle calculation method of the AoA positioning algorithm based on a convolutional neural network as claimed in claim 1, It is characterized in that The maximum positioning range of the base station is: horizontal angle 0~360 degrees, and pitch angle 0~90 degrees.

3. The angle calculation method of the AoA positioning algorithm based on a convolutional neural network as claimed in claim 1, It is characterized in that The figure is hemispherical.

4. The angle calculation method of the AoA positioning algorithm based on a convolutional neural network as claimed in claim 3, It is characterized in that The hemisphere represents a horizontal angle of 0 to 360 degrees and a pitch angle of 0 to 90 degrees.

Images