A method for estimating the angle of a target object with high angle resolution using a large array antenna

Abstract

The present invention provides a method for estimating the angle of a target object with high angular resolution using a large array antenna. The array antenna receives an input signal matrix, where the input signal matrix is a transmitted signal or a reflected signal of at least one target object. The method comprises: step S1: inputting the input signal matrix into a first computational model to obtain the number of targets and a rough target angle corresponding to the target object's location; step S2: inputting the number of targets and the input signal matrix into a second computational model to perform singular value decomposition and obtain a noise matrix; step S3: obtaining an iterative angle range using the rough target angle obtained in step S1; and step S4: inputting the complex tracking matrix and the noise matrix corresponding to the iterative angle range into a third computational model and iterating the angle range to obtain a precise target angle.

Description

Technical Field

[0001] The present invention relates to an estimation method, and more particularly to a method for estimating the angle of a target object using a large array antenna for high angle resolution. Background Art

[0002] Antenna modules are widely used in today's society. For example, in wireless communications and radar detection, they are required to transmit and receive wireless signals for signal transmission or location detection. In environments with multiple antenna signal sources, issues such as signal interference, optimal signal transmission paths, and selection of the best signal sources are all areas of significant development and resolution in the antenna field.

[0003] For example, mobile phones transmit network and telecommunication signals through antennas connected to base stations. Therefore, how to select the best base station for connection among multiple base stations at varying distances and along varying transmission paths, thereby obtaining optimal telecommunication and network signals, has become a pressing challenge in the mobile communications field. Another example is automotive radar detection systems, which are primarily used to detect objects around the vehicle, identifying obstacles and assisting drivers in avoiding collisions. Furthermore, they can even proactively enable autonomous driving. The detection principle of automotive radar systems is to use radar antennas to measure the distance between objects and the vehicle as it moves. Traditional radar solutions rely on two to three transmitting antennas and three to four receiving antennas. Imaging radar utilizes multiple multiple-input, multiple-output (MIMO) antenna arrays to achieve high-resolution mapping of the surrounding environment, providing high-precision image quality. This eliminates Doppler ambiguity in all environmental conditions, enabling long-range and wide-field decision-making to avoid collisions between vehicles and objects, and prevent traffic accidents.

[0004] In the aforementioned example of optimal signal search technology for mobile phones and base stations, since searching for base station signals is less urgent and dangerous, the target signal detection method used by the mobile phone's radar antenna can use a signal detection method with longer computation time to repeatedly search for the base station's location and determine the base station's signal strength, thereby receiving a better signal. Furthermore, the base station's location is fixed, making it easier to determine its location.

[0005] However, a car's radar antenna is designed to prevent collisions while the vehicle is moving. Therefore, the vehicle radar system must accurately detect the number and location of objects in a short period of time. Therefore, a more accurate and rapid signal calculation method is needed to achieve this goal. Summary of the Invention

[0006] The main purpose of the present invention is to solve the problems of the target signal detection method in the prior art, such as the inability to accurately determine the target position and the excessively long detection calculation time.

[0007] To achieve the above-mentioned objectives, the present invention provides a method for estimating the angle of a target object with high angular resolution using a large array antenna. The method receives an input signal matrix through an array antenna, where the input signal matrix is a transmitted signal or a reflected signal of at least one target object. The method for estimating the angle of a target object with high angular resolution using a large multi-input multi-output array antenna includes the following steps: Step S1: Inputting the input signal matrix into a first computational model and inputting a complex tracking matrix related to angles for comparison. When the complex tracking matrix satisfies the conditions of the first computational model, a number of targets and a rough target angle corresponding to the target location are obtained; Step S2: Inputting the number of targets and the input signal matrix into a second computational model for singular value decomposition and obtaining a noise matrix; Step S3: Obtaining an iterative angle range using the rough target angle obtained in Step S1; Step S4: Inputting the complex tracking matrix and the noise matrix corresponding to the iterative angle range into a third computational model and iterating the angle range to obtain a precise target angle.

[0008] In another embodiment of the present invention, step S1 further includes a step S1a, in which the first computational model compares the input signal matrix with the complex tracking matrix for correlation. If one of the complex tracking matrices is not correlated with the input signal matrix, step S1a is repeated to compare the correlation between the next complex tracking matrix and the input signal matrix. If one of the complex tracking matrices is correlated with the input signal matrix, step S1b is performed: a residual value is calculated based on the correlated tracking matrix and the input signal matrix.

[0009] In another embodiment of the present invention, in step S1b, a tracking matrix correlated with the input signal matrix is defined as a first tracking matrix, and the first tracking matrix and the input signal matrix are used to calculate the residual value.

[0010] In another embodiment of the present invention, step S1 further includes a step S1c, in which the residual value is compared with a residual threshold. If the residual value shows a one-time significant decrease and is lower than or equal to the residual threshold, the first calculation model obtains the number of target objects and roughly estimates the target angle; if the residual value is greater than the residual threshold, a step S1d is performed: the reduction range of the residual value is compared.

[0011] In another embodiment of the present invention, in step S1d, if the reduction in the residual value is greater than a significant threshold, step S1e is performed: the first tracking matrix is stored, and the input signal matrix is converted into a second input signal matrix independent of the first tracking matrix, and the process returns to step S1a to perform a correlation comparison between the second input signal matrix and other complex tracking matrices; if the reduction in the residual value is less than the significant threshold, the process directly returns to step S1a to perform a correlation comparison between the input signal matrix and other complex tracking matrices.

[0012] In another embodiment of the present invention, in step S1a, the angle range of the complex tracking matrix is from -90 degrees to 90 degrees, and the corresponding complex tracking matrix is sequentially correlated with the input signal matrix in units of a first interval angle.

[0013] In another embodiment of the present invention, the first interval angle is between 1 / 180 and 1 / 18 of the angle range of the complex tracking matrix.

[0014] In another embodiment of the present invention, in step S1b, the first calculation model calculates the residual value by using the least squares method.

[0015] In another embodiment of the present invention, in step S2 , the input signal matrix is first converted into a covariance matrix, and then singular value decomposition is performed on the matrix together with the number of target objects.

[0016] In another embodiment of the present invention, in step S3 , the rough estimated target angle is added to and subtracted from an allowable angle to form an iteration angle range.

[0017] In another embodiment of the present invention, the allowable angle is smaller than the first interval angle.

[0018] In another embodiment of the present invention, in step S4 , the iteration angle range is sequentially input into the third operation model in units of a second interval angle, and the angle range is iterated with the noise matrix.

[0019] In another embodiment of the present invention, the second interval angle is between 1 / 100 and 1 / 10 of the iteration angle range.

[0020] In another embodiment of the present invention, the third operation model performs iteration of the angle range and the noise matrix through an orthogonal operation.

[0021] In another embodiment of the present invention, in step S4, a curve of angle and power can be obtained after iteration of the third calculation model, and the angle corresponding to the highest peak on the curve is the precise target angle.

[0022] In another embodiment of the present invention, the input signal matrix is obtained by reflecting signals from a target object from a radar array antenna installed on a vehicle body.

[0023] Thus, the present invention combines the first computing model, the second computing model, and the third computing model to more accurately obtain information related to the number and angle of the target objects, and can increase the computing speed, so that the present invention can be applied to more different types of equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 1 is a flow chart of a method for estimating the angle of a target object using a large array antenna for high-angle resolution according to an embodiment of the present invention;

[0025] Figure 2 is a detailed flow chart of step S1 of an embodiment of the present invention;

[0026] Figure 3 This is a schematic diagram of the implementation state of the first embodiment of the present invention, used to illustrate the situation where there is only one target object;

[0027] Figure 4 is a schematic diagram of an implementation state of the second embodiment of the present invention, used to illustrate a situation where there are more than two targets;

[0028] Figure 5 is a graph of a second embodiment of the present invention, used to represent the positions of the target at an angle of -12 degrees and an angle of 21 degrees;

[0029] Figure 6 is a graph of another embodiment of the present invention, used to represent the positions of the target at an angle of -1 degree and an angle of 1 degree respectively.

[0030] Description of Reference Numerals

[0031] 100: Method for estimating target angles with high angle resolution using large array antennas

[0032] S1~S4: Steps S1a~S1e: Steps

[0033] 1: Car radar 2: Object

[0034] 3a: First object 3b: Second object

[0035] L: Detection baseline. DETAILED DESCRIPTION

[0036] To facilitate the description of the central concept of the present invention as described in the Summary of the Invention, specific embodiments are now presented. Various objects in the embodiments are drawn to a scale suitable for illustration rather than to the scale of the actual components.

[0037] Radar array antennas have a variety of uses, such as object detection, weather observation, and tracking. Radar array antennas used for object detection utilize various algorithms, such as Orthogonal Matching Pursuit (OMP), Multiple Signal Classification (MUSIC), and Iterative Sparse Asymptotic Minimum Variance (SAMV), to perform signal analysis and calculations.

[0038] The Orthogonal Matching Pursuit (OMP) object detection method first compares the correlation between the input signal matrix (the reflected signal of the target object received by the radar array antenna) and the angle-related steering vector. Based on the correlation between the input signal matrix and the steering vector, residual analysis in regression analysis is used to calculate the residual value to determine the number of objects and the angle at which the objects are located. Orthogonal Matching Pursuit (OMP) has the ability to calculate quickly. However, to achieve this fast calculation capability, OMP's detection accuracy is not high. Orthogonal Matching Pursuit (OMP) cannot obtain the precise angle information of the target object, only that the target is located within a certain angle range.

[0039] The object detection method of Multiple Signal Classification (MUSIC) first converts the target's reflected signals received by the radar array antenna into a covariance matrix. Then, based on the number of targets and the covariance matrix, it performs singular value decomposition (SVD) to obtain a noise matrix. Finally, the noise matrix and an angle-dependent steering vector are iterated to obtain the target's angle information. While MUSIC can obtain high-precision angle information, the computational effort required to achieve this accuracy is extremely large, resulting in very slow computational speeds. Furthermore, MUSIC requires the number of targets to be known before it can be calculated, which limits its application.

[0040] See also Figures 1 to 2As shown, a method 100 for estimating the angle of an object using a large array antenna for high-angle resolution is disclosed according to an embodiment of the present invention. Specifically, the method is a method for estimating the angle of an object using a large multi-input multi-output (MIMO) array antenna. An array antenna receives an input signal matrix, wherein the input signal matrix is a transmitted signal or a reflected signal of at least one target object. In an embodiment of the present invention, the array antenna is a multi-input multi-output (MIMO) array antenna, which can independently transmit signals using multiple antennas at the transmitting end and simultaneously receive and recover the original information using multiple antennas at the receiving end. The input signal matrix is obtained by receiving the reflected signals of the target object using a radar array antenna installed on the vehicle body. This input signal matrix is used as a basis for the vehicle's advanced driver assistance systems (ADAS), such as forward collision warning (FCW) and blind spot detection (Blind Spot Detection). Detection, BSD), etc., but not limited to this. Any part that requires the number, position, and angle detection of targets can be performed using the method of the present invention. Another example is the connection determination between a mobile phone and a base station. Due to the connection signal, the mobile phone needs to perform connection detection on multiple base stations to determine the optimal base station for connection. Therefore, the method of the present invention is also applicable. The present invention includes the following steps:

[0041] Step S1: Input the input signal matrix into a first computational model (e.g., Orthogonal Matching Pursuit, OMP) and compare it with a complex angle-dependent tracking matrix. When the complex tracking matrix satisfies the conditions of the first computational model, the number of targets and a rough estimate of the target angle corresponding to the target's location are obtained. The complex tracking matrix is composed of angle-dependent steering vectors, which can be expressed as Equation (1).

[0042] Where v(k) is the steering vector.

[0043] In an embodiment of the present invention, a step S1a, a step S1b, a step S1c, a step S1d, and a step S1e are used to determine whether the complex pursuit matrix satisfies the conditions of the first computational model. The contents of steps S1a, S1b, S1c, S1d, and S1e are described below:

[0044] Step S1a: The first computational model compares the input signal matrix with the complex tracking matrix to determine whether there is any correlation. If one of the complex tracking matrices is not correlated with the input signal matrix, step S1a is repeated to perform a correlation comparison between the next complex tracking matrix and the input signal matrix. If one of the complex tracking matrices is correlated with the input signal matrix, step S1b is performed.

[0045] Furthermore, the complex tracking matrix has an angle range from -90 degrees to 90 degrees, and correlation comparisons are performed sequentially between the corresponding complex tracking matrices and the input signal matrix at a first interval angle. The correlation comparison compares the angular correlations between the complex tracking matrix and the input signal matrix. The value of the first interval angle is between 1 / 180 and 1 / 18 (1 degree to 10 degrees) of the angular range of the complex tracking matrix, and the user can set the first interval angle as needed.

[0046] Step S1b: Define the tracking matrix correlated with the input signal matrix as a first tracking matrix, and calculate a residual value using the first tracking matrix and the input signal matrix. In this embodiment, the first calculation model calculates the residual value using the least squares method.

[0047] Step S1c: Compare the residual value with a residual threshold. If the residual value shows a one-time, significant decrease and is lower than or equal to the residual threshold (indicating that the first tracking matrix satisfies the conditions of the first computational model), the first computational model can obtain the number of targets and the rough estimate of the target angle. If the residual value is greater than the residual threshold, step S1d is performed. If the residual value is greater than the residual threshold, it indicates that there may be multiple targets, or that the first tracking matrix and the input signal matrix have a low correlation.

[0048] Step S1d: Compare the magnitude of the reduction in the residual values. If the magnitude of the reduction in the residual values is greater than a significant threshold, it indicates that the first tracking matrix has a high correlation with the input signal matrix, and the process proceeds to step S1e. If the magnitude of the reduction in the residual values is less than the significant threshold, the process returns directly to step S1a and continues to compare the correlations between the input signal matrix and the other complex tracking matrices until the residual values are less than or equal to the residual threshold, allowing the first computational model to obtain the number of targets and the roughly estimated target angle. If the magnitude of the reduction in the residual values is less than the significant threshold, it indicates that the first tracking matrix has a low correlation with the input signal matrix.

[0049] Step S1e: Store the first tracking matrix and convert the input signal matrix into a second input signal matrix that is independent of the first tracking matrix, and return to step S1a to perform a correlation comparison between the second input signal matrix and other complex tracking matrices until the residual value is lower than or equal to the residual threshold, so that the first computational model can obtain the number of target objects and the roughly estimated target angle.

[0050] Step S2: Convert the input signal matrix into a covariance matrix, and input the covariance matrix and the number of target objects obtained by the first computational model into a second computational model for singular value decomposition to obtain a noise matrix. The input signal matrix is converted into the covariance matrix using Equation (2).

[0051] R=E{yy H}…(2), where R is the covariance matrix; y is the input signal matrix.

[0052] Step S3: obtaining an iterative angle range through the roughly estimated target angle obtained in step S1, wherein the iterative angle range is obtained by adding and subtracting an allowable angle from the roughly estimated target angle, and the allowable angle is smaller than the first interval angle.

[0053] Step S4: Input the complex tracking matrix and the noise matrix corresponding to the iteration angle range into a third computational model (e.g., Multiple Signal Classification, MUSIC) and perform an iteration of the angle range to obtain a precise target angle. The iteration angle range is sequentially input into the third computational model in units of a second interval angle, and the angle range is iterated with the noise matrix.

[0054] Furthermore, the second interval angle is between 1 / 100 and 1 / 10 of the iterative angle range; the third computing model iterates the iterative angle range and the noise matrix through orthogonal operation, and after the iteration of the third computing model, a curve of angle and power can be obtained, and the angle corresponding to the obvious peak on the curve is the precise target angle.

[0055] Please cooperate Figure 3 and Figure 4 As shown, there is a vehicle radar 1, which has a detection reference line L, which is perpendicular to the vehicle radar 1. The upward direction of the detection reference line L is the negative direction, and the downward direction of the detection reference line L is the positive direction. Figure 3 As shown in FIG. 1 , which is a first embodiment of the present invention, it is assumed that there is an object 2, and the detection reference line L is used as a reference, and the angle of the object 2 is -12 degrees; Figure 4 The second embodiment of the present invention is shown in FIG. Assuming a first object 3a and a second object 3b, and based on the detection reference line L, the first object 3a is located at an angle of -12 degrees, and the second object 3b is located at an angle of 21 degrees. The following first describes the actual detection steps of the first embodiment of the present invention:

[0056] Please cooperate Figures 1 to 3 As shown, in step S1, the input signal matrix is input into the first computational model and the complex tracking matrix is input for comparison. When the complex tracking matrix satisfies the conditions of the first computational model, the number of targets and the rough estimated target angle are obtained. The input signal matrix is obtained by receiving the reflected signal of object 2 by the vehicle radar 1; the complex tracking matrix is composed of steering vectors related to the angle, and the range of the complex tracking matrix is from -90 degrees to 90 degrees. Step S1 determines whether the complex tracking matrix satisfies the conditions of the first computational model through steps S1a to S1e.

[0057] In step S1a, the first computational model compares the correlation between the complex tracking matrix and the input signal matrix, sequentially from -90 degrees to 90 degrees, using the first interval angle as a unit. If one of the complex tracking matrices has no correlation with the input signal matrix, step S1a is repeated.

[0058] Taking the first embodiment as an example, the first interval angle is set to 5 degrees. For example, when the first computational model performs a correlation comparison between the tracking matrix at an angle of -40 degrees and the input signal matrix, since the angle of object 2 (-12 degrees) is significantly different from the angle of the tracking matrix (-40 degrees), the first computational model determines that the tracking matrix at an angle of -40 degrees has no correlation with the input signal matrix. Therefore, the first computational model repeats step S1a to sequentially perform a correlation comparison between the complex tracking matrix at the next angle and the input signal matrix.

[0059] Conversely, if one of the complex tracking matrices is correlated with the input signal matrix, step S1b is performed. Taking the first embodiment as an example, when the first computational model compares the correlation between the tracking matrix at an angle of -10 degrees and the input signal matrix, since the angle of object 2 (-12 degrees) is very close to the angle of the tracking matrix (-10 degrees), the first computational model determines that the tracking matrix at an angle of -10 degrees is correlated with the input signal matrix, and the process continues with step S1b.

[0060] In step S1b, the tracking matrix correlated with the input signal matrix is defined as the first tracking matrix, and the first computational model calculates the residual value using the least squares method based on the first tracking matrix and the input signal matrix.

[0061] Taking the first embodiment as an example, the first calculation model defines the tracking matrix with an angle of -10 degrees as the first tracking matrix, and performs the calculation of the residual value, wherein the initial value of the residual value is 100%.

[0062] In step S1c, the residual value is compared with the residual threshold. If the residual value shows a one-time significant decrease and is lower than or equal to the residual threshold, it means that the first tracking matrix meets the conditions of the first computing model, and the first computing model can obtain the number of targets and the roughly estimated target angle. If the residual value is greater than the residual threshold, step S1d is performed.

[0063] Taking the first embodiment as an example, the residual threshold is set to 10%, and the residual value is calculated to be 5%. Since there is only one object 2, and the angle at which it is located (-12 degrees) is very close to the angle of the first tracking matrix (-10 degrees), the residual value will show a one-time significant decrease and fall below the residual threshold (5% < 10%). Therefore, the first computational model can obtain the rough estimate of the target angle as -10 degrees. Furthermore, since the residual value shows only a one-time significant decrease, it is below the residual threshold (5% < 10%), so the first computational model can obtain the number of targets as one.

[0064] Since the first embodiment has obtained the number of target objects (1) and the roughly estimated target angle (-10 degrees) in step S1c, it means that the purpose of step S1 (obtaining the number of target objects and the roughly estimated target angle) has been achieved. Therefore, there is no need to perform steps S1d and S1e, and step S2 can be continued.

[0065] In step S2, the input signal matrix is converted into the covariance matrix, and the covariance matrix and the number of target objects are input into the second computational model for singular value decomposition (SVD) to obtain the noise matrix. Taking the first embodiment as an example, the covariance matrix and the number of target objects (1) are input into the second computational model for singular value decomposition to obtain the noise matrix.

[0066] In step S3, the rough estimated target angle plus and minus the allowable angle is used to determine the iteration angle range, and the allowable angle is smaller than the first interval angle. For example, in the first embodiment, the allowable angle is set to 3 degrees. The rough estimated target angle (-10 degrees) is added to and subtracted from the allowable angle (3 degrees), resulting in an iteration angle range of -7 degrees to -13 degrees. The allowable angle (3 degrees) is smaller than the first interval angle (5 degrees).

[0067] In step S4, the complex tracking matrix and the noise matrix corresponding to the iteration angle range are input into the third computational model. The third computational model then performs an orthogonal operation to iterate the angles of the complex tracking matrix and the noise matrix to obtain the precise target angle. The iteration angle range is based on the second interval angle, and the corresponding complex tracking matrices are sequentially input into the third computational model.

[0068] Taking the first embodiment as an example, the second interval angle is set to 0.1 degrees. The complex tracking matrices corresponding to the iteration angle range (-7 degrees to -13 degrees) are sequentially input into the third computational model (e.g., Multiple Signal Classification, MUSIC) with the noise matrix in units of the second interval angle (0.1 degrees). The third computational model sequentially performs angle iterations on the complex tracking matrices corresponding to the iteration angle range (-7 degrees to -13 degrees) and the noise matrix through orthogonal operations to obtain the precise target angle (-12 degrees).

[0069] Among them, after the third operation model is iterated, the angle and power curve can be obtained. Since the angle of object 2 is -12 degrees, the curve will only have an obvious highest peak at the position of -12 degrees, indicating that -12 degrees is the precise target angle.

[0070] As mentioned above, Figure 3 As shown, in the first embodiment, the vehicle radar 1 can detect an object 2 at an angle of -12 degrees and no other objects at other angles through steps S1 to S4 of the present invention.

[0071] The following continues to describe the actual detection steps in the second embodiment of the present invention:

[0072] Please cooperate Figures 1 to 4As shown, in step S1, the input signal matrix is input into the first computational model and the complex tracking matrix is input for comparison. When the complex tracking matrix satisfies the conditions of the first computational model, the number of targets and the rough estimated target angle are obtained. The input signal matrix is obtained by the vehicle radar 1 receiving reflected signals from the first object 3a and the second object 3b. The complex tracking matrix is composed of steering vectors related to angles, and the range of the complex tracking matrix is from -90 degrees to 90 degrees. Step S1 determines whether the complex tracking matrix satisfies the conditions of the first computational model through steps S1a to S1e.

[0073] In step S1a, the first computational model compares the correlation between the complex tracking matrix and the input signal matrix, sequentially from -90 degrees to 90 degrees, using the first interval angle as a unit. If one of the complex tracking matrices has no correlation with the input signal matrix, step S1a is repeated.

[0074] Taking the second embodiment as an example, the first interval angle is set to 5 degrees. For example, when the first computational model performs a correlation comparison between the tracking matrix at an angle of -40 degrees and the input signal matrix, since the angles of the first object 3a (-12 degrees) and the second object 3b (21 degrees) differ significantly from the angle of the tracking matrix (-40 degrees), the first computational model determines that the tracking matrix at an angle of -40 degrees has no correlation with the input signal matrix. Therefore, the first computational model repeats step S1a to sequentially perform a correlation comparison between the complex tracking matrix at the next angle and the input signal matrix.

[0075] Conversely, if one of the complex tracking matrices is correlated with the input signal matrix, step S1b is performed. Taking the second embodiment as an example, when the first computational model compares the correlation between the tracking matrix at an angle of -15 degrees and the input signal matrix, since the difference between the angle of the first object 3a (-12 degrees) and the angle of the tracking matrix (-15 degrees) is small, the first computational model determines that the tracking matrix at an angle of -15 degrees is correlated with the input signal matrix, and the process continues to step S1b.

[0076] In step S1b, the tracking matrix correlated with the input signal matrix is defined as the first tracking matrix, and the first computational model calculates the residual value using the least squares method based on the first tracking matrix and the input signal matrix.

[0077] Taking the second embodiment as an example, the first calculation model defines the tracking matrix with an angle of -15 degrees as the first tracking matrix, and performs the residual value calculation, wherein the initial value of the residual value is 100%.

[0078] In step S1c, the residual value is compared with the residual threshold. If the residual value shows a one-time significant decrease and is lower than or equal to the residual threshold, it means that the first tracking matrix meets the conditions of the first computing model, and the first computing model can obtain the number of targets and the roughly estimated target angle; if the residual value is greater than the residual threshold, step S1d is performed.

[0079] Taking the second embodiment as an example, the residual threshold is set to 10%, and the residual value is calculated to be 90%. The residual value is still greater than the residual threshold (90%>10%), so the next step (step S1d) must be performed.

[0080] In step S1d, the magnitude of the reduction in the residual value is compared. If the magnitude of the reduction in the residual value is less than the significance threshold, the process returns directly to step S1a and continues to compare the correlations between the input signal matrix and the other complex tracking matrices until the residual value falls below the residual threshold. If the magnitude of the reduction in the residual value is greater than the significance threshold, the process proceeds to step S1e.

[0081] Taking the second embodiment as an example, the significance threshold is set to 35%. The residual value is calculated to be 90%, but the decrease in the residual value (10%) is less than the significance threshold (35%). This indicates that the first tracking matrix (angle of -15 degrees) is correlated with the input signal matrix, but the correlation is not high. Therefore, the first computational model returns to step S1a to perform the next correlation comparison between the tracking matrix (angle of -10 degrees) and the input signal matrix.

[0082] Continuing with the above, the first computational model repeats steps S1a through S1d. In step S1a, because the difference between the angle of the first object 3a (-12 degrees) and the angle of the tracking matrix (-10 degrees) is small, the first computational model determines that the tracking matrix at an angle of -10 degrees is correlated with the input signal matrix. In step S1b, the tracking matrix at an angle of -10 degrees is defined as the first tracking matrix, and the residual value calculation is performed. At this time, the starting value of the residual value is 90%. In step S1c, the calculated residual value is 50%, which is still greater than the residual threshold (50% > 10%), so step S1d must be performed. In step S1d, the decrease in the residual value (from 90% to 50%, a 40% decrease) exceeds the significance threshold (35%), so step S1e is continued. The reduction in the residual value (40%) is greater than the significant threshold, indicating that the first tracking matrix (with an angle of -10 degrees) has a high correlation with the input signal matrix.

[0083] In step S1e, the first tracking matrix is stored, and the input signal matrix is converted into the second input signal matrix that is independent of the first tracking matrix. Then, the process returns to step S1a to perform a correlation comparison between the second input signal matrix and other complex tracking matrices until the residual value is lower than or equal to the residual threshold, so that the first computational model can obtain the number of target objects and the roughly estimated target angle.

[0084] Taking the second embodiment as an example, the first tracking matrix (angle of -10 degrees) is stored, and the input signal matrix is converted into a second input signal matrix (independent of -10 degrees) that is independent of the first tracking matrix (angle of -10 degrees). The process then returns to step S1a to perform a correlation comparison between the next tracking matrix (angle of -5 degrees) and the second input signal matrix until the residual value is less than or equal to the residual threshold (10%). The storage of the first tracking matrix (angle of -10 degrees) indicates that one of the roughly estimated target angles is -10 degrees.

[0085] As described above, in the second embodiment, the first computation model passes through steps S1a to S1e, and finally obtains the number of target objects as 2 (the first object 3a and the second object 3b), and the roughly estimated target angles as -10 degrees and 20 degrees.

[0086] In step S2, the input signal matrix is converted into the covariance matrix, and the covariance matrix and the number of target objects are input into the second computational model for singular value decomposition (SVD) to obtain the noise matrix. Taking the second embodiment as an example, the covariance matrix and the number of target objects (two) are input into the second computational model for singular value decomposition to obtain the noise matrix.

[0087] In step S3, the rough estimated target angle plus and minus the allowable angle is used to determine the iteration angle range, and the allowable angle is smaller than the first interval angle. For example, in the second embodiment, the allowable angle is set to 3 degrees. The rough estimated target angles (-10 degrees and 20 degrees) are added to and subtracted from the allowable angle (3 degrees), resulting in an iteration angle range of -7 degrees to -13 degrees and 17 degrees to 23 degrees. The allowable angle (3 degrees) is smaller than the first interval angle (5 degrees).

[0088] In step S4, the complex tracking matrix and the noise matrix corresponding to the iteration angle range are input into the third computational model. The third computational model then performs an orthogonal operation to iterate the angles of the complex tracking matrix and the noise matrix to obtain the precise target angle. The iteration angle range is based on the second interval angle, and the corresponding complex tracking matrices are sequentially input into the third computational model.

[0089] Taking the second embodiment as an example, the second interval angle is set to 0.1 degrees. The complex tracking matrices corresponding to the iteration angle ranges (-7 degrees to -13 degrees and 17 degrees to 23 degrees) are sequentially input into the third computational model with the noise matrix in units of the second interval angle (0.1 degrees). The third computational model then sequentially iterates the angles of the complex tracking matrices corresponding to the iteration angle ranges (-7 degrees to -13 degrees and 17 degrees to 23 degrees) and the noise matrix through an orthogonal operation to obtain the precise target angles (-12 degrees and 21 degrees).

[0090] Among them, see Figure 5 After iterating the third calculation model, the angle and power curve can be obtained. Since the angle of the first object 3a is -12 degrees and the angle of the second object 3b is 21 degrees, the curve will only have obvious peaks at the positions of -12 degrees and 21 degrees, indicating that -12 degrees and 21 degrees are the precise target angles.

[0091] As mentioned above, Figure 4As shown, in the second embodiment, the vehicle radar 1 can detect a first object 3a at an angle of -12 degrees and a second object 3b at an angle of 21 degrees through steps S1 to S4 of the present invention, and there are no other objects at other angles.

[0092] For further information, see Figure 6 In other embodiments of the present invention, after the calculations of steps S1 to S4 of the present invention, the curve has obvious peaks at the positions of -1 degree and 1 degree, indicating that there is a target at each of the positions of -1 degree and 1 degree. Figure 6 As shown, it represents that using the method of the present invention, even if the angles between the targets are very close (angles of -1 degree and 1 degree), the curve can still accurately present obvious highest peaks at the corresponding angle positions (angles of -1 degree and 1 degree). Therefore, the user can directly know from the curve that there is a target at the position of angles of -1 degree and 1 degree respectively.

[0093] Thus, the present invention has the following advantages:

[0094] 1. The present invention uses the first computational model to perform a preliminary, larger-scale computation to eliminate most angles irrelevant to the input signal matrix, thereby obtaining the number of target objects and the roughly estimated target angles. This further improves the computational speed of the present invention.

[0095] 2. The present invention utilizes the combination of the second and third computing modules to further decompose the rough target angle and, through angle iteration, obtain the precise target angle. This allows the present invention to more accurately determine the target's position.

[0096] 3. The present invention utilizes singular value decomposition (SVD) to calculate the noise matrix and employs an orthogonal operation to perform angle iteration, so that the curve obtained by the third operation module will have a clear peak only at the precise target angle, thereby allowing the user to more clearly understand the angle information of the target object.

[0097] Although the present invention is described based on a preferred embodiment, those skilled in the art will be able to make various modifications without departing from the spirit and scope of the invention. The above embodiments are intended only to illustrate the present invention and are not intended to limit the scope of the invention. All modifications and variations that do not violate the spirit of the invention are intended to be within the scope of the present invention.

Claims

1. A method for estimating the angle of an object with high angular resolution using a large array antenna, wherein an array antenna receives an input signal matrix, the input signal matrix being a transmitted signal or a reflected signal of at least one object, and wherein: The method comprises the following steps: Step S1: Inputting the input signal matrix into a first computational model and inputting a plurality of tracking matrices related to angles for comparison. When the plurality of tracking matrices satisfy the conditions of the first computational model, a number of targets and a rough estimate of target angles corresponding to the locations of the targets are obtained. Step S2: Inputting the number of target objects and the input signal matrix into a second computation model for singular value decomposition to obtain a noise matrix; Step S3: obtaining an iterative angle range using the roughly estimated target angle obtained in step S1; and Step S4: Input the complex tracking matrix and the noise matrix corresponding to the iterative angle range into a third computation model and perform iteration of the angle range to obtain a precise target angle.

2. The method for estimating the angle of an object using a large array antenna with high angle resolution according to claim 1, wherein: In step S1, a step S1a is also included, in which the first computational model compares the input signal matrix with the complex tracking matrix for correlation. If one of the complex tracking matrices has no correlation with the input signal matrix, step S1a is repeated to perform a correlation comparison between the next complex tracking matrix and the input signal matrix. If one of the complex tracking matrices has a correlation with the input signal matrix, step S1b is performed: a residual value is calculated based on the correlated tracking matrix and the input signal matrix.

3. The method for estimating the angle of an object using a large array antenna for high angle resolution according to claim 2, wherein: In step S1b, the tracking matrix correlated with the input signal matrix is defined as a first tracking matrix, and the residual value is calculated using the first tracking matrix and the input signal matrix.

4. The method for estimating the angle of an object using a large array antenna for high angle resolution according to claim 3, wherein: In step S1, a step S1c is also included, in which the residual value is compared with a residual threshold. If the residual value shows a one-time significant decrease and is lower than or equal to the residual threshold, the first calculation model obtains the number of target objects and the roughly estimated target angle; if the residual value is greater than the residual threshold, a step S1d is performed: the reduction range of the residual value is compared.

5. The method for estimating the angle of an object using a large array antenna for high angle resolution according to claim 4, wherein: In step S1d, if the reduction in the residual value is greater than a significant threshold, step S1e is performed: the first tracking matrix is stored, and the input signal matrix is converted into a second input signal matrix independent of the first tracking matrix, and the process returns to step S1a to perform a correlation comparison between the second input signal matrix and the other complex tracking matrices; if the reduction in the residual value is less than the significant threshold, the process directly returns to step S1a to perform a correlation comparison between the input signal matrix and the other complex tracking matrices.

6. The method for estimating the angle of an object using a large array antenna for high angle resolution according to claim 2, wherein: In step S1a, the angle range of the complex tracking matrix is from -90 degrees to 90 degrees, and the corresponding complex tracking matrix is sequentially correlated with the input signal matrix in units of a first interval angle.

7. The method for estimating the angle of an object using a large array antenna for high angle resolution according to claim 6, wherein: The first interval angle is between 1 / 180 and 1 / 18 of the angle range of the complex tracking matrix.

8. The method for estimating the angle of an object using a large array antenna with high angle resolution according to claim 3, wherein: In step S1b, the first calculation model calculates the residual value using the least squares method.

9. The method for estimating the angle of an object using a large array antenna with high angle resolution according to claim 1, wherein: In step S2 , the input signal matrix is first converted into a covariance matrix, and then singular value decomposition is performed on the matrix together with the target object number.

10. The method for estimating the angle of a target object using a large array antenna with high angle resolution according to claim 6, wherein: In step S3 , the rough estimated target angle is added to and subtracted from an allowable angle to form the iteration angle range.

11. The method for estimating the angle of an object using a large array antenna with high angle resolution according to claim 10, wherein: The allowable angle is smaller than the first interval angle.

12. The method for estimating the angle of an object using a large array antenna with high angle resolution according to claim 10, wherein: In step S4 , the iteration angle range is sequentially input into the third operation model in units of a second interval angle, and the angle range is iterated with the noise matrix.

13. The method for estimating the angle of an object using a large array antenna with high angle resolution according to claim 12, wherein: The second interval angle is between 1 / 100 and 1 / 10 of the iteration angle range.

14. The method for estimating the angle of an object using a large array antenna with high angle resolution according to claim 1, wherein: In step S4 , the third computation model performs iteration on the iteration angle range and the noise matrix through an orthogonal computation.

15. The method for estimating the angle of a target object using a large array antenna with high angle resolution according to claim 1, wherein: In step S4 , a curve of angle and power is obtained after iteration of the third calculation model, and the angle corresponding to the highest peak on the curve is the precise target angle.

16. The method for estimating the angle of an object with high angle resolution using a large array antenna according to claim 1, wherein: The input signal matrix is obtained by the reflected signal of the radar array antenna installed on the vehicle body passing through the target object.

Images