Blur reduction for public field of view based on radar systems

Abstract

A method includes obtaining an initial point cloud for each of two or more radar systems that share a common field of view; each initial point cloud is generated by processing reflected energy at each of the two or more radar systems; each point of the initial point cloud indicates one or more hypotheses for range, Doppler, and direction of arrival (DOA) of an object that caused the reflected energy; a point cloud derived from the initial point cloud has the same number of hypotheses for range, Doppler, and direction of arrival; resolving ambiguity in the common field of view is based on the point cloud to obtain resolved points and unresolved points in the common field of view; and radar images obtained from each of the two or more radar systems are used to control an aspect of vehicle operation.

Description

Technical Field

[0001] The subject disclosure relates to blur mitigation for public field of view based radar systems. Background Art

[0002] Vehicles (e.g., cars, trucks, construction equipment, automated factory equipment) use sensors to perform semi-autonomous or autonomous operations. Exemplary sensors (e.g., cameras, radar systems, lidar systems, inertial measurement units, accelerometers) provide information about the vehicle and its surroundings. Exemplary semi-autonomous operations include adaptive cruise control (ACC) and collision avoidance. Radar systems typically transmit radio frequency (RF) signals and receive energy reflected from the transmitted signal encountering one or more objects. Processing the reflected energy provides a point cloud, each point associated with a range, a direction of arrival (e.g., an azimuth and elevation angle relative to an object), and a Doppler (rate of change of range). Instead of indicating a single range, direction of arrival, and Doppler for an object, each point can be associated with two or more hypotheses of range values, two or more hypotheses of direction of arrival values, and two or more hypotheses of Doppler values. In order to perform detection using a point cloud, the ambiguity created by the assumptions must first be resolved. Therefore, it is desirable to provide ambiguity mitigation based on a common field of view of the radar system. Summary of the Invention

[0003] In one exemplary embodiment, a method includes obtaining an initial point cloud for each of two or more radar systems sharing a common field of view. Each initial point cloud is generated by processing reflected energy at each of the two or more radar systems, and each point of the initial point cloud indicates one or more hypotheses about the range, Doppler, and direction of arrival (DOA) of the object causing the reflected energy. The method also includes obtaining a point cloud from the initial point cloud for each of the two or more radar systems. Each point of the point cloud for each of the two or more radar systems has the same number of hypotheses about range as other points in the point cloud, the same number of hypotheses about Doppler as other points in the point cloud, and the same number of hypotheses about direction of arrival as other points in the point cloud. Ambiguities in the common field of view are resolved based on the point cloud for each of the two or more radar systems to obtain resolved points and unresolved points in the common field of view. The resolved points indicate a range value, a Doppler value, and a direction of arrival value. A radar image is obtained from each of the two or more radar systems based on the resolved and unresolved points in the common field of view, wherein the radar image is used to control an aspect of vehicle operation.

[0004] In addition to one or more features described herein, obtaining a point cloud from the initial point cloud includes, for each initial point cloud, determining a plurality of hypotheses for range, a plurality of hypotheses for Doppler, and a plurality of hypotheses for direction of arrival for each point of each initial point cloud, outputting the initial point cloud as a point cloud based on a number of hypotheses, the number of hypotheses for range for each point being a first number, the number of hypotheses for Doppler for each point being a second number, and the number of hypotheses for direction of arrival for each point being a third number, and processing the initial point cloud to obtain the point cloud based on the number of hypotheses, the number of hypotheses for range for at least one point being less than the first number, the number of hypotheses for Doppler for at least one point being less than the second number, or the number of hypotheses for direction of arrival for at least one point being less than the third number.

[0005] In addition to one or more features described herein, processing the initial point cloud includes generating a subset of the initial point cloud to include each point having fewer than a first number of hypotheses for range, a second number of hypotheses for Doppler, and a third number of hypotheses for direction of arrival, and using the ambiguity function to obtain a full subset of the initial point cloud that includes the first number of hypotheses for range, the second number of hypotheses for Doppler, and the third number of hypotheses for direction of arrival for each point in the subset of the initial point cloud.

[0006] In addition to one or more features described herein, processing the initial point cloud includes obtaining a Mahalanobis distance between each point in the subset of the initial point cloud and each hypothesis in the full subset of the initial point cloud, and identifying each point in the subset of the initial point cloud as a hypothesis in the full subset of the initial point cloud based on the Mahalanobis distance.

[0007] In addition to one or more features described herein, processing the initial point cloud includes discarding each point in the subset of the initial point cloud that is a hypothesis in the complete subset of the initial point cloud from the complete subset of the initial point cloud and retaining a remainder of the complete subset of the initial point cloud to generate the point cloud.

[0008] In addition to one or more features described herein, the method also includes identifying each point in the point cloud of each of the two or more radar systems that is in the common field of view.

[0009] In addition to one or more features described herein, resolving ambiguity in the common field of view includes obtaining Mahalanobis distances between each set of hypotheses for each point in the common field of view of one of the two or more radar systems and each set of hypotheses for each point in the common field of view of the other of the two or more radar systems, using points in the common field of view of one of the two or more radar systems and the other of the two or more radar systems one at a time, the set of hypotheses including a combination of one of the one or more hypotheses for range, one of the one or more hypotheses for Doppler, and one of the one or more hypotheses for direction of arrival.

[0010] In addition to one or more features described herein, resolving ambiguity in the common field of view includes identifying a point pair that results in a lowest Mahalanobis distance, and based on the point pair passing a gating condition, retaining only a set of hypotheses for the point pair associated with the lowest Mahalanobis distance as unambiguous.

[0011] In addition to one or more features described herein, checking the gating conditions for a point pair includes using each of range, Doppler, and direction of arrival of a set of hypotheses for the point pair.

[0012] In addition to one or more features described herein, one or more objects are identified based on radar images of one or more of the two or more radar systems.

[0013] In another exemplary embodiment, a vehicle includes two or more radar systems. The vehicle also includes a controller configured to obtain an initial point cloud for each of the two or more radar systems that share a common field of view. Each initial point cloud is generated by processing reflected energy at each of the two or more radar systems, and each point in the initial point cloud indicates one or more hypotheses about the range, Doppler, and direction of arrival (DOA) of the object that caused the reflected energy. The controller obtains a point cloud from the initial point cloud for each of the two or more radar systems. Each point in the point cloud for each of the two or more radar systems has the same number of hypotheses about range, the same number of hypotheses about Doppler, and the same number of hypotheses about direction of arrival as other points in the point cloud. The controller resolves ambiguity in the common field of view based on the point cloud for each of the two or more radar systems to obtain resolved points and unresolved points in the common field of view. The resolved points indicate a range value, a Doppler value, and a direction of arrival value, and a radar image is obtained from each of the two or more radar systems based on the resolved and unresolved points in the common field of view. The radar image is used to control one aspect of vehicle operation.

[0014] In addition to one or more features described herein, the controller obtains the point cloud by, for each initial point cloud, determining a plurality of hypotheses for range, a plurality of hypotheses for Doppler, and a plurality of hypotheses for direction of arrival for each point of each initial point cloud, outputting the initial point cloud as a point cloud based on a number of hypotheses, the number of hypotheses for range for each point being a first number, the number of hypotheses for Doppler for each point being a second number, and the number of hypotheses for direction of arrival for each point being a third number, and processing the initial point cloud to obtain the point cloud based on the number of hypotheses, the number of hypotheses for range for at least one point being less than the first number, the number of hypotheses for Doppler for at least one point being less than the second number, or the number of hypotheses for direction of arrival for at least one point being less than the third number.

[0015] In addition to one or more features described herein, the controller processes the initial point cloud by generating a subset of the initial point cloud to include each point having fewer than a first number of hypotheses for range, a second number of hypotheses for Doppler, and a third number of hypotheses for direction of arrival, and using an ambiguity function to obtain a complete subset of the initial point cloud that includes the first number of hypotheses for range, the second number of hypotheses for Doppler, and the third number of hypotheses for direction of arrival for each point in the subset of the initial point cloud.

[0016] In addition to one or more features described herein, the controller processes the initial point cloud by obtaining a Mahalanobis distance between each point in the subset of the initial point cloud and each hypothesis in the full subset of the initial point cloud, and identifying each point in the subset of the initial point cloud that is a hypothesis in the full subset of the initial point cloud based on the Mahalanobis distance.

[0017] In addition to one or more features described herein, the controller processes the initial point cloud by discarding each point in the subset of the initial point cloud that is a hypothesis in the complete subset of the initial point cloud from the complete subset of the initial point cloud, and retaining a remainder of the complete subset of the initial point cloud to generate a point cloud.

[0018] In addition to one or more features described herein, the controller identifies each point in the point cloud of each of the two or more radar systems that is in a common field of view.

[0019] In addition to one or more features described herein, the controller resolves ambiguity in the common field of view of one of the two or more radar systems by obtaining a Mahalanobis distance between each set of hypotheses for each point in the common field of view of one of the two or more radar systems and each set of hypotheses for each point in the common field of view of another of the two or more radar systems using points in the common field of view of one of the two or more radar systems and another of the two or more radar systems at a time, the set of hypotheses comprising a combination of one of the one or more hypotheses for range, one of the one or more hypotheses for Doppler, and one of the one or more hypotheses for direction of arrival.

[0020] In addition to one or more features described herein, the controller resolves ambiguity in the common field of view by identifying a point pair that results in a lowest Mahalanobis distance, and based on the point pair passing a gating condition, retaining only a set of hypotheses associated with the point pair with the lowest Mahalanobis distance as unambiguous.

[0021] In addition to one or more features described herein, the controller checks gating conditions for the point pair including each of range, Doppler, and direction of arrival using a set of hypotheses for the point pair.

[0022] In addition to one or more features described herein, the controller identifies one or more objects based on radar images of one or more of the two or more radar systems.

[0023] The above features and advantages and other features and advantages of the present disclosure will become apparent from the following detailed description when taken in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Additional features, advantages, and details appear by way of example only in the following detailed description, which refers to the accompanying drawings, in which:

[0025] Figure 1 is a block diagram of a vehicle including common field of view ambiguity mitigation based on a radar system;

[0026] Figure 2 is a process flow for a method of performing blur mitigation based on a common field of view of a radar system according to one or more embodiments;

[0027] Figure 3 The following diagram shows the radar system for each radar system according to one or more embodiments. Figure 2 The processes performed as part of the process flow shown in the figure; and

[0028] Figure 4 FIG. 1 shows a schematic diagram of a radar system for two or more radar systems according to one or more embodiments. Figure 2 A process that is performed as part of another part of the process flow shown in . DETAILED DESCRIPTION

[0029] The following description is merely exemplary in nature and is in no way intended to limit the present disclosure, its application, or uses. It should be understood that throughout the drawings, corresponding reference numerals indicate like or corresponding parts and features.

[0030] As previously mentioned, processing the reflected energy received by the radar system helps to obtain the range, direction of arrival, and Doppler of the object that caused the reflection. It is also noted that the point cloud obtained by processing the radar data can include ambiguity, so that each point in the point cloud includes two or more hypotheses for range, direction of arrival, and Doppler. The assumptions in one or more domains (i.e., range, Doppler, direction of arrival) may be the result of waveform parameters, transmission mode, antenna radiation pattern, grating lobes, or increased sidelobe levels. Therefore, each point in the point cloud obtained by the same radar system is expected to have the same number of hypotheses in each domain. For example, for a given radar system, each point in the point cloud can be expected to have two range hypotheses, three Doppler hypotheses, and one direction of arrival hypothesis (i.e., no ambiguity). Embodiments of the systems and methods detailed herein relate to ambiguity mitigation based on a common field of view of radar systems. When the fields of view of two or more radar systems overlap, the ambiguity associated with the points in the point cloud corresponding to the overlapping area (i.e., the common field of view) can be resolved by selecting the best hypothesis.

[0031] According to an exemplary embodiment, Figure 1 is a block diagram of a vehicle 100 that includes blur mitigation based on a common field of view 125 of radar systems 110a, 110b. Figure 1 The exemplary vehicle 100 shown is an automobile 101. Three radar systems 110a, 110b, 110c (generally referred to as 110) are shown with corresponding fields of view 115a, 115b, 115c (generally referred to as 115). As shown, the field of view (FOV) 115a corresponding to radar system 110a and the FOV 115b corresponding to radar system 110b share a common area indicated as common FOV 125. The FOV 115c of radar system 110c does not have any common portions with the fields of view 115a, 115b of radar systems 110a, 110b. In addition to the radar systems 110, the vehicle 100 may include additional sensors 130 (e.g., cameras, lidar systems). The number of radar systems 110, their locations around the vehicle 100, the number of radar systems 110 whose FOV 115 is part of one or more common FOVs 125, and the number and location of the additional sensors 130 are not intended to be limiting. Figure 1 The limitations of the exemplary illustrations in .

[0032] The vehicle 100 includes a controller 120 that can obtain information from the radar system 110 and the additional sensors 130 to control operational aspects of the vehicle 100. The controller 120, alone or in combination with processing circuitry of each radar system 110, can also perform ambiguity reduction according to one or more embodiments detailed herein. The controller 120 includes processing circuitry that can include application-specific integrated circuits (ASICs), electronic circuitry, processors (shared, dedicated, or grouped) and memory that executes one or more software or firmware programs, combinational logic circuitry, and / or other suitable components that provide the described functionality.

[0033] Figure 2 1 is a process flow for a method 200 for performing blur reduction based on a common field of view 125 of radar systems 110, according to one or more embodiments. Method 200 may be performed solely by controller 120 of vehicle 100, or in conjunction with processing circuitry as part of radar system 110. At block 210, the process of block 215 is performed at each radar system 110 whose FOV 115 overlaps with a common FOV 125. Generally, the process at block 215 is performed for each radar system 110, regardless of any overlap in its FOV 115, but only radar systems 110 associated with one or more common FOVs 125 are relevant to the discussion of one or more embodiments. A given radar system 110 may have a FOV 115 that overlaps with the fields of view 115 of two or more other radar systems 110. Furthermore, radar systems 110 of a given vehicle 100 may have multiple common FOVs 125. The processes performed at blocks 210 and 220 apply to each radar system 110 whose FOV 115 overlaps the FOV 115 of another radar system 110 (i.e., each radar system 110 associated with at least one common FOV 125). As described in further detail, the process at block 230 is performed with one common FOV 125 in mind at a time.

[0034] At block 215, for each radar system 110 whose FOV 115 is part of the common FOV 125, the process includes obtaining and processing reflected energy. These processes are well known, and a brief overview of an exemplary embodiment for processing reflected energy is provided herein. As previously described, RF energy is transmitted at each radar system 110 and reflected energy is received. Typically, two Fast Fourier Transforms (FFTs) are performed on the received reflected energy. The first Fast Fourier Transform is along the range, and the second Fast Fourier Transform is along the Doppler corresponding to radial velocity. More specifically, a first (range) Fast Fourier Transform is performed on each transmitted signal to implement range matched filter processing. A second (Doppler) Fast Fourier Transform is performed on each range bin of the first Fast Fourier Transform results for all simultaneously transmitted signals to implement Doppler matched filter processing.

[0035] After the fast Fourier transform, a range-Doppler map is obtained. In the case of a multiple-input multiple-output (MIMO) radar system 110 having multiple transmitting elements and multiple receiving elements, a range-Doppler map is obtained for each combination of transmitting elements and receiving elements. Each range-Doppler map indicates a set of range bins, a set of Doppler bins, and the intensity associated with each range bin and Doppler bin combination. A beamforming process is performed on the range-Doppler map estimate for each bin direction (i.e., obtaining the direction of arrival (DOA)). After beamforming, a range-Doppler beammap is obtained. A detection process is performed on the range-Doppler beammap to obtain a detection associated with the object. A detection is a point in the initial point cloud (IPC) and has one or more corresponding hypotheses for range, Doppler, and direction of arrival associated with it.

[0036] At block 210, after obtaining an initial point cloud IPC for each radar system 110 according to block 215, the process of block 225 is then performed separately for each radar system 110 at block 220. At block 225, the process includes ensuring that all points of the initial point cloud IPC have all expected hypotheses. For a given radar system 110, all points of the initial point cloud IPC should generally have the same number of hypotheses. The number of hypotheses expected for each point in the initial point cloud may generally be different for each radar system 110 (e.g., six hypotheses for each point in IPC1 and four hypotheses for each point in IPC2). For example, each point in the initial point cloud IPC1 may be expected to include two range hypotheses (R1, R2), one Doppler hypothesis (D1), and three direction-of-arrival hypotheses (DOA1, DOA2, DOA3). Therefore, each point in the initial point cloud IPC1 is expected to be associated with the six combinations shown in Table 1 (i.e., six hypotheses (H1 through H6)).

[0037] H1 R1-D1-DOA1 H2 R1-D1-DOA2 H3 R1-D1-DOA3 H4 R2-D1-DOA1 H5 R2-D1-DOA2 H6 R2-D1-DOA3

[0038] Table 1. Example hypothesis H of the points of the initial point cloud IPC.

[0039] Typically, the initial point cloud IPC may have an expected number of assumptions for each point, or it may have no assumptions for any point. For explanation purposes, refer to Figure 3 An initial point cloud IPC is discussed that has a mixture of points that have and do not have the expected number of hypotheses. In block 225, it may be found that one or more points of the initial point cloud IPC1 do not have all of the expected hypotheses (e.g., a point in IPC1 has only one or two hypotheses instead of six hypotheses). In this case, a check is performed to determine whether the point is not a point of the initial point cloud IPC1 at all, but rather a hypothesis of another compliant point of the initial point cloud IPC1. If, based on the check, no hypothesis is found that the point is another point, the correct number of hypotheses may be obtained by using a known fuzzy function. Figure 3 The process at block 225 is described in further detail.

[0040] The output of block 220 is a point cloud PC that is generated by ensuring that each initial point cloud IPC obtained at each radar system 110 has a complete set of hypotheses for each point. For a given radar system 110, the input initial point cloud IPC and the output point cloud PC can be the same. This is true if the initial point cloud IPC already has the expected number of hypotheses for each point. At block 230, resolving the ambiguity of the points in the common FOV 125 refers to selecting one of the hypotheses and discarding the rest before performing detection. Figure 4 This option is further explained in detail.

[0041] As previously described, different radar systems 110 may share different common FOVs 125. For example, a radar system 110 may have a FOV 115 that overlaps with the FOV 115 of a second radar system 110 (i.e., a first common FOV 125), and also have a separate portion that overlaps with the FOV 115 of a third radar system 110 (i.e., a second common FOV 125). The process at block 230 is performed separately for each different common FOV 125 (i.e., differently for the first common FOV 125 and the second common FOV 125). Thus, at block 230, points from point clouds PC of two or more radar systems 110 that are part of a given common FOV 125 are sequentially considered. Once at least some ambiguity has been resolved at block 230, additional known detection processes may be performed at block 240.

[0042] That is, each radar system 110 provides a point cloud. Based on one or more embodiments discussed herein, some points of the point cloud (i.e., some points that are part of common FOV 125) can clearly indicate range, Doppler, and direction of arrival. The point cloud from each radar system 110 essentially provides an image of field of view 115 corresponding to that radar system 110. This image can be used to identify the type of object (e.g., a person, another vehicle). The image indicated by the point cloud of radar systems 110 with common FOV 125 can be used to determine accuracy (e.g., two radar systems 110 may indicate different ranges for the same portion of a person in their respective images).

[0043] Figure 3 225 ( FIG. 226 ) for each radar system 110 according to one or more embodiments. Figure 2 ) is performed by the processing flow of FIG. 225 . As previously described, the process at box 225 is performed separately for each radar system 110 and starts with an initial point cloud IPC. At box 310 , a check is first made to see if each point in the initial point cloud IPC for a given radar system 110 has a full set of hypotheses. If so, then at box 315 , the initial point cloud IPC is the point cloud PC for the radar system 110 . If each point of the initial point cloud IPC does not have a full set of hypotheses (i.e., the same number of range hypotheses, Doppler hypotheses, and direction of arrival hypotheses as every other point), then the process at box 320 is reached. As previously described, typically the radar system 110 either provides an initial point cloud IPC that is a point cloud PC (i.e., all points have the expected number of hypotheses) or provides an initial point cloud IPC in which no point includes the expected number of hypotheses. For purposes of explanation, a mixture of these scenarios is addressed through the exemplary cases discussed herein.

[0044] In the exemplary case of a given initial point cloud IPC obtained (at block 215) from a given radar system 110, point P1 has the expected number of hypotheses (e.g., three hypotheses H1, H2, H3), point P2 is found to have only two of the three expected hypotheses, H1 and H2, and points P3 and P4 each have only one range, Doppler, and direction of arrival value. At block 320, the process includes using an ambiguity function to obtain additional hypotheses for points with an incomplete number of hypotheses (e.g., fewer than three). In the exemplary case, the ambiguity function is used to obtain hypothesis H3 for point P2 and all three hypotheses H1, H2, H3 for points P3 and P4.

[0045] At block 330, obtaining the Mahalanobis distance d of some or all points in the initial point cloud IPC refers to obtaining the Mahalanobis distance d of each point in the initial point cloud IPC for a given radar system 110 without the expected number of assumptions, as shown in Table 2. The Mahalanobis distance d is given by:

[0046]

[0047] In Equation 1, R, D, Az, and El refer to the range, Doppler, azimuth, and elevation angles, respectively, which define the specific assumptions. In addition, σ R , σ D , σ Az , σ El are the accuracy parameters associated with the range, Doppler, azimuth, and elevation of a given radar system 110, respectively. That is, for each calculation of the Mahalanobis distance d belonging to the same radar system 110, σ R , σ D , σ Az , σ El are the same. As shown in Table 2, index i represents a point in the initial point cloud IPC that is not associated with a hypothesis (e.g., point P3 and point P4 in the example), or represents a point and hypothesis for which there are fewer hypotheses than expected (e.g., point P2). For simplicity, each of points P3 and P4 can be considered to have a single hypothesis or estimate for range, Doppler, and direction of arrival. In addition, in Table 2, index j represents a point and hypothesis associated with a complete set based on the ambiguity function (at box 320). Thus, for example, the Mahalanobis distance d322 is between point P3 (i.e., i=3) and the second hypothesis of the second point (i.e., j=22). As another example, d2243 is between the second hypothesis H2 of the second point P2 (i.e., i=22) and the third hypothesis H3 of the fourth point P4 (i.e., j=43), which is obtained based on the ambiguity function at box 320.

[0048] As previously mentioned, Table 2 shows the Mahalanobis distances dij obtained between points obtained without a hypothesis or with an incomplete hypothesis set (shown along the columns of Table 2) and points with a full hypothesis set based on the fuzzy function (shown along the rows of Table 2). Points with a full hypothesis set (e.g., point P1) are not part of the process of blocks 320 to 340.

[0049]

[0050]

[0051] Table 2. Distances between points of the initial point cloud IPC.

[0052] At block 340, the process includes identifying points that are actually hypotheses rather than points. This can be based on, for example, sequentially checking the lowest distance value d in Table 2. In the exemplary case, based on the distance d323 (i.e., i=3 for point P3 and j=23 for hypothesis H3 for point P2) being the minimum value in Table 2, it can be determined that point P3 in the initial point cloud IPC is actually the third hypothesis for point P2. In this case, point P3 is removed as an independent point and used as hypothesis H3 for point P2 in the resulting point cloud PC. As another example, no distance d4j (the distance in the last row of Table 2) is below a threshold or is the lowest value. In this case, point P4 is retained as an independent point, and the hypotheses H1, H2, and H3 determined by using the fuzzy function are retained as hypotheses for point P4 in the point cloud PC.

[0053] At block 350, generating the point cloud PC refers to providing all points from the initial point cloud IPC that remain after processing at blocks 320, 330, and 340. Figure 2 As shown, the point cloud PC from each radar system 110 is provided to the process at box 230, which will refer to Figure 4 Further discussion.

[0054] Figure 4 Two or more radar systems 110 are shown in block 230 ( Figure 2 ). As previously described, the process at block 230 is performed for each common FOV 125 and therefore involves points from the point clouds PCs of at least two radar systems 110. At block 410, the process includes identifying points within the common FOV 125 in the point clouds PCs of two or more radar systems 110. For example, this identification may be based on range and direction-of-arrival values ​​associated with the point hypothesis.

[0055] In the exemplary case discussed herein for exemplary purposes, it is assumed that the first radar system 110 (e.g., Figure 1 The radar system 110a in FIG. 1 has three points Pa, Pb, and Pc, which are connected to the second radar system 110 (e.g., Figure 1 The two points Px, Py of the radar system 110b) are part of the same common FOV 125. Figure 3 For the example discussed, the three points P1, P2, P4, each with three hypotheses, generated by the process of block 225 would be one of the points of one of radar systems 110 shown in Table 3. Points Pa, Pb, Pc of the first radar system 110 may each have two hypotheses H1 and H2, while points Px, Py of the second radar system 110 may each have three hypotheses H1, H2, and H3.

[0056] At block 420, the process includes obtaining the Mahalanobis distance d between points of a pair of radar systems 110 (within a common FOV 125). In the exemplary case, the distance d is between points Pa, Pb, and Pc of a first radar system 110 and points Px and Py of a second radar system 110, as shown in Table 3. Specifically, points Pa, Pb, and Pc of the first radar system 110 and their corresponding hypotheses H1 and H2 are shown along the top, while points Px and Py of the second radar system 110 and their corresponding hypotheses H1, H2, and H3 are shown along the side. Each Mahalanobis distance dij according to Equation 1 is between a hypothesis (H1, H2, or H3) for one of points Px or Py and a hypothesis (H1 or H2) for one of points Pa, Pb, Pc, or Pd. For example, dx1a1 is the Mahalanobis distance between hypothesis H1 for point Px (i.e., i=x1) and hypothesis H1 for point Pa (j=a1). Similarly, dy3c2 is the Mahalanobis distance between hypothesis H3 (i.e., i=y3) at point Py and hypothesis H2 (j=c2) at point Pc. In Equation 1, the accuracy parameter σ is R , σ D , σ Az , σ El The worst case value between the two radar systems 110 is selected.

[0057]

[0058] Table 3. Distances between points in the common field of view.

[0059] At block 430, the process includes selecting a hypothesis based on the Mahalanobis distances dij of a point pair or pairs of points. Candidate pairs may be selected and then examined as described in detail. For example, the candidate pairs may be selected sequentially by selecting the lowest Mahalanobis distances dij. The candidate pairs may then be examined (e.g., filtered, gated) to determine whether the hypotheses for both points in the pair can be selected as the same unambiguous hypothesis for the point pair, and other hypotheses for the point pair may be eliminated.

[0060] For example, if the Mahalanobis distance dy3b2 (i.e., i=y3 and j=b2) is the lowest, then hypothesis H3 of point Py and hypothesis H2 of point Pb are candidate point pairs. In this case, check whether i=y3 and j=b2 satisfy all of the following conditions:

[0061]

[0062]

[0063]

[0064]

[0065] The checks represented by Equations 2-5 ensure that the candidate point pairs have similar hypotheses in each domain (i.e., range, Doppler, and direction of arrival). Because the candidate point pairs are associated with two different radar systems 110, the accuracy parameter σ R , σ D , σ Az , σ El is not identical for two points. The worst case scenario (i.e., the higher value of the precision parameter) is used in each domain. If the check is satisfied, the hypothesis (e.g., hypothesis H3 for point Py and hypothesis H2 for point Pb in the example) is considered to represent an unambiguous point, and the other hypotheses are discarded.

[0066] Therefore, in this exemplary case, hypotheses H1 and H2 for point Py and hypothesis H1 for point Pb can be discarded, and points Py and Pb are considered solved. Specifically, the range, Doppler, and direction of arrival indicated by hypothesis H3 for point Py are retained for the point cloud of second radar system 110, while hypothesis H2 for point Pb is retained for the point cloud of first radar system 110. Some point pairs in common FOV 125 may not qualify as candidate point pairs or may not pass the checks represented by equations 2-5. These issues remain unresolved. At block 440, obtaining resolved and unresolved points in common FOV 125 facilitates further processing in block 240. Specifically, resolving ambiguity for at least a subset of points in common FOV 125 improves the accuracy of any subsequent applications of the point clouds from each radar system 110.

[0067] Although the above disclosure has been described with reference to exemplary embodiments, it will be understood by those skilled in the art that various changes may be made and equivalents may be substituted for elements thereof without departing from the scope thereof. In addition, many modifications may be made to adapt a particular situation or material to the teachings of the present disclosure without departing from the basic scope of the present disclosure. Therefore, it is intended that the present disclosure is not limited to the particular embodiments disclosed, but rather includes all embodiments falling within its scope.

Claims

1. A method for alleviating blur based on a common field of view of a radar system, comprising: obtaining an initial point cloud for each of two or more radar systems sharing a common field of view, wherein each initial point cloud is generated by processing reflected energy at each of the two or more radar systems, and each point of the initial point cloud indicates one or more hypotheses about range, Doppler, and direction of arrival (DOA) of an object that caused the reflected energy; obtaining a point cloud for each of the two or more radar systems from the initial point cloud, wherein each point of the point cloud for each of the two or more radar systems has the same number of assumptions about range as other points of the point cloud, the same number of assumptions about Doppler as other points of the point cloud, and the same number of assumptions about direction of arrival as other points of the point cloud; identifying each point in the point cloud of each of the two or more radar systems that is in the common field of view; resolving ambiguities in the common field of view based on the point cloud of each of the two or more radar systems to obtain resolved points and unresolved points in the common field of view, wherein the resolved points indicate a range value, a Doppler value, and a direction of arrival value; and obtaining a radar image from each of two or more radar systems based on resolved points and unresolved points in a common field of view, wherein the radar image is used to control an aspect of vehicle operation; wherein resolving ambiguity in the common field of view comprises obtaining a Mahalanobis distance between each set of hypotheses for each point in the common field of view of one of the two or more radar systems and each set of hypotheses for each point in the common field of view of the other of the two or more radar systems using points in the common field of view of one of the two or more radar systems and the other of the two or more radar systems one at a time, the set of hypotheses comprising a combination of one of one or more hypotheses for range, one of one or more hypotheses for Doppler, and one of one or more hypotheses for direction of arrival.

2. The method of claim 1 , wherein obtaining a point cloud from the initial point cloud comprises, for each initial point cloud, determining a plurality of hypotheses for range, a plurality of hypotheses for Doppler, and a plurality of hypotheses for direction of arrival for each point of each initial point cloud, outputting the initial point cloud as a point cloud based on a number of hypotheses, the number of hypotheses for range for each point being a first number, the number of hypotheses for Doppler for each point being a second number, and the number of hypotheses for direction of arrival for each point being a third number, and processing the initial point cloud to obtain the point cloud based on the number of hypotheses, the number of hypotheses for range for at least one point being less than the first number, the number of hypotheses for Doppler for at least one point being less than the second number, or the number of hypotheses for direction of arrival for at least one point being less than the third number.

3. The method of claim 2 , wherein processing the initial point cloud comprises generating a subset of the initial point cloud to include each point having fewer than a first number of hypotheses for range, a second number of hypotheses for Doppler, and a third number of hypotheses for direction of arrival, and using an ambiguity function to obtain a complete subset of the initial point cloud including the first number of hypotheses for range, the second number of hypotheses for Doppler, and the third number of hypotheses for direction of arrival for each point in the subset of the initial point cloud, processing the initial point cloud further comprises obtaining a Mahalanobis distance between each point in the subset of the initial point cloud and each hypothesis in the complete subset of the initial point cloud, and identifying each point in the subset of the initial point cloud as a hypothesis in the complete subset of the initial point cloud based on the Mahalanobis distance, and processing the initial point cloud further comprises discarding each point in the subset of the initial point cloud as a hypothesis in the complete subset of the initial point cloud from the complete subset of the initial point cloud and retaining a remainder of the complete subset of the initial point cloud to generate the point cloud.

4. The method according to claim 1, wherein Resolving ambiguity in the common field of view also includes identifying a point pair that results in a lowest Mahalanobis distance, and based on the point pair passing a gating condition, retaining only the hypothesis set associated with the point pair with the lowest Mahalanobis distance as unambiguous, and checking the gating condition for the point pair, including each of a range, a Doppler, and a direction of arrival using the hypothesis set for the point pair. The method of claim 1 , wherein one or more objects are identified based on radar images of one or more of the two or more radar systems.

6. A vehicle capable of achieving blur reduction for a public field of view based on a radar system, comprising: Two or more radar systems; and a controller configured to obtain an initial point cloud for each of two or more radar systems that share a common field of view, wherein each initial point cloud is generated by processing reflected energy at each of the two or more radar systems, and each point of the initial point cloud indicates one or more hypotheses about range, Doppler, and direction of arrival (DOA) of an object that caused the reflected energy, obtain a point cloud from the initial point cloud for each of the two or more radar systems, wherein each point of the point cloud for each of the two or more radar systems has the same number of hypotheses about range as other points of the point cloud, the same number of hypotheses about Doppler as other points of the point cloud, and the same number of hypotheses about direction of arrival as other points of the point cloud, resolve ambiguities in the common field of view based on the point cloud for each of the two or more radar systems to obtain resolved points and unresolved points in the common field of view, wherein the resolved points indicate a range value, a Doppler value, and a direction of arrival value, and obtain a radar image from each of the two or more radar systems based on the resolved points and the unresolved points in the common field of view, wherein the radar image is used to control an aspect of vehicle operation; The controller is further configured to identify each point in the point cloud of each of the two or more radar systems that is in a common field of view, and the controller is configured to resolve ambiguity in the common field of view by obtaining a Mahalanobis distance between each set of hypotheses for each point in the common field of view of one of the two or more radar systems and each set of hypotheses for each point in the common field of view of the other of the two or more radar systems using the points in the common field of view of one of the two or more radar systems and the other of the two or more radar systems one at a time, the set of hypotheses comprising a combination of one of the one or more hypotheses for range, one of the one or more hypotheses for Doppler, and one of the one or more hypotheses for direction of arrival.

7. The carrier according to claim 6, wherein: For each initial point cloud, the controller is configured to obtain a point cloud by determining, for each point of each initial point cloud, a plurality of hypotheses for range, a plurality of hypotheses for Doppler, and a plurality of hypotheses for direction of arrival, outputting the initial point cloud as a point cloud based on a number of hypotheses, the number of hypotheses for range for each point being a first number, the number of hypotheses for Doppler for each point being a second number, and the number of hypotheses for direction of arrival for each point being a third number, and processing the initial point cloud to obtain a point cloud based on the number of hypotheses, the number of hypotheses for range for at least one point being less than the first number, the number of hypotheses for Doppler for at least one point being less than the second number, or the number of hypotheses for direction of arrival for at least one point being less than the third number.

8. The carrier according to claim 7, wherein: The controller is configured to process the initial point cloud by generating a subset of the initial point cloud to include each point having fewer than a first number of hypotheses for range, a second number of hypotheses for Doppler, and a third number of hypotheses for direction of arrival, and using an ambiguity function to obtain a complete subset of the initial point cloud including the first number of hypotheses for range, the second number of hypotheses for Doppler, and the third number of hypotheses for direction of arrival for each point in the subset of the initial point cloud, the controller is further configured to process the initial point cloud by obtaining a Mahalanobis distance between each point in the subset of the initial point cloud and each hypothesis in the complete subset of the initial point cloud, and identifying each point in the subset of the initial point cloud that is a hypothesis in the complete subset of the initial point cloud based on the Mahalanobis distance, and the controller is further configured to process the initial point cloud by discarding each point in the subset of the initial point cloud that is a hypothesis in the complete subset of the initial point cloud from the complete subset of the initial point cloud, and retaining a remainder of the complete subset of the initial point cloud to generate a point cloud.

9. The carrier according to claim 6, wherein: The controller is configured to resolve ambiguity in the common field of view by identifying a point pair that results in a lowest Mahalanobis distance, and based on the point pair passing a gating condition, retain only the hypothesis set for the point pair associated with the lowest Mahalanobis distance as unambiguous, and the controller is configured to check the gating condition for the point pair, including each of range, Doppler, and direction of arrival using the hypothesis set for the point pair.

10. The carrier according to claim 6, wherein: The controller is configured to identify one or more objects based on radar images from one or more of the two or more radar systems.

Images