Efficient Compression of Radar Data

Abstract

A two-stage technique for compressing radar data is disclosed, wherein, in a first stage, constant false alarm rate (CFAR) compression is performed using a median value calculated from at least one dimension of the radar data and then used to determine a filter threshold for all dimensions of the radar data. The radar data is then compressed by filtering the radar data based on the filter threshold. In a second stage, peak detection is then performed on the compressed data to identify detected objects (targets). These and other embodiments are described herein.

Description

Background Art

[0001] Radar uses radio frequency (RF) signals to determine the location and / or movement of an object. Traditionally, "ping" radar (low-resolution radar) has been used in air defense systems, flight control systems, and many other applications. Recently, however, "imaging" radar (high-resolution radar) has been able to provide images at much higher resolutions than ping radar is capable of. The higher resolution provided by imaging radar can be used for object detection in automotive and other applications.

[0002] However, the throughput and memory requirements of imaging radars can be quite demanding. Constant false alarm rate (CFAR) or peak detection can be used, but traditional techniques may require higher complexity and may not have sufficient power to process radar data from high-resolution imaging radar systems. In addition, alternative techniques may not be accurate enough for automotive and / or other modern applications. Summary of the invention

[0003] The technology described herein solves these and other problems by utilizing a two-stage scheme. In the first stage, CFAR compression is performed using the median value calculated from at least one dimension of the radar data and then used to determine the filter thresholds for all dimensions of the radar data. The radar data is then compressed by filtering the radar data based on the filter thresholds. In the second stage, the compressed data is then peak detected to identify the detected object (target). These and other embodiments are described herein.

[0004] An example method according to the present description includes obtaining input samples including values ​​of multiple dimensions measured by a radar, determining a median of the values ​​of the input samples of at least one of the multiple dimensions, and filtering the input samples based on the median to create a compressed radar data set. The filtering includes, for each of the multiple dimensions: determining a filter threshold for the respective dimension, and excluding from the compressed radar data set one or more values ​​of the input samples of the respective dimension that are less than the filter threshold above the median. The method also includes performing peak detection on the compressed radar data set to generate a list of one or more identified targets.

[0005] An example device for compressing radar data according to the present description includes: a memory, and one or more processing units communicatively coupled to the memory. The one or more processing units are configured to: obtain input samples including values ​​of multiple dimensions measured by a radar, determine a median of the values ​​of the input samples of at least one of the multiple dimensions, and filter the input samples based on the median to create a compressed radar data set. The filtering includes, for each of the multiple dimensions: determining a filtering threshold for the respective dimension, and excluding from the compressed radar data set one or more values ​​of the input samples of the respective dimensions that are less than the filtering threshold above the median. The one or more processing units are also configured to perform peak detection on the compressed radar data set to generate a list of one or more identified targets.

[0006] Another example device according to the present description includes: a component for obtaining input samples including values ​​of multiple dimensions measured by a radar, a component for determining a median value of the values ​​of the input samples of at least one dimension of the multiple dimensions, and a component for filtering the input samples based on the median value to create a compressed radar data set. The filtering includes, for each dimension of the multiple dimensions: determining a filtering threshold for the respective dimension, and excluding from the compressed radar data set one or more values ​​of the input samples of the respective dimension that are less than the filtering threshold above the median value. The device also includes a component for performing peak detection on the compressed radar data set to generate a list of one or more identified targets.

[0007] An example non-transitory computer-readable medium according to the present description has instructions embedded therein for compressing radar data. When executed by one or more processing units, the instructions cause the one or more processing units to perform one or more operations, the one or more operations including: obtaining input samples including values ​​of multiple dimensions measured by a radar, determining a median of the values ​​of the input samples of at least one of the multiple dimensions, and filtering the input samples based on the median to create a compressed radar data set. The filtering includes, for each of the multiple dimensions: determining a filtering threshold for each dimension, and excluding from the compressed radar data set one or more values ​​of the input samples of each dimension that are less than the filtering threshold above the median. When executed by the one or more processing units, the instructions also cause the one or more processing units to perform one or more operations including performing peak detection on the compressed radar data set to generate a list of one or more identified targets. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 is a block diagram illustrating an embodiment of a technique for compressing radar data.

[0009] Figure 2A is a diagram illustrating a binary search for determining a median in one dimension of an input sample, according to an embodiment.

[0010] Figure 2B is a flow chart illustrating a generalized form of a binary search process according to an embodiment.

[0011] Figure 3A-3C is a diagram illustrating how embodiments may allow values ​​in a one-dimensional array of values ​​to be partitioned into subsets for positioning.

[0012] Figure 4 is an illustration of a graph that may be used for threshold determination according to some embodiments.

[0013] Figure 5A and Figure 5B Two diagrams showing how the first stage is used to compress data are shown, each according to an embodiment.

[0014] Figure 6 is an illustration of a data structure that may be used for storage of compressed data output.

[0015] Figure 7 is an illustration of an example data flow architecture according to an embodiment.

[0016] Figure 8 is an illustration of an example DSP block diagram according to an embodiment.

[0017] Fig. 9 is a flow chart of a method of compressing radar data according to an embodiment.

[0018] Fig.10 An embodiment of a computer system that can be utilized as described herein is shown.

[0019] According to certain exemplary embodiments, similar reference symbols in each figure indicate similar elements. In addition, multiple instances of an element can be represented by following the first digit of the element with a letter or hyphen and a second digit. For example, multiple instances of element 110 can be indicated as 110-1, 110-2, 110-3, etc., or as 110a, 110b, 110c, etc. When only the first digit is used to refer to such an element, it should be understood that any instance of the element (for example, the element 110 in the aforementioned example will refer to element 110-1, 110-2 and 110-3 or to element 110a, 110b and 110c). DETAILED DESCRIPTION

[0020] Several exemplary embodiments will now be described with respect to the accompanying drawings which form a part hereof. Although specific embodiments are described below in which one or more aspects of the present disclosure may be implemented, other embodiments may be used and various modifications may be made without departing from the scope of the present disclosure or the spirit of the appended claims.

[0021] Imaging radars can be used in modern electronic devices and systems for a variety of industrial, commercial, military, and consumer applications. In particular, radars using RF frequencies with relatively small wavelengths (e.g., centimeters or millimeters, such as millimeter wave ("mmWave") radars) can be used for applications such as imaging for autonomous driving, object detection, etc. In these applications, radar systems can typically operate at 57-71 GHz, but can also include frequencies ranging from 30-300 GHz. For example, in some embodiments, a 5G band (e.g., 28 GHz) can be used. Nevertheless, depending on the desired functionality, manufacturing considerations, and / or other factors, the radar data compression techniques provided herein can be implemented in imaging radar systems using higher and / or lower radio frequency (RF) frequencies (e.g., outside the 30-300 GHz range).

[0022] As noted, imaging radars can generate large amounts of data. Each radar scan can generate radar data in multiple dimensions, including, for example, azimuth, elevation, time, Doppler, and range. CFAR can be used to significantly compress data by estimating the noise level within the radar data and then removing (compressing) the values ​​in the radar data that do not exceed a threshold minimum value above the noise level. Ordered Statistics CFAR (OS-CFAR) uses sorting to obtain the median of at least part of the radar data and is considered to be very effective when compared to cell average CFAR (CA-CFAR), which takes the average of the pixel values ​​within a defined "cell". However, implementations of OS-CFAR can be very complex, particularly implementations of OS-CFAR for sliding windows that process radar data in multiple dimensions. Peak CFAR involves determining the peak of one or more dimensions by comparing the value to the surrounding values. However, the problem is that peak CFAR does not work well in environments with relatively large amounts of noise. Furthermore, regardless of which type of CFAR is used, the high throughput of imaging radar makes the radar data particularly difficult to compress in software, and traditional hardware implementations are often inflexible and unable to adapt to changes that may be required by various vendors.

[0023] The embodiments described herein address these and other issues by utilizing a two-stage radar compression technique that produces a high degree of compression and can be implemented at least in part in configurable hardware that can accommodate differences in vendor requirements. Figure 1 An example embodiment is shown in FIG.

[0024] Figure 11 is a block diagram illustrating an embodiment of a technique for compressing radar data. As shown, the technique can be divided into a first stage 110 and a second stage 120. The functions shown in each stage 110, 120 can be performed by a computer system such as Fig.10 Executed by hardware and / or software components of a computer system as shown in FIG. 1 and described below. For example, this may include Figure 7 The Segment Routing (SR) flow architecture shown in and / or Figure 8 The digital signal processor (DSP) architecture shown in FIG.

[0025] Here, input samples 130 include radar samples from the following four dimensions: azimuth, elevation, range, and Doppler (which may be abbreviated herein and in the figures as AZ, EL, R, and D, respectively). That is, input samples 130 include a 4D radar image, where a "pixel" has a measured value for each azimuth, elevation, range, and Doppler. (e.g. Figure 1 As shown, input samples 130 may include absolute values ​​of measured values.) Depending on the desired functionality, hardware limitations, processing issues, and / or other factors, the size of each dimension may vary. For example, in Figure 1 In the example shown in , there are 128 Doppler bins for each azimuth, elevation, and range. (Alternative embodiments may have a greater or lesser number of values.) The values ​​of input samples 130 may include digital data (e.g., 16-bit or 12-bit values) converted by the radar from analog readings.

[0026] The input samples 130 are used for noise estimation, where a median is calculated for the values ​​in at least one dimension of the input samples 130. That is, if a strong target is present, it can be assumed that there will be a single peak in at least one dimension that represents the target, and the rest is noise that can be filtered out. Depending on the desired functionality, Doppler is Figure 1 The dimensions chosen in the example shown in , although any dimension could be chosen. Even so, using Doppler may be advantageous because the noise for Doppler is primarily due to the crystal and is relatively uniform compared to the noise in the other dimensions.

[0027] In some embodiments, the first stage 110 may include determining medians in more than one dimension. For example, some embodiments may perform a form of dual CFAR, where a median determination and median-based data compression 145 (discussed in more detail below) may be performed for a first dimension (e.g., Doppler), and then the data may be further compressed by performing similar steps with respect to a second dimension (e.g., azimuth), which will have their own respective thresholds. In some embodiments, compression may be performed for additional dimensions in a similar manner. Thus, according to some embodiments, the first stage 110 may include functionality for median determination 135, threshold comparison, and subsequent data compression 145 with respect to more than one dimension.

[0028] Although in the form of OS-CFAR, the determination of the median here can be performed efficiently because it can process a single dimension and does not involve data windowing. (In addition, for the case where the median is determined for multiple dimensions, such embodiments can determine the median for those individual dimensions one at a time.) As will be appreciated by those of ordinary skill in the art, the determination of the median in a one-dimensional array can be performed in any of a variety of ways. As previously described, some embodiments can utilize hardware-based solutions to accommodate the bandwidth required to process the input samples 130.

[0029] Figure 2A is a diagram illustrating a binary search according to an embodiment for determining the median in one dimension of input samples 130. Here, instead of performing a sort, an embodiment may utilize a vectorized array of comparators (e.g., 128 comparators) that calculate the median every n-bit period. All input samples of the dimension may be compared to a threshold value that is based on the number of iterations and samples (values) in the array, which provides a cumulative distribution histogram of the one-dimensional array from which the median value may be determined.

[0030] As a simple example, for a 1D array of 8 values:

[0031] 1 2 5 67 69 32 1 67

[0032] Where each value represents a sample of the radar in the dimension (e.g., Doppler) for which the median is to be determined, each of which can have a maximum possible value of 128. To obtain a solution of order 4 (half the number of values), the array can undergo an iterative process as shown in Table 1.

[0033]

[0034]

[0035] Table 1: Example binary search values

[0036] As can be seen, the threshold is set to 64 for the first iteration (because it is half of the maximum value of 128), and the number of values ​​equal to or greater than 64 is returned (3). With each subsequent iteration, Figure 2A As shown, a new threshold is determined based on whether the number of values ​​is equal to or greater than the previous threshold, where the new threshold is the old threshold ± half of the absolute value of the previous change (e.g., based on the number of values ​​greater than or equal to the threshold, the second iteration increases or decreases the threshold by 32, the third iteration increases or decreases the threshold by 16, the fourth iteration increases or decreases the threshold by 8, etc.). If the number of values ​​that meet or exceed the threshold is less than the sequence number, the threshold is lowered, and if the number is greater than or equal to the sequence number, the threshold is increased. Therefore, as can be seen from Table 1, the threshold is selected at each iteration so that the number of values ​​that meet or exceed the threshold converges to sequence number 4. Then, the value 32 is the median. As will be appreciated by those of ordinary skill in the art, a similar process can be used to determine the median of values ​​of a larger one-dimensional array.

[0037] Figure 2B is a flow chart illustrating a general form of a binary search process according to an embodiment. However, it will be appreciated that alternative techniques may be used to perform a binary search to determine the median of radar samples in a particular dimension. Furthermore, in alternative embodiments, some functions may be combined, separated, and / or performed in an alternative order depending on the desired functionality.

[0038] The functionality of block 210 includes setting an ordered statistics (OS) search value, which may be based on an ordered statistics technique. (In the above example, the OS search value corresponds to a threshold. Initially, this is 64, corresponding to half of the maximum possible sample value.) As shown in block 220, each value in the one-dimensional array of radar sample values ​​is compared to the OS search value (e.g., using a comparator array) to determine the number of samples in the array that are equal to or greater than the OS search value.

[0039] The result is then compared to the OS order to determine if further processing is required. Again, the determination of the OS order can be based on an ordered statistical technique. (In the above example, the OS order number is 4, corresponding to half the number of values ​​in the one-dimensional array.) As shown in block 230, if the number of samples having values ​​greater than the OS search value is equal to or greater than the OS order, the OS search value can be incremented (as shown in block 240) and the comparison at block 220 repeated. (The incrementation of the OS search value can be performed in a binary search manner, such as Figure 2A , by incrementing the value to the value at the midpoint between the current OS search value and the maximum or previously searched OS search value).

[0040] If the number of samples having values ​​greater than the OS search value is not equal to or greater than the OS order, the number of samples having values ​​equal to the OS search value is compared to the OS order, as shown in block 250. If the number of samples having values ​​equal to the OS search value is not greater than or equal to the OS order, the OS search value may be decremented (as shown in block 260) and the comparison at block 220 repeated. (Decrementing the OS search value may also be performed in a binary search manner by reducing the value to a minimum value or a value at a midpoint between previously searched OS search values).

[0041] Finally, as indicated at block 270, if the number of samples having a value equal to the OS search value is greater than or equal to the OS order, the OS value (median) is determined to be the current OS search value.

[0042] It may be noted that while some embodiments may utilize a binary search rather than a moving window to compute the median, some embodiments may allow for the values ​​to be split into multiple subsets for positioning. Figure 3A-3C An example is shown in . Although alternative embodiments may have a larger or smaller number of input samples in a one-dimensional array, Figure 3A The example shown in assumes a one-dimensional array of 128 input samples (values).

[0043] Figure 3A is a diagram showing how a single median value may be calculated from an entire one-dimensional array according to an embodiment (e.g., using the previous description of Figure 2A and 2B description method).

[0044] Figure 3B is a diagram of an alternative embodiment in which a one-dimensional array is partitioned into four subsets. More specifically, a one-dimensional array of 128 samples is partitioned into four equal, mutually exclusive subsets of 32 samples each. The median of each subset can then be calculated. Thus, each subset can have a different noise estimate, which can be helpful in instances where there may be a relatively large number of targets within a certain range. (For example, in the case of using Doppler, there may be a large number of targets within a range of velocities).

[0045] Figure 3C is a diagram of another alternative embodiment, in which the one-dimensional array is partitioned into seven subsets. Figure 3B In the example in , the subsets again consist of 32 samples each. However, here the subsets overlap (i.e., are not mutually exclusive). This can result in Figure 3B However, due to the overlap in the subsets, the median may produce a more accurate solution when the fruit sets are smaller (e.g., with little or no overlap).

[0046] Back to Figure 1Once the median determination 135 has been made, a threshold determination 140 is made, wherein the threshold is added to the determined median for the purpose of compressing the input samples 130 by filtering out values ​​that do not meet the threshold. In other words, when added to the determined median, the threshold represents a value that exceeds the noise floor, which, if exceeded, represents a detected target. As described above, different thresholds may be determined for and applied to values ​​of the input samples 130 in different dimensions. The threshold may be determined in any of a variety of ways. Figure 4 Helps illustrate one technique for determining a threshold.

[0047] Figure 4 is an illustration of a graph that may be used for threshold determination in accordance with some embodiments. Here, the number of targets per frame (manually verified for visible targets) is plotted for various thresholds in increasing dB. As can be seen, the number of targets is a non-linear function with two regions of interest. The first region 410 at a relatively low threshold represents a rapid increase in target detection, while the second region 420 at a higher threshold represents a limited increase in target detection (noise statistics begin to dominate). A transition point 430 may be identified at the approximate boundary between the first region 410 and the second region 420. In accordance with some embodiments, the transition point 430 may represent an optimal threshold. Depending on the desired functionality, the threshold may be determined based on the location of the transition point 430. For example, in some embodiments, the threshold may be selected to be a predetermined amount dB lower than the transition point 430 to help ensure that targets are not filtered out of the data. For example, at Figure 4 If the predetermined amount is 3 dB, then the resulting threshold would be 10 dB (3 dB lower than the 13 dB transition point 430).

[0048] Back to Figure 1 , as the thresholds are determined, the thresholds can then be added to the determined median to create a threshold map. During data compression 145, if input samples 130 of a given dimension fail to meet the respective thresholds for that dimension, they can be filtered out. This can result in a large amount of compression. Depending on the thresholds and the scene itself, this compression can result in a ratio of 50:1-1500:1.

[0049] Figure 5A and Figure 5B Each embodiment shows two graphs, upper graph 510 and lower graph 520, which illustrate how the first stage 110 is used to compress data. However, one of ordinary skill in the art will appreciate that embodiments may utilize different thresholds, and the situations in which multiple targets are identified may vary. Therefore, Figure 5A and Figure 5B These are provided as non-limiting examples only.

[0050] exist Figure 5A In the upper graph 510 of , the median Doppler value is mapped onto the azimuth and range values. In the lower graph 520, a first graph 530 of Doppler values ​​for a corresponding first point (with a given azimuth and range) on the upper graph 510 is close to a strong target. (An arrow 540 connects the first graph 530 to the corresponding first point in the upper graph 510). As indicated by the shading of the first point in the upper graph 510, the median value of the first graph 530 exceeds 110 dB. This first median value 550 is shown in the lower graph 520. Using the techniques described above, a predetermined threshold value (approximately 10 dB in the example shown in FIG. 5 ) can be multiplied by the first median value 550 to provide a first threshold value 560 for filtering out data. As can be seen, most of the data from the first graph 530 can be filtered out. Therefore, in the compressed data ( Figure 1 In the output of the first stage 110), only the central peak of the first graph 530 may be retained.

[0051] A similar process can be used to filter out Figure 5B The data of the second graph 570 shown in FIG. Figure 5B 5, the lower graph 520 includes a second graph 570 of Doppler values ​​for the corresponding second point on the upper graph 510 away from the strong target. (Again, arrow 540 connects the second graph 570 to the corresponding second point in the upper graph 510). The median of the second graph 570 is shown as a second median value 580 in the lower graph 520, and a second threshold value 590 is provided that is 10 dB higher than the second median value 580. As can be seen, all of the data from the second graph 570 can be filtered out.

[0052] According to some embodiments, the compressed data output of the first stage 110 (e.g., Figure 5A and 5B As shown, after filtering the input samples 130 using a threshold value, may be stored in a manner that allows efficient data retrieval while maintaining small memory requirements.

[0053] Figure 6 6 is an illustration of a data structure that may be used for storage of compressed data output according to an embodiment. The data structure includes an indirect access table 610 (for each elevation angle) and compressed data 620. Such a data structure may allow for exceptionally high access times while maintaining relatively low memory usage. Example sizes of various entries in the data store are indicated at the top of the entry (e.g., "16 bits," "1 bit," "7 bits"), although it may be noted that alternative embodiments may include the same Figure 6 Furthermore, different embodiments may have dimensions of different sizes than those shown.

[0054] The indirect access table 610 comprises a fixed size table having a series of entries for each range and azimuth angle for a given elevation angle. (Thus, Figure 6 The example indirect access table 610 in shows only an entry for a single elevation angle. Other elevation angles will have their own indirect access table 610. Even so, in some alternative embodiments, a single table may be used for all elevation angles. Here, for each range, the azimuth index spans from 0 to X, representing X+1 azimuths. The range index spans from 0 to Y, representing Y+1 ranges. It will be appreciated that alternative embodiments may have azimuth, range, and elevation entries in different ways.

[0055] Indirect access table 610 indicates for each entry whether a value exceeding a threshold has been detected. If, for a given range, azimuth, and elevation, the Doppler value exceeds the threshold, a pointer is used to point to the location in compressed data 620 that has the Doppler data. Otherwise, indirect access table 610 will include a null value ("FFFF") instead of a pointer. Figure 6 In the example shown, the values ​​in the indirect access table 610 for range 0, azimuth 3; range 1, azimuth 1; and range 1, azimuth 5 have pointer values ​​of 0, 17, and 20, respectively. These are pointer values ​​that point to locations within the compressed data 620 (shown as individual rows). (For simplicity, only a portion of the values ​​in the compressed data 620 are shown. That is, in some instances, each entry will include a value).

[0056] Depending on the desired functionality, the formatting information of the compressed data 620 may vary. In some embodiments, the compressed data 620 may include a Doppler index at the location where the detection occurred (at a given azimuth, range, and elevation), as well as the detection value. Figure 6 , software memory access may be simplified by separating the compressed data 620 into two parts: one part for indexes (shown in index "buckets" 630), and one part for detected values ​​(shown in value "buckets" 640). In some embodiments, the compressed data 620 may further include a "last" designation to indicate whether the corresponding Doppler is the last Doppler value detected for a particular elevation, range, and azimuth angle.

[0057] In some embodiments, the compressed data 620 can be optimized for parallel access. For example, different portions of the compressed data 620 (e.g., ranges of data) can be stored in different memory banks of a memory block. Due to parallelism, this can allow data stored in and / or retrieved from the compressed data 620 with very little latency and high throughput. For example, in some embodiments, the peak detection 150 (discussed in more detail below) performed in the second stage 120 can be performed using parallel processes that access multiple memory banks simultaneously. That is, in some embodiments, a particular machine can operate in parallel, and retrieving and decompressing data from compressed memory is very little latency and very high throughput.

[0058] Back again Figure 1 , the second stage 120 may be performed by performing peak detection 150 on the compressed data 160. Performing peak detection may involve performing 3D or 4D peak detection by determining whether there are peaks across three or four dimensions in the compressed data 160. This may mean checking corresponding values ​​and adjacent values ​​across all dimensions. More generally, for the number of dimensions N, this may involve checking 3 N -1 point (26 points for 3D peak detection, and 80 points for 4D peak detection). Peaks found from peak detection 150 may be added to a target list, which may be an output of the second stage 120. The determination of whether to perform 3D or 4D peak detection may be based on preference, and may be a configurable aspect, as discussed in more detail below.

[0059] Because peak detection 150 can be performed on the efficiently stored compressed data 160, peak detection can be performed quickly. That is, because compressed data 160 can be randomly accessible (e.g., as Figure 6 ), so nearby values ​​can be identified and compared across three or four dimensions (for 3D or 4D peak detection) more quickly than if the compressed data were stored in a grid or other order. In addition, peak detection 150 can provide additional compression. For example, the target list can be 10-40 times smaller than the compressed data 160.

[0060] Figure 7 is an illustration of an example data flow architecture 700 according to an embodiment. The blocks of threshold CFAR shown correspond to Figure 1 The functionality of the first stage 110 shown in FIG. 1 and the blocks of 3D peak and 4D peak shown correspond to the functionality of the second stage 120 . It will be appreciated that Figure 7 The time delays shown in are associated with a specific embodiment based on the hardware utilized. Alternative embodiments may have different delays.

[0061] Figure 8 is an illustration of an example DSP block diagram showing an embodiment of a DSP 800 that may be used to perform one or more of the functions described herein. Figure 8 Aspects of the various components shown in can be used to perform the techniques for data compression described herein, which may include data compression and storage as taught in the embodiments described above. Figure 8 The block diagrams provided in the foregoing are provided only as non-limiting examples. One of ordinary skill in the art will appreciate that various changes may be made while still allowing the DSP to provide the functionality described herein. The DSP 800 may be included in a computer system such as Fig.10 (Item 1020) shown and described below.

[0062] CFAR and peak detection in the manner described herein may be performed by components within the respective "CFAR" and "Peak" blocks of the DSP 800. The multiple arrows extending from the Peak block to the Compressed Memory block illustrate how multiple peak detection engines may operate in parallel and access multiple locations and memories simultaneously. The peak detection engines may then perform peak detection on the accessed values ​​(e.g., by performing comparisons). As previously described, peak detection may be performed by comparing values ​​across three or four dimensions.

[0063] The peak detection performed by the peak detection engine can be customized in any of a variety of ways to employ different types of peak detection depending on the desired functionality. For example, the peak detection engine can not only find peaks among adjacent values, it can also be customized to further include a threshold value so that the peak value is only identified as a peak if it has a value greater than the adjacent values ​​by at least a threshold amount. Additionally or alternatively, the peak detection engine can be customized to simply identify a peak as a value greater than the average value among the adjacent values, and / or greater than the average value by at least a threshold amount. In some embodiments, the peak control engine can utilize control registers representing values ​​from different dimensions (e.g., different registers for different dimensions) to perform comparisons of values ​​so that a given value can be quickly compared against a maximum value or an average value among adjacent values. In some embodiments, the selection of a threshold value (which can be positive or negative), a mathematical configuration (e.g., comparison against a maximum or an average), and / or the size of the region for analysis (e.g., the amount of adjacent values ​​to be compared) can be configurable.

[0064] Ultimately, the functionality described herein may be implemented by hardware and / or software components in any of a variety of ways. Furthermore, some embodiments may allow for additional types of configurability in one or more aspects. For example, embodiments may allow for the use of dimensions other than Doppler for median determination and / or for determining local medians (e.g., as Figure 3A-3C), and / or determination of the dimensions of 3D and / or 4D peak detection. According to some embodiments, the configuration may optionally use a "passing by" threshold in peak detection (rather than a simple "greater than" comparison).

[0065] Fig. 9 900 is a flowchart of a method 900 for compressing radar data according to an embodiment. Alternative embodiments may be combined, separated or otherwise changed. Fig. 9 The functions described in the blocks shown in the figure vary in functionality. Fig. 9 One or more of the functional components of the blocks shown may include hardware and / or software components of a computer system, such as Fig.10 As described above, the computer system may include a computer system 1000 as shown in FIG. Figure 7 and / or Figure 8 One or more components shown in .

[0066] At block 910, the functionality includes obtaining input samples that include values ​​of radar samples in multiple dimensions. As described in the embodiments provided herein, these dimensions may include two or more of azimuth, elevation, Doppler, or range. The input samples include raw radar data, or we include radar data that has undergone some preliminary processing (e.g., where the absolute value of the raw radar data is taken). The components for performing the functionality at block 910 may include, for example, radar 1075, wireless communication interface 1030, and / or other component(s) of computer system 1000, such as Fig.10 shown and described in more detail below.

[0067] At block 920, a median of the values ​​of the input samples of at least one of the plurality of dimensions is determined. As described above with respect to FIG. 2, in some embodiments, determining the median may include performing a binary search using a comparator array. In some embodiments, this may include iteratively comparing the values ​​of the input samples to a search value using the comparator array, and for each iteration: (i) comparing the number of input samples having values ​​greater than the search value to an ordered statistical order, and (ii) changing the search value based on the comparison. Means for performing the functions at block 920 may include, for example, processing unit(s) 1010, DSP 1020, and / or other component(s) of computer system 1000, such as Fig.10 As shown and described in more detail below, these components can be used to perform Figure 8 The DSP 800 is shown as one or more components of the CFAR block.

[0068] The functionality at block 930 includes filtering the input samples based on the median value to create a compressed radar data set, wherein the filtering includes: for each dimension in a plurality of dimensions, determining a filter threshold for the respective dimension, and excluding from the compressed radar data set one or more values ​​of the input samples for the respective dimension that are less than the filter threshold above the median value. As indicated in the embodiments described above, the filter threshold may be set at a predetermined value above the determined median value.

[0069] like Figure 3A-3C As described in , multiple medians can be calculated for a dimension by dividing the dimension into subsets. Therefore, in some embodiments, determining the median of the values ​​of the input samples of the dimension includes determining the median of each of the multiple subsets within the input samples of the dimension. In such embodiments, for each dimension in the multiple dimensions, determining the filtering threshold for the respective dimension may include determining multiple filtering thresholds including the filtering threshold for each subset in the multiple subsets, and / or excluding one or more values ​​of the input sample from the compressed radar data set of the respective dimension includes excluding one or more values ​​of the input sample that are less than at least one of the multiple filtering thresholds. In some embodiments (such as Figure 3B ), multiple subsets within the input samples of the dimension are mutually exclusive. However, as indicated, in some embodiments, multiple subsets within the input samples of the dimension may not be mutually exclusive.

[0070] The means for performing the functions at block 930 may include, for example, processing unit(s) 1010, DSP 1020, and / or other components of computer system 1000, such as Fig.10 As shown and described in more detail below, these components can be used to perform Figure 8 The DSP 800 is shown as one or more components of the CFAR block.

[0071] At block 940, the functionality includes performing peak detection on the compressed radar data set to generate a list of one or more identified targets. The compressed radar data may be stored in a specific memory that may be randomly accessed via an indirect access table (e.g., Figure 6). In addition, according to some embodiments, the compressed data may be distributed among memory banks of the memory such that different portions of the compressed data are simultaneously accessible, which may increase throughput when writing to and / or reading from the compressed data. Depending on the desired functionality, 3D and / or 4D peak detection may be performed on the compressed data. Thus, according to some embodiments, the plurality of dimensions may include four dimensions, and performing peak detection on the compressed radar data set includes identifying peaks in at least three of the plurality of dimensions. Means for performing the functions at block 920 may include, for example, processing unit(s) 1010, DSP 1020, and / or other components of computer system 1000, such as Fig.10 As shown and described in more detail below, these components can again be used to perform Figure 8 The DSP 800 is shown as one or more components of the CFAR block.

[0072] Fig.10 1000 is an example of a computer system 1000 that may be utilized as described above. For example, computer system 1000 may include a computer system that is incorporated into a vehicle and used to manage one or more systems related to navigation and / or autonomous driving of the vehicle, as well as communicate with other onboard systems and / or other traffic entities. Computer system 1000 may be used to collect radar data, and thus may perform Fig. 9 One or more of the functions of method 900. It should be noted that Fig.10 It is intended only to provide a broad illustration of various components, any or all of which may be utilized as appropriate. It may be noted that in some cases, Fig.10 The components shown may be located in a single physical device and / or distributed among various network devices that may be placed in different physical locations on the vehicle. Nevertheless, embodiments are not limited to vehicles and may include computer system 1000 incorporated into static or mobile devices other than vehicles.

[0073] Computer system 1000 is shown as including hardware elements that may be electrically coupled (or may otherwise be in communication, as appropriate) via bus 1005. The hardware elements may include (multiple) processing units 1010, which may include, but are not limited to, one or more general-purpose processors, one or more special-purpose processors (such as DSPs, graphics acceleration processors, application-specific integrated circuits (ASICs), etc.), and / or other processing structures or components. Fig.10 As shown in , some embodiments may have a separate DSP 1020, depending on the desired functionality. The DSP 1020 and / or the processing unit(s) 1010 may include Figure 8 The DSP shown and / or performs Figure 8 The computer system 1000 may also include one or more input devices 1070, which may include devices related to a user interface (e.g., a touch screen, a touchpad, a microphone, (multiple) buttons, (multiple) dials, (multiple) switches, etc.), and / or devices related to navigation, autonomous driving, etc. Similarly, one or more output devices 1015 may be related to user interaction (e.g., via a display, (multiple) light emitting diodes ((multiple) LEDs), (multiple) speakers, etc.), and / or devices related to navigation, autonomous driving, etc.

[0074] The computer system 1000 may also include a wireless communication interface 1030, which may include but is not limited to a modem, a network card, an infrared communication device, a wireless communication device and / or a chipset (such as Device, IEEE 802.11 device, IEEE 802.15.4 device, Wi-Fi device, WiMAX device, WAN device and / or various cellular devices, etc.), etc., which can enable the computer system 1000 to communicate with other devices via any of various communication standards. Communication can be performed via one or more wireless communication antennas 1032 that send and / or receive wireless signals 1034. In some embodiments, the wireless communication interface 1030 may include a radar system. And therefore, the wireless communication interface 1030 can be configured not only for RF communication, but also for imaging as described in the embodiments of this article. To this end, the wireless communication interface 1030 may include a separate transceiver and / or a separate receiver and transmitter, or any combination of a transceiver, a transmitter and a receiver capable of sending and receiving RF signals for radar functions. Therefore, the antenna 1032 may include multiple discrete antennas, antenna arrays, or any combination.

[0075] Computer system 1000 may also include sensor(s) 1040. In some embodiments, sensor(s) 1040 may include a radar system 1075 as described in the embodiments detailed herein. Here, radar 1075 may be external to computer system 1000 or integrated into computer system 1000 and may include one or more antennas for transmitting and / or receiving radar RF signals. Additionally or alternatively, as described above, the radar may be incorporated into wireless communication interface 1030.

[0076] Additionally or alternatively, the sensor(s) 1040 may include, but are not limited to, one or more accelerometers, gyroscopes, cameras, magnetometers, altimeters, microphones, proximity sensors, light sensors, barometers, LIDAR, and the like.

[0077] Embodiments of computer system 1000 may also include a global navigation satellite system (GNSS) receiver 1080 capable of receiving signals 1084 from one or more GNSS satellites using antenna 1082 (which may be the same as antenna 1032). Positioning based on GNSS signal measurements may be utilized in conjunction with imaging radar to provide higher-order functions, including complex navigation and position determination. GNSS receiver 1080 may extract the location of computer system 1000 from GNSS satellites of a GNSS system, such as a global positioning system (GPS) and / or similar systems, using conventional techniques.

[0078] Computer system 1000 may also include and / or be in communication with memory 1060. Memory 1060 may include, but is not limited to, local and / or network accessible storage, disk drives, drive arrays, optical storage devices, solid-state storage devices such as random access memory (RAM) and / or read-only memory (ROM), which may be programmable, flash-updatable, and the like. Such storage devices may be configured to implement any suitable data storage, including but not limited to various file systems, database structures, and the like. In some embodiments, memory 1060 may include a plurality of storage devices, such as a plurality of storage devices, and a plurality of storage devices, such as a plurality of storage devices, and a plurality of storage devices. Figure 1 In addition, those memory banks may be incorporated into DSP 1020 and / or processing unit(s) 1010.

[0079] The memory 1060 of the computer system 1000 may also include software elements ( Fig.10 ), the software elements include operating systems, device drivers, executable libraries and / or other codes such as one or more applications, which may include computer programs provided by various embodiments, and / or as described herein, may be designed to implement methods provided by other embodiments, and / or configure systems provided by other embodiments. Merely by way of example, one or more processes described with respect to the (multiple) methods discussed above may be implemented as code and / or instructions in memory 1060 executable by computer system 1000 (and / or (multiple) processing units 1010 or DSP 1020 within computer system 1000). Then, in one aspect, such code and / or instructions may be used to configure and / or adapt a general purpose computer (or other device) to perform one or more operations according to the described methods.

[0080] It will be apparent to those skilled in the art that substantial variations may be made depending on specific requirements. For example, customized hardware may also be used, and / or specific elements may be implemented in hardware, software (including portable software such as applets, etc.), or both. In addition, connections to other computing devices (such as network input / output devices) may be employed.

[0081] With reference to the accompanying drawings, a component that may include a memory may include a non-transitory machine-readable medium. As used herein, the terms "machine-readable medium" and "computer-readable medium" refer to any storage medium that participates in providing data that causes a machine to operate in a particular manner. In the embodiments provided above, various machine-readable media may be related to providing instructions / codes to a processing unit and / or other (multiple) devices for execution. Additionally or alternatively, a machine-readable medium may be used to store and / or carry such instructions / codes. In many embodiments, a computer-readable medium is a physical and / or tangible storage medium. Such a medium may take many forms, including but not limited to non-volatile media, volatile media, and transmission media. The common form of a computer-readable medium includes, for example, magnetic and / or optical media, any other physical medium with a hole pattern, RAM, programmable ROM (PROM), erasable PROM (EPROM), FLASH-EPROM, any other memory chip or box, a carrier wave as described below, or any other medium from which a computer can read instructions and / or codes.

[0082] The methods, systems and devices discussed herein are examples. Various embodiments may omit, replace or add various processes or components as appropriate. For example, the features described with respect to certain embodiments may be combined in various other embodiments. Different aspects and elements of the embodiments may be combined in a similar manner. The various components in the figures provided herein may be embodied in hardware and / or software. In addition, technology is evolving, so many elements are examples, and the scope of this disclosure is not limited to these specific examples.

[0083] It has proven convenient at times, primarily for reasons of common usage, to refer to such signals as bits, information, values, elements, symbols, characters, variables, terms, numbers, reference numerals, and the like. It will be understood, however, that all of these or similar terms are associated with the appropriate physical quantities and are merely convenient labels. Unless otherwise specifically stated, it will be apparent from the above discussion that discussions throughout this specification utilizing terms such as "processing," "computing," "calculating," "determining," "determining," "identifying," "associating," "performing," and the like refer to the actions or processes of a specific apparatus, such as a special-purpose computer or similar special-purpose electronic computing device. Thus, in the context of this specification, a special-purpose computer or similar special-purpose electronic computing device is capable of manipulating or converting signals, which are typically represented as physical electronic, electrical, or magnetic quantities within a memory, register, or other information storage device, transmission device, or display device of the special-purpose computer or similar special-purpose electronic computing device.

[0084] The terms "and" and "or" used herein may include a variety of meanings, which are also expected to depend at least in part on the context in which the terms are used. Generally, if "or" is used in an associative list, such as A, B, or C, it is intended to mean A, B, and C used in an inclusive sense here, and A, B, or C used in an exclusive sense here. In addition, the term "one or more" used herein may be used to describe any feature, structure, or characteristic of the singular, or may also be used to describe some combination of features, structures, or characteristics. However, it should be noted that this is merely an illustrative example, and the claimed subject matter is not limited to this example. In addition, if the term "at least one of..." is used in an associative list, such as A, B, or C, it may be interpreted to mean any combination of A, B, and / or C, such as A, AB, AA, AAB, AABBCCC, etc.

[0085] Having described several embodiments, various modifications, alternative constructions, and equivalents may be used without departing from the spirit of the present disclosure. For example, the above elements may be merely components of a larger system, wherein other rules may take precedence over or otherwise modify the application of the various embodiments. In addition, several steps may be taken before, during, or after considering the above elements. Therefore, the above description does not limit the scope of the present disclosure.

[0086] In view of this description, embodiments may include different combinations of features. Implementation examples are described in the following numbered clauses:

[0087] 1. A method for compressing radar data, the method comprising:

[0088] obtaining input samples comprising values ​​of a plurality of dimensions measured by the radar;

[0089] determining a median of the values ​​of the input samples in at least one of a plurality of dimensions;

[0090] Filtering the input samples based on a median value to create a compressed radar dataset, wherein the filtering comprises, for each of a plurality of dimensions:

[0091] Determine the filtering thresholds for each dimension; and

[0092] excluding from the compressed radar data set one or more values ​​of input samples for each dimension that are less than a filtering threshold above the median value; and

[0093] Peak detection is performed on the compressed radar data set to generate a list of one or more identified targets.

[0094] 2. The method of clause 1, wherein the plurality of dimensions comprises two or more of the following:

[0095] Azimuth,

[0096] Elevation angle,

[0097] Doppler, or

[0098] scope.

[0099] 3. A method according to clause 1 or 2, wherein:

[0100] Determining a median of the values ​​of the input samples in at least one dimension includes determining a median of each of a plurality of subsets within the input samples in at least one dimension; and

[0101] For each of the multiple dimensions:

[0102] Determining the filtering threshold for each dimension includes determining a plurality of filtering thresholds including a filtering threshold for each subset of the plurality of subsets; and

[0103] Excluding one or more values ​​of the input samples in each dimension from the compressed radar data set includes excluding one or more values ​​of the input samples that are less than at least one filter threshold of the plurality of filter thresholds.

[0104] 4. A method as in any of clauses 1-3, wherein the multiple subsets within the input samples in at least one dimension are mutually exclusive.

[0105] 5. The method of any of clauses 1-4, further comprising storing the compressed radar data set on a plurality of memory banks, wherein at least two memory banks of the plurality of memory banks are accessible in parallel.

[0106] 6. The method of any of clauses 1-5, further comprising storing the compressed radar data set in a memory bank accessible via one or more indirect access tables.

[0107] 7. The method of any of clauses 1-6, wherein determining the median comprises performing a binary search using a comparator array.

[0108] 8. The method of clause 7, wherein performing a binary search using a comparator array comprises:

[0109] iteratively comparing the value of the input sample to the search value using the comparator array; and

[0110] For each iteration:

[0111] compares the number of input samples that have values ​​greater than the search value to an ordered statistical order, and

[0112] Changes the search value based on the comparison.

[0113] 9. A method according to clauses 1 to 8, wherein:

[0114] The multiple dimensions include four dimensions; and

[0115] Performing peak detection on the compressed radar data set includes identifying peaks in at least three of the plurality of dimensions.

[0116] 10. A device for compressing radar data, the device comprising:

[0117] Memory; and

[0118] One or more processing units communicatively coupled to the memory and configured to:

[0119] obtaining input samples comprising values ​​of a plurality of dimensions measured by the radar;

[0120] determining a median of the values ​​of the input samples in at least one of a plurality of dimensions;

[0121] Filtering the input samples based on a median value to create a compressed radar dataset, wherein the filtering comprises, for each of a plurality of dimensions:

[0122] Determine the filtering thresholds for each dimension; and

[0123] Excluding from the compressed radar data set one or more values ​​of input samples for each dimension that are less than a filtering threshold above a median value; and

[0124] Peak detection is performed on the compressed radar data set to generate a list of one or more identified targets.

[0125] 11. The apparatus of clause 10, wherein the plurality of dimensions comprises two or more of:

[0126] Azimuth,

[0127] Elevation angle,

[0128] Doppler, or

[0129] scope.

[0130] 12. The apparatus of clause 10 or 11, wherein the one or more processing units are configured to:

[0131] determining a median value of the values ​​of the input samples in at least one dimension at least in part by determining a median value for each of a plurality of subsets within the input samples in at least one dimension; and

[0132] For each of the multiple dimensions:

[0133] determining a filter threshold for each dimension at least in part by determining a plurality of filter thresholds including a filter threshold for each subset of the plurality of subsets;

[0134] One or more values ​​of the input samples in each dimension are excluded from the compressed radar data set at least in part by excluding one or more values ​​of the input samples that are less than at least one filter threshold of the plurality of filter thresholds.

[0135] 13. An apparatus as in any of clauses 10-12, wherein the plurality of subsets within the input samples of at least one dimension are mutually exclusive.

[0136] 14. The apparatus of any of clauses 10-13, wherein the one or more processing units are configured to store the compressed radar data set on a plurality of memory banks, wherein at least two of the plurality of memory banks are accessible in parallel.

[0137] 15. The apparatus of any of clauses 10-14, wherein the one or more processing units are configured to store the compressed radar data set in a memory bank accessible via one or more indirect access tables.

[0138] 16. The apparatus of any of clauses 10-15, wherein the one or more processing units are configured to determine the median value based on a result of a binary search performed by the comparator array.

[0139] 17. The apparatus of clause 16, wherein the apparatus is configured to perform a binary search using the comparator array as follows:

[0140] iteratively comparing the value of the input sample to the search value using the comparator array; and

[0141] For each iteration:

[0142] compares the number of input samples that have values ​​greater than the search value to an ordered statistical order, and

[0143] Changes the search value based on the comparison.

[0144] 18. A device as described in any of clauses 10 to 17, wherein:

[0145] The multiple dimensions include four dimensions; and

[0146] The one or more processing units are configured to perform peak detection on the compressed radar data set based on identifying peaks in at least three of the plurality of dimensions.

[0147] 19. A device comprising:

[0148] means for obtaining input samples comprising values ​​of a plurality of dimensions measured by the radar;

[0149] means for determining a median of the values ​​of the input samples in at least one of a plurality of dimensions;

[0150] Means for filtering input samples based on a median value to create a compressed radar data set, wherein the filtering comprises, for each of a plurality of dimensions:

[0151] Determine the filtering thresholds for each dimension; and

[0152] excluding from the compressed radar data set one or more values ​​of input samples for each dimension that are less than a filtering threshold above the median value; and

[0153] Means for performing peak detection on a compressed radar data set to generate a list of one or more identified targets.

[0154] 20. The apparatus of clause 19, wherein the plurality of dimensions comprises two or more of:

[0155] Azimuth,

[0156] Elevation angle,

[0157] Doppler, or

[0158] scope.

[0159] 21. Apparatus as claimed in clause 19 or 20, wherein:

[0160] means for determining a median of values ​​of the input samples of at least one dimension, comprising means for determining a median of each of a plurality of subsets within the input samples of at least one dimension; and

[0161] A component that performs the following operations for each of multiple dimensions:

[0162] Determining a filter threshold for each dimension based on determining a plurality of filter thresholds including a filter threshold for each subset of the plurality of subsets;

[0163] Based on the exclusion of one or more values ​​of the input samples that are less than at least one of a plurality of filter thresholds, one or more values ​​of the input samples in each dimension are excluded from the compressed radar data set.

[0164] 22. The apparatus of clause 21, wherein the plurality of subsets within the input samples in at least one dimension are mutually exclusive.

[0165] 23. The apparatus of any of clauses 19-22, further comprising means for storing the compressed radar data set on a plurality of memory banks, wherein at least two memory banks of the plurality of memory banks are accessible in parallel.

[0166] 24. The apparatus of any of clauses 19-23, further comprising means for storing the compressed radar data set in a memory bank accessible via one or more indirect access tables.

[0167] 25. A device as described in any of clauses 19 to 24, wherein:

[0168] The multiple dimensions include four dimensions; and

[0169] Performing peak detection on the compressed radar data set includes identifying peaks in at least three of the plurality of dimensions.

[0170] 26. A non-transitory computer readable medium having instructions embedded therein for compressing radar data, wherein the instructions, when executed by one or more processing units, cause the one or more processing units to perform one or more operations, the one or more operations comprising:

[0171] obtaining input samples comprising values ​​of a plurality of dimensions measured by the radar;

[0172] determining a median of the values ​​of the input samples in at least one of a plurality of dimensions;

[0173] Filtering the input samples based on a median value to create a compressed radar dataset, wherein the filtering comprises, for each of a plurality of dimensions:

[0174] Determine the filtering thresholds for each dimension; and

[0175] excluding from the compressed radar data set one or more values ​​of input samples for each dimension that are less than a filtering threshold above the median value; and

[0176] Peak detection is performed on the compressed radar data set to generate a list of one or more identified targets.

[0177] 27. The non-transitory computer-readable medium of clause 26, wherein the plurality of dimensions comprises two or more of:

[0178] Azimuth,

[0179] Elevation angle,

[0180] Doppler, or

[0181] scope.

[0182] 28. A non-transitory computer readable medium as defined in clause 26 or 27, wherein:

[0183] Determining a median of the values ​​of the input samples in at least one dimension includes determining a median of each of a plurality of subsets within the input samples in at least one dimension; and

[0184] For each of the multiple dimensions:

[0185] Determining the filter thresholds for each dimension includes determining a plurality of filter thresholds including a filter threshold for each subset of the plurality of subsets;

[0186] Excluding one or more values ​​of the input samples in each dimension from the compressed radar data set includes excluding one or more values ​​of the input samples that are less than at least one filter threshold of the plurality of filter thresholds.

[0187] 29. A non-transitory computer-readable medium as in any of clauses 26-28, wherein the instructions, when executed by one or more processing units, further cause the one or more processing units to perform one or more operations including storing the compressed radar data set on a plurality of memory banks, wherein at least two of the plurality of memory banks are accessible in parallel.

[0188] 30. The non-transitory computer-readable medium of any of clauses 26-29, wherein the instructions, when executed by one or more processing units, further cause the one or more processing units to perform one or more operations including storing the compressed radar data set in a memory bank accessible via one or more indirect access tables.

Claims

1. A method for compressing radar data, the method comprising: obtaining input samples comprising values ​​of a plurality of dimensions measured by the radar; determining a median of the values ​​of the input samples in at least one of the plurality of dimensions; Filtering input samples based on the median value to create a compressed radar dataset, wherein the filtering comprises, for each dimension of the plurality of dimensions: Determine the filtering thresholds for each dimension; and excluding from the compressed radar data set one or more values ​​of input samples for each dimension that are less than a filter threshold above the median value; and Peak detection is performed on the compressed radar data set to generate a list of one or more identified targets.

2. The method of claim 1, wherein: The multiple dimensions include two or more of the following: Azimuth, Elevation angle, Doppler, or scope.

3. The method of claim 1, wherein: Determining a median of the values ​​of the input samples in at least one dimension includes determining a median of each of a plurality of subsets within the input samples in at least one dimension; as well as For each dimension in the multiple dimensions: Determining the filter thresholds for each dimension includes determining a plurality of filter thresholds including the filter threshold for each subset of the plurality of subsets; as well as Excluding one or more values ​​of the input samples in each dimension from the compressed radar data set includes excluding one or more values ​​of the input samples that are less than at least one filter threshold of the plurality of filter thresholds.

4. The method of claim 3, wherein: The multiple subsets within the input samples of the at least one dimension are mutually exclusive. 5 . The method of claim 1 , further comprising storing the compressed radar data set on a plurality of memory banks, wherein at least two of the plurality of memory banks are accessible in parallel.

6. The method of claim 1 , further comprising storing the compressed radar data set in a memory bank accessible via one or more indirect access tables.

7. The method of claim 1, wherein: Determining the median value includes performing a binary search using an array of comparators.

8. The method of claim 7, wherein: Using a comparator array to perform a binary search involves: iteratively comparing the value of the input sample to a search value using the comparator array; and For each iteration: comparing the number of input samples having values ​​greater than the search value to an ordered statistical order, and The search value is changed based on the comparison.

9. The method of claim 1, wherein: The plurality of dimensions comprises four dimensions; and Performing peak detection on the compressed radar data set includes identifying peaks in at least three of the plurality of dimensions.

10. A device for compressing radar data, the device comprising: Memory; as well as One or more processing units, communicatively coupled to the memory and configured to: obtaining input samples comprising values ​​of a plurality of dimensions measured by the radar; determining a median of the values ​​of the input samples in at least one of the plurality of dimensions; Filtering input samples based on the median value to create a compressed radar dataset, wherein the filtering comprises, for each dimension of the plurality of dimensions: Determine the filtering thresholds for each dimension; and excluding from the compressed radar data set one or more values ​​of input samples for each dimension that are less than a filter threshold above the median value; and Peak detection is performed on the compressed radar data set to generate a list of one or more identified targets.

11. The device of claim 10, wherein: The multiple dimensions include two or more of the following: Azimuth, Elevation angle, Doppler, or scope.

12. The device of claim 10, wherein: The one or more processing units are configured to: determining a median value of the values ​​of the input samples in the at least one dimension at least in part by determining a median value of each of a plurality of subsets within the input samples in the at least one dimension; as well as For each dimension in the multiple dimensions: determining a filter threshold for each dimension at least in part by determining a plurality of filter thresholds including a filter threshold for each subset of the plurality of subsets; One or more values ​​of the input samples in each dimension are excluded from the compressed radar data set at least in part by excluding one or more values ​​of the input samples that are less than at least one filter threshold of the plurality of filter thresholds.

13. The apparatus of claim 12, wherein: The multiple subsets within the input samples of the at least one dimension are mutually exclusive.

14. The apparatus of claim 10, wherein: The one or more processing units are configured to store the compressed radar data set on a plurality of memory banks, wherein at least two memory banks of the plurality of memory banks are accessible in parallel.

15. The apparatus of claim 10, wherein: The one or more processing units are configured to store the compressed radar data set in a memory bank accessible via one or more indirect access tables.

16. The apparatus of claim 10, wherein: The one or more processing units are configured to determine a median value based on a result of a binary search performed by the comparator array.

17. The apparatus of claim 16, wherein: The device is configured to perform a binary search using a comparator array to: iteratively comparing the value of the input sample to a search value using the comparator array; and For each iteration: comparing the number of input samples having values ​​greater than the search value to an ordered statistical order, and The search value is changed based on the comparison.

18. The apparatus of claim 10, wherein: The plurality of dimensions comprises four dimensions; and The one or more processing units are configured to perform peak detection on the compressed radar data set based on identifying peaks in at least three of the plurality of dimensions.

19. An apparatus comprising: means for obtaining input samples comprising values ​​of a plurality of dimensions measured by the radar; means for determining a median of the values ​​of the input samples in at least one of the plurality of dimensions; means for filtering input samples based on the median value to create a compressed radar data set, wherein the filtering comprises, for each dimension of the plurality of dimensions: Determine the filtering thresholds for each dimension; and excluding from the compressed radar data set one or more values ​​of input samples for each dimension that are less than a filter threshold above the median value; and Means for performing peak detection on the compressed radar data set to generate a list of one or more identified targets.

20. The apparatus of claim 19, wherein: The multiple dimensions include two or more of the following: Azimuth, Elevation angle, Doppler, or scope.

21. The apparatus of claim 19, wherein: The means for determining a median of the values ​​of the input samples of at least one dimension comprises means for determining a median of each of a plurality of subsets within the input samples of at least one dimension; as well as Means for performing the following operations for each dimension of the plurality of dimensions: determining a filter threshold for each dimension based on determining a plurality of filter thresholds including a filter threshold for each subset of the plurality of subsets; Based on the exclusion of one or more values ​​of the input samples that are less than at least one of the plurality of filter thresholds, one or more values ​​of the input samples in each dimension are excluded from the compressed radar data set.

22. The apparatus of claim 21, wherein: The multiple subsets within the input samples of the at least one dimension are mutually exclusive.

23. The apparatus of claim 19, further comprising means for storing the compressed radar data set on a plurality of memory banks, wherein at least two of the plurality of memory banks are accessible in parallel.

24. The apparatus of claim 19, further comprising means for storing the compressed radar data set in a memory bank accessible via one or more indirect access tables.

25. The apparatus of claim 19, wherein: The plurality of dimensions comprises four dimensions; and Performing peak detection on the compressed radar data set includes identifying peaks in at least three of the plurality of dimensions.

26. A non-transitory computer readable medium having embedded therein instructions for compressing radar data, wherein: When the instruction is executed by one or more processing units, the one or more processing units are caused to perform one or more operations, the one or more operations comprising: obtaining input samples comprising values ​​of a plurality of dimensions measured by the radar; determining a median of the values ​​of the input samples in at least one of the plurality of dimensions; Filtering input samples based on the median value to create a compressed radar dataset, wherein the filtering comprises, for each dimension of the plurality of dimensions: Determine the filtering thresholds for each dimension; and excluding from the compressed radar data set one or more values ​​of input samples for each dimension that are less than a filter threshold above the median value; and Peak detection is performed on the compressed radar data set to generate a list of one or more identified targets.

27. The non-transitory computer readable medium of claim 26, wherein: The multiple dimensions include two or more of the following: Azimuth, Elevation angle, Doppler, or scope.

28. The non-transitory computer readable medium of claim 26, wherein: Determining a median of the values ​​of the input samples in at least one dimension includes determining a median of each of a plurality of subsets within the input samples in at least one dimension; as well as For each dimension in the multiple dimensions: Determining the filter thresholds for each dimension includes determining a plurality of filter thresholds including the filter threshold for each subset of the plurality of subsets; Excluding one or more values ​​of the input samples in each dimension from the compressed radar data set includes excluding one or more values ​​of the input samples that are less than at least one filter threshold of the plurality of filter thresholds.

29. The non-transitory computer readable medium of claim 26, wherein: The instructions, when executed by one or more processing units, further cause the one or more processing units to perform one or more operations including storing the compressed radar data set on a plurality of memory banks, wherein at least two of the plurality of memory banks are accessible in parallel.

30. The non-transitory computer readable medium of claim 26, wherein: The instructions, when executed by one or more processing units, further cause the one or more processing units to perform one or more operations including storing the compressed radar data set in a memory bank accessible via one or more indirect access tables.

Images