Intelligent scooter blind area monitoring method and system based on millimeter wave radar

Abstract

The present application relates to the technical field of blind area monitoring, and particularly relates to a smart scooter blind area monitoring method and system based on a millimeter wave radar, which comprises the following steps: obtaining target information detected by the radar every set time, and simultaneously obtaining the moving distance and rotation angle of the scooter between the current frame acquisition time point and the previous frame acquisition time point; calculating the predicted positions of each target in the previous frame acquisition according to the moving distance and rotation angle of the scooter; comparing each target in the current frame and the previous frame, and classifying each target in the current frame as a fixed target or a moving target according to the comparison result; matching the moving targets in the current frame and the previous frame, and screening out threat moving targets that affect the safety of the scooter; calculating the motion state of the threat moving targets, and combining the positions of the threat moving targets to calculate the threat scores of the threat moving targets, and outputting warning signals of different levels according to the threat score sizes. The present application can effectively improve the accuracy of radar monitoring.

Description

Technical Field

[0001] This invention relates to the field of blind spot monitoring technology, and in particular to a method and system for monitoring blind spots of intelligent scooters based on millimeter-wave radar. Background Technology

[0002] Millimeter-wave radar uses ultra-high-frequency millimeter-wave signals. It is small, lightweight, and has high spatial resolution. Leveraging the advantages of millimeter-wave signals, they can penetrate fog, smoke, and dust, allowing operation at night or in darkness, providing excellent all-weather performance. Therefore, it is suitable for monitoring blind spots behind vehicles. In most operating environments, millimeter-wave radar is placed at a relatively high height, making the identification of dangerous targets relatively simple. However, because smart scooters are installed at a low position, the millimeter-wave radar is very close to the ground. This results in many ground targets being easily detected by the radar, causing fixed and moving targets to be mixed together, which seriously affects the accuracy of millimeter-wave radar in detecting threats. Summary of the Invention

[0003] This invention provides a method and system for monitoring blind spots of intelligent scooters based on millimeter-wave radar, which can effectively solve the problems in the background art.

[0004] This invention provides a method for monitoring blind spots of intelligent scooters based on millimeter-wave radar, comprising the following steps:

[0005] The radar detects target information at set intervals, including the target's position and reflection intensity, and also acquires the scooter's movement distance and rotation angle between the acquisition time point in the current frame and the acquisition time point in the previous frame.

[0006] Based on the scooter's travel distance and rotation angle, calculate the predicted positions of each target acquired in the previous frame;

[0007] The positions of each target acquired in this frame are compared with those acquired in the previous frame. Based on the comparison results, each target acquired in this frame is classified as a fixed target or a moving target.

[0008] Match the moving targets in this frame with those in the previous frame to filter out threatening moving targets that could affect the safety of the scooter;

[0009] The system calculates the movement state of the moving targets and, based on their positions, calculates their threat scores. It then outputs warning signals of different levels according to the threat scores.

[0010] Furthermore, when acquiring target information detected by the radar in each frame, an intensity threshold is set, and targets with a reflection intensity lower than the set intensity threshold are deleted as clutter.

[0011] Furthermore, after removing clutter, each detected target is connected to the radar. If multiple targets are collinear, the one furthest from the radar is taken as the detected target, and the remaining targets on the line are removed as clutter.

[0012] Furthermore, the predicted positions of each target acquired in the previous frame are calculated as follows:

[0013] Set the radar coordinates, with the radar position as the origin, the x-axis pointing to the front of the vehicle, and the y-axis pointing to the left side of the vehicle.

[0014] Let this frame be the t-th frame where the radar detects the target information. Then, between the two acquisition time points of the t-th frame and the (t-1)-th frame, the scooter's moving distance ΔL and rotation angle Δθ are:

[0015] Let x(t-1,i) be the x-coordinate of the i-th target acquired in frame t-1, and y(t-1,i) be the y-coordinate. Then the x-coordinate xpre(t-1,i) and ypre(t-1,i) of the predicted position of the i-th target are:

[0016] xpre(t-1,i)=x(t-1,i)·cos(△θ)-y(t-1,i)·sin(△θ)-△L;

[0017] ypre(t-1,i)=x(t-1,i)·sin(△θ)+y(t-1,i)·cos(△θ).

[0018] Furthermore, the classification of each target acquired in this frame into fixed targets or moving targets is as follows:

[0019] Among all targets in frame t, targets with x-coordinates between [-ΔL, 0] are considered fixed targets;

[0020] Matching is performed using a fixed target from frame t-1:

[0021] For each fixed target in frame t-1, take its predicted position as the center, find the target closest to the center among all targets acquired in frame t, and calculate whether the distance between the center and the target exceeds a threshold. If it does not exceed the threshold, then the target in frame t is determined to be a fixed target.

[0022] The other targets obtained in the t-th frame after removing the fixed targets are the moving targets.

[0023] Furthermore, matching moving targets specifically involves:

[0024] Multiple search areas are simultaneously formed with radius R centered on each moving target in frame t; the radius of each search area starts from 0 and then gradually increases.

[0025] When the search area of ​​a moving target covers the predicted position of the moving target in frame t-1, the moving target in frame t is considered to match the moving target in frame t-1, and the search area corresponding to the moving target in frame t is deleted.

[0026] The process continues until the radius R increases to a set threshold.

[0027] Connect all matching moving targets in frame t and frame t-1. Determine if any line segments intersect. If any line segments intersect, swap the matching results of moving targets in frame t that belong to intersecting line segments.

[0028] Furthermore, the specific moving targets that pose a threat to scooter safety are identified as follows:

[0029] For all mutually matching moving targets in frame t and frame (t-1), calculate the distance between the predicted positions of the moving targets in frame t and frame (t-1), and denote the moving targets in frame t with a distance greater than ΔL as threatening moving targets.

[0030] Furthermore, the threat score is calculated as follows:

[0031] For the nth pair of successfully matched moving targets in the current frame and the previous frame, calculate the corresponding velocity magnitude as Vn, and calculate the distance between the moving target in the current frame and the radar as dn;

[0032] Calculate the distance threat factor: Fdisn = 1 - dn / Dmax;

[0033] Where Dmax is the radar's maximum monitoring distance;

[0034] Calculate the speed threat factor: Fspen=min(Vn / △V,1);

[0035] Wherein, △V is the average velocity between the current frame and the previous frame, with a minimum value of 1;

[0036] Calculate the directional threat factor: Fdirn = B / Hn;

[0037] Where B is the set safe width;

[0038] Hn is the distance from the origin to the straight line containing the two successfully matched moving targets of the nth pair;

[0039] Calculate the threat score for the nth moving target:

[0040] Fmovn=0.5Fdisn+0.2Fspen+0.1Fdirn.

[0041] Furthermore, a danger zone centered on the radar is established within the radar's monitoring range, and different levels of warning signals are output based on the distance between the target in the danger zone and the radar.

[0042] The present invention also provides a blind spot monitoring system for intelligent scooters based on millimeter-wave radar, comprising a storage device and a processor. The storage device is used to store one or more program instructions; the processor is used to run one or more program instructions to perform the steps of the above-described blind spot monitoring method for intelligent scooters based on millimeter-wave radar.

[0043] The technical solution of this invention can achieve the following technical effects:

[0044] This method uses the scooter's own motion information to filter and match millimeter-wave radar data. It can quickly classify and identify fixed and moving targets from radar monitoring results when the scooter chassis is low and the radar is close to the ground. It also matches the detection results of each frame to quickly and accurately assess the threat level of moving targets in the blind zone, effectively improving the radar's monitoring accuracy.

[0045] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 This is a flowchart illustrating a method for monitoring blind spots in smart scooters based on millimeter-wave radar.

[0048] Figure 2 A schematic diagram illustrating the principle of matching moving targets between two frames. Detailed Implementation

[0049] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0050] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0051] This invention relates to a method for monitoring blind spots of intelligent scooters based on millimeter-wave radar, such as... Figure 1 As shown, the process includes three main steps: data acquisition, data filtering and matching, and data evaluation. The specific details of each step are as follows:

[0052] Data Acquisition: This step is mainly used to acquire the basic data for subsequent calculations, including radar data and scooter data. The radar will emit millimeter waves into the monitoring area at regular intervals, and then determine the position of objects in the monitoring area based on the reflected signals. When collecting radar data, the target information detected by the radar is acquired at set intervals, including the position and reflection intensity of the target. Each acquisition is recorded as a frame, and each frame will correspond to an uncertain number of target information.

[0053] While acquiring target information for each frame, the system also acquires the scooter's distance traveled and rotation angle between the acquisition time of the current frame and the acquisition time of the previous frame. The distance traveled is obtained by directly measuring the wheel's movement using a wheel speed sensor and then converting the data. The rotation angle is obtained by measuring the distance using an inertial measurement unit (such as a gyroscope) and then converting the data, ensuring the accuracy of the scooter's data.

[0054] Data filtering and matching: In data acquisition, each frame contains a lot of target information. Since the scooter may be in motion, the monitored targets may be in motion or stationary, and the number is uncertain, the target information in adjacent frames cannot be simply matched. The algorithm in this step needs to distinguish between fixed targets and moving targets and match the target information in two frames in order to calculate the motion state of the target.

[0055] Firstly, the movement of the scooter will cause all target positions within the monitoring area to move uniformly in the radar coordinate system. Based on the scooter's movement distance and rotation angle, the predicted position of each target acquired in the previous frame is calculated. This predicted position is the position in the current frame if each target in the previous frame has not moved.

[0056] The positions of each target acquired in this frame are compared with those acquired in the previous frame. Based on the comparison results, each target acquired in this frame is classified. If the predicted position of a target in the previous frame coincides with the target position in this frame, then this target is very likely to be a fixed target; if they do not coincide, then it is very likely to be a moving target.

[0057] The current frame is matched with the moving targets in the previous frame. Fixed targets can be matched directly based on overlap. After the fixed targets are matched, a large amount of target information can be eliminated. The remaining targets that cannot overlap will be the moving targets.

[0058] Data Evaluation: This step is mainly used to evaluate the previously screened moving targets. After identifying the moving targets, since there are many moving targets in the scooter usage scenarios, such as pedestrians, electric vehicles, and cars, the number of targets that actually pose a safety threat to the scooter is relatively small. If a complete calculation and evaluation is performed on each moving target, it would waste a lot of computing power. Therefore, it is necessary to first screen out the threatening moving targets that directly affect the safety of the scooter from all the moving targets, and then evaluate the status of the threatening moving targets.

[0059] During the assessment, the motion state of the moving threat target is calculated, and the threat score T of each moving threat target is calculated based on the position of each moving threat target. Different levels of warning signals are output according to the magnitude of the threat score T.

[0060] High threat level (T≥0.6) triggers a level 2 warning, providing the user with a strong visual and vibration alert;

[0061] Medium threat level (0.3 ≤ T < 0.6) triggers a Level 1 warning, providing the user with a gentle visual cue;

[0062] Low threat (T < 0.3), record, no alarm.

[0063] When an alarm is triggered, the target can be divided into left-side targets (y>0) and right-side targets (y<0) based on the sign of its y-coordinate in the radar coordinate system. Then, different levels of warnings are triggered on the left and right sides according to the degree of threat on the left and right sides, thereby helping users to better avoid threats.

[0064] Millimeter-wave radar is extremely sensitive to the detection of metal targets. When the radar approaches the ground, many metal targets on the ground (such as screws, metal lights, metal guardrails, etc.) will cause the radar to react and form clutter in the radar's monitoring results. For smaller metal parts, their reflection ability is limited, that is, the strength of the signal transmitted back will not be very high. Therefore, when acquiring the target information detected by the radar in each frame, an intensity threshold can be set, and targets with a reflection intensity lower than the set intensity threshold can be deleted as clutter, thereby eliminating the interference of small metal parts.

[0065] After initial clutter removal, it's necessary to further separate larger metal objects or other clutter. At this point, each detected target can be connected to the radar. Generally, targets that pose a threat to scooter safety are relatively large and will block the millimeter waves emitted by the low-lying radar. According to radar monitoring results, after connecting each target to the radar, there shouldn't be multiple targets collinear. If multiple targets are collinear, it physically indicates that the radar's millimeter waves can penetrate some targets, which is obviously unrealistic. It's more likely that clutter is generated by metal objects on the ground (such as metal manhole covers, iron plates, etc.). In this case, the target farthest from the radar is taken as the detected target, and the remaining targets on the line are removed as clutter.

[0066] The specific steps for calculating the predicted positions of each target acquired in the previous frame are as follows:

[0067] Set the radar coordinates with the radar position as the origin, the x-axis pointing to the front of the vehicle, and the y-axis pointing to the left side of the vehicle.

[0068] Let this frame be the t-th frame where the radar detects the target information. Then, between the two acquisition time points of the t-th frame and the (t-1)-th frame, the scooter's moving distance ΔL and rotation angle Δθ are:

[0069] Let x(t-1,i) be the x-coordinate of the i-th target acquired in frame t-1, and y(t-1,i) be the y-coordinate. Then the x-coordinate xpre(t-1,i) and ypre(t-1,i) of the predicted position of the i-th target are:

[0070] xpre(t-1,i)=x(t-1,i)·cos(△θ)-y(t-1,i)·sin(△θ)-△L;

[0071] ypre(t-1,i)=x(t-1,i)·sin(△θ)+y(t-1,i)·cos(△θ).

[0072] After acquiring target information detected by the radar each time, it is necessary to classify each target acquired in this frame into either stationary or moving targets, specifically as follows:

[0073] Let this frame be the t-th frame from which the radar detects the target information.

[0074] First, for the targets detected in frame t within the range [-ΔL, 0], these are targets newly entered into the monitoring area by the scooter due to its movement. These targets will not have corresponding matches in frame t-1, so they need to be defined first. Since the radar's monitoring range is typically 10-25 meters, and the time difference between two frames is usually only about 0.2 seconds, it's unlikely that an object of such high speed would directly enter the scooter's [-ΔL, 0] area from outside the radar's monitoring area. Therefore, targets with x-coordinates between [-ΔL, 0] in frame t can be considered fixed targets in frame t. After determining the target classification for a portion of the targets in frame t, the targets already marked as fixed targets can be removed and no longer need to participate in subsequent calculations.

[0075] Then, the fixed target in frame t-1 is used for matching:

[0076] For each fixed target in frame t-1, with its predicted position as the center, find the target closest to the center among all targets acquired in frame t, and calculate whether the distance between the center (i.e., the predicted position) and the target exceeds a set threshold. If it does not exceed the threshold, then the target in frame t is determined to be a fixed target.

[0077] After removing the fixed targets mentioned above, the remaining targets acquired in frame t are the moving targets. It should be noted that, due to reasons such as objects from outside the monitoring area entering the monitoring area from behind or objects moving out of the monitoring area between consecutive frames, the number of moving targets in frame t and frame t-1 may not be exactly the same. However, no additional processing is required for this discrepancy; simply record the aforementioned judgment results accurately.

[0078] Matching moving targets in two frames (frame t and frame t-1) specifically involves:

[0079] Because the interval between two frames is relatively short, there are not many objects that can move a large distance in such a short time, and the objects that can move so fast are also relatively large (usually motor vehicles). Slow-moving objects such as pedestrians are easily identified as stationary targets. Therefore, after removing stationary targets, there is usually only one moving target left, which is relatively easy to match.

[0080] If there are multiple targets, further algorithms are needed for matching.

[0081] At this point, taking each moving target in frame t as the center and radius R, multiple search areas are simultaneously formed; the radius of each search area starts from 0 and then gradually increases.

[0082] When the search area of ​​a moving target covers the predicted position of the moving target in frame t-1, the moving target in frame t is considered to match the moving target in frame t-1, and the search area corresponding to the moving target in frame t is deleted.

[0083] The process continues until the radius R increases to a set threshold.

[0084] Connect all matching moving targets in frame t and frame t-1. Determine if any line segments intersect. If any line segments intersect, swap the matching results of moving targets in frame t that belong to intersecting line segments.

[0085] The physical principle behind the above search is:

[0086] Within a short time interval between two frames, the same target cannot move too far, and the more moving targets there are, the slower the same target moves. Thus, by searching all moving targets in frame t simultaneously, we can find the most likely location of the moving target in the previous frame. However, once the search area radius R exceeds a certain range, such as 5 meters, if we assume that there is a moving target in the previous frame that can match the moving target in this frame, it means that the target moved more than 5 meters in 0.2 seconds, which is obviously unreasonable when there are so many moving targets. Therefore, the moving target that did not match may be a newly entered moving target from behind. Its threat level can be calculated in subsequent calculations, but it can be ignored in the calculation of this frame.

[0087] After initially determining the correspondence, connect all the moving targets in the t-th and t-1-th frames that match each other. This connecting line segment is the movement path of each object. In the short time between two frames, the movement paths of each object are unlikely to intersect. Therefore, if the line segments intersect, it may be that the target matching between the two frames is incorrect and needs to be corrected.

[0088] The matching process is Figure 2 Taking the hypothetical scenario shown as an example, the (t-1)th frame is the upper part of the image, and the tth frame is the lower part of the image;

[0089] At frame t-1, there are three moving targets in the monitored area: a square, a triangle, and a circle. The bottom of the figure shows the three predicted positions of the square, triangle, and circle, which are represented by the dashed lines.

[0090] In frame t, four moving targets were identified: a square, a triangle, a circle, and a pentagon. The square, triangle, and circle represent the positions of each target after they moved, while the pentagon is a moving target that entered from behind.

[0091] It should be noted that squares, triangles, circles, and pentagons are shapes assigned to targets only to represent actual correspondences. In actual radar monitoring results, each target is a point, and correspondences cannot be matched by shape.

[0092] At the start of the matching process, R is initially set to 0.1 meters. Then, within frame t, circular areas with a radius of 0.1 meters are formed, centered on four moving targets: a square, a triangle, a circle, and a pentagon. Each search area is then checked to see if it covers a target with a dashed line shape (i.e., the predicted position of the moving target in the previous frame). R is then continuously increased, successively to 0.2 meters, 0.3 meters, 0.4 meters, and so on. If, at R = 0.2 meters, the circular search area covers the target with the dashed line shape, then the correspondence between the two moving targets is established, and the circular search area is no longer checked when R is subsequently increased. This continues until R reaches a maximum of 5 meters. If, at this point, the pentagonal target's search area still does not contain a target with the dashed line shape, then the pentagonal target is the newly appearing moving target.

[0093] After all moving targets have been matched, suppose that the movement distance of the square and triangle targets is now unusual, causing the aforementioned matching errors to occur, such as a square target matching a dashed triangle target and a triangle target matching a dashed square target. In this case, if we connect the two lines, we will find that the two line segments intersect, thus identifying the matching error. Then we swap the two intersecting matching results, so that the square target matches the dashed square target and the triangle target matches the dashed triangle target.

[0094] Of course, since the movement of objects is unpredictable, when multiple moving objects are clustered together, there are numerous moving targets in two frames, which can easily lead to matching errors, even though the connection verification may pass despite intersections. However, in such cases, the moving distances of the objects are very short, and the objects are clustered together. Therefore, even if some of the moving targets have matching errors, the calculated movement speeds are very low and will not be considered threatening moving targets, resulting in false alarms. On the other hand, if there are multiple fast-moving objects, their movement distances are long, and the length of the connection formed after matching verification will also be long. In this case, there will be virtually no matching errors but successful verifications. When a matching error occurs, the long connection length will generally form intersections, thus ensuring the accuracy of detecting high-speed moving objects.

[0095] The specific moving targets that pose a threat to scooter safety are:

[0096] For all mutually matching moving targets in frame t and frame (t-1), calculate the distance between the predicted positions of the moving targets in frame t and frame (t-1). This distance can be regarded as the distance the object moves within two frames. If this distance < ΔL, it means that the object moves slower than the scooter and will not pose a threat to the scooter. The moving target in frame t with a distance > ΔL is recorded as a threatening moving target.

[0097] The threat score is calculated as follows:

[0098] For the nth pair of successfully matched moving targets in the current frame and the previous frame, calculate the corresponding velocity magnitude as Vn, and calculate the distance between the moving target in the current frame and the radar as dn;

[0099] Calculate the distance threat factor: Fdisn = 1 - dn / Dmax;

[0100] Where Dmax is the radar's maximum monitoring distance;

[0101] Calculate the speed threat factor: Fspen=min(Vn / △V,1);

[0102] Wherein, △V is the average velocity between the current frame and the previous frame, with a minimum value of 1;

[0103] min(Vn / △V,1) means taking the smaller value between Vn / △V and 1. Since the threat score is calculated only when a threatening moving target has appeared, it means that there is already a certain threat behind. This min algorithm ensures that the calculation can have a basic numerical response to the threat.

[0104] Calculate the directional threat factor: Fdirn = B / Hn;

[0105] Where B is the set safe width, which is the width of the scooter plus the minimum safe distance when a vehicle passes by the scooter;

[0106] Hn is the distance from the origin to the straight line between the two successfully matched moving targets in the nth pair; the smaller Hn is, the more directly the fast-moving target behind is facing the scooter, and the value of Fdirn will increase rapidly.

[0107] The aforementioned threat factors can be used individually for assessment, or they can be combined to calculate a total threat score. The threat score for the nth moving target can then be calculated.

[0108] Fmovn=0.5Fdisn+0.2Fspen+0.1Fdirn;

[0109] The values ​​0.5, 0.2, and 0.1 are weighting coefficients and can be modified according to specific usage requirements.

[0110] In addition to detecting threats from moving objects far from the blind spot, it is also necessary to monitor threats from any target located nearby. In this case, a danger zone centered on the radar can be established within the radar's monitoring range, such as a danger zone with a radius of 1 meter. Different levels of warning signals can be output based on the distance between the target in the danger zone and the radar.

[0111] The present invention also relates to a blind spot monitoring system for intelligent scooters based on millimeter-wave radar, comprising a storage device and a processor, wherein the storage device is used to store one or more program instructions; and the processor is used to run one or more program instructions to perform the steps of the above-described blind spot monitoring method for intelligent scooters based on millimeter-wave radar.

[0112] Similarly, the above-mentioned optimization schemes for the system can also achieve the optimization effects corresponding to the methods in Embodiment 1, which will not be repeated here.

[0113] Although this application has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made thereto without departing from the spirit and scope of this application. Accordingly, this specification and drawings are merely exemplary illustrations of the application as defined herein, and are to be considered as covering any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Thus, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A method for monitoring blind spots of intelligent scooters based on millimeter-wave radar, characterized by the following steps: include: The radar detects target information at set intervals, including the target's position and reflection intensity, and also acquires the scooter's movement distance and rotation angle between the acquisition time point in the current frame and the acquisition time point in the previous frame. Based on the scooter's travel distance and rotation angle, calculate the predicted positions of each target acquired in the previous frame; The positions of each target acquired in this frame are compared with those acquired in the previous frame. Based on the comparison results, each target acquired in this frame is classified as a fixed target or a moving target. Match the moving targets in this frame with those in the previous frame to filter out threatening moving targets that could affect the safety of the scooter; The system calculates the movement state of the moving targets and, based on their positions, calculates their threat scores. It then outputs warning signals of different levels according to the threat scores.

2. The method for monitoring blind spots of intelligent scooters based on millimeter-wave radar according to claim 1, characterized in that, When acquiring target information detected by the radar in each frame, an intensity threshold is set, and targets with a reflection intensity lower than the set intensity threshold are deleted as clutter.

3. The method for monitoring blind spots of intelligent scooters based on millimeter-wave radar according to claim 2, characterized in that, After removing clutter, each detected target is connected to the radar. If multiple targets are collinear, the one furthest from the radar is taken as the detected target, and the remaining targets on the line are removed as clutter.

4. The method for monitoring blind spots of intelligent scooters based on millimeter-wave radar according to claim 1, characterized in that, The specific steps for calculating the predicted positions of each target acquired in the previous frame are as follows: Set the radar coordinates, with the radar position as the origin, the x-axis pointing to the front of the vehicle, and the y-axis pointing to the left side of the vehicle. Let this frame be the t-th frame where the radar detects the target information. Then, between the two acquisition time points of the t-th frame and the (t-1)-th frame, the scooter's moving distance ΔL and rotation angle Δθ are: Let x(t-1,i) be the x-coordinate of the i-th target acquired in frame t-1, and y(t-1,i) be the y-coordinate. Then the x-coordinate xpre(t-1,i) and ypre(t-1,i) of the predicted position of the i-th target are: xpre(t-1,i)=x(t-1,i)·cos(△θ)-y(t-1,i)·sin(△θ)-△L; ypre(t-1,i)=x(t-1,i)·sin(△θ)+y(t-1,i)·cos(△θ).

5. The method for monitoring blind spots of intelligent scooters based on millimeter-wave radar according to claim 4, characterized in that, The targets acquired in this frame are classified into stationary targets or moving targets as follows: Among all targets in frame t, targets with x-coordinates between [-ΔL, 0] are considered fixed targets; Matching is performed using a fixed target from frame t-1: For each fixed target in frame t-1, take its predicted position as the center, find the target closest to the center among all targets acquired in frame t, and calculate whether the distance between the center and the target exceeds a threshold. If it does not exceed the threshold, then the target in frame t is determined to be a fixed target. The other targets obtained in the t-th frame after removing the fixed targets are the moving targets.

6. The method for monitoring blind spots of intelligent scooters based on millimeter-wave radar according to claim 5, characterized in that, Matching moving targets specifically involves: Multiple search areas are simultaneously formed with radius R centered on each moving target in frame t; the radius of each search area starts from 0 and then gradually increases. When the search area of ​​a moving target covers the predicted position of the moving target in frame t-1, the moving target in frame t is considered to match the moving target in frame t-1, and the search area corresponding to the moving target in frame t is deleted. The process continues until the radius R increases to a set threshold. Connect all matching moving targets in frame t and frame t-1. Determine if any line segments intersect. If any line segments intersect, swap the matching results of moving targets in frame t that belong to intersecting line segments.

7. The method for monitoring blind spots of intelligent scooters based on millimeter-wave radar according to claim 6, characterized in that, The specific moving targets that pose a threat to scooter safety are: For all mutually matching moving targets in frame t and frame (t-1), calculate the distance between the predicted positions of the moving targets in frame t and frame (t-1), and denote the moving targets in frame t with a distance greater than ΔL as threatening moving targets.

8. The method for monitoring blind spots of intelligent scooters based on millimeter-wave radar according to claim 1, characterized in that, The threat score is calculated as follows: For the nth pair of successfully matched moving targets in the current frame and the previous frame, calculate the corresponding velocity magnitude as Vn, and calculate the distance between the moving target in the current frame and the radar as dn; Calculate the distance threat factor: Fdisn = 1 - dn / Dmax; Where Dmax is the radar's maximum monitoring distance; Calculate the speed threat factor: Fspen=min(Vn / △V,1); Wherein, △V is the average velocity between the current frame and the previous frame, with a minimum value of 1; Calculate the directional threat factor: Fdirn = B / Hn; Where B is the set safe width; Hn is the distance from the origin to the straight line containing the two successfully matched moving targets of the nth pair; Calculate the threat score for the nth moving target: Fmovn=0.5Fdisn+0.2Fspen+0.1Fdirn.

9. The method for monitoring blind spots of intelligent scooters based on millimeter-wave radar according to claim 1, characterized in that, A danger zone centered on the radar is established within the radar's monitoring range, and different levels of warning signals are output based on the distance between the target in the danger zone and the radar.

10. A blind spot monitoring system for intelligent scooters based on millimeter-wave radar, characterized in that, The device includes a storage unit and a processor, wherein the storage unit is used to store one or more program instructions; and the processor is used to run one or more program instructions to perform the steps of the millimeter-wave radar-based smart scooter blind spot monitoring method as described in any one of claims 1 to 9.

Images