Motion classification using low-level detection

Abstract

A technique and apparatus for implementing motion classification using low-level detection are described. Specifically, a radar system identifies fusion detections associated with an object and determines whether the fusion detections indicate that the object is moving. If it is determined that the object is moving or moving perpendicular to the main vehicle, a current motion counter or a vertical motion counter is incremented, respectively. If either the current motion counter or the vertical motion counter has a value greater than a threshold, the current motion flag and / or the vertical motion flag are set to true. In response to setting either flag to true, the radar system increments a historical motion counter to true. The main vehicle is then operated based on the current motion flag, the vertical motion flag, and the historical motion counter. In this way, the radar system introduces hysteresis to improve the reliability and stability of motion classification.

Description

Background Technology

[0001] Advances in autonomous driving could lead to the widespread use of self-driving cars and autonomous robots in business and daily life. Many autonomous driving technologies rely on information provided by one or more sensors to visualize the surrounding environment. With this information, autonomous driving logic can identify the presence (or absence) of objects and their current position and speed. However, even with this information, determining the appropriate actions to provide a sufficient safety margin to avoid collisions can be challenging for autonomous driving logic.

[0002] Autonomous driving logic can determine appropriate actions based on whether an object is stationary or moving, and on a correct estimate of the object's orientation or heading. In some instances, the time available to determine the appropriate action may be very short, and within this time, sensors may not be able to obtain sufficient observations of the object to determine its action classification and accurately predict future movement. For sensors, rapidly and accurately classifying objects for use in autonomous driving applications can be challenging. Summary of the Invention

[0003] Techniques and apparatus for implementing motion classification using low-level detection are described. Specifically, a radar system mounted on a mobile platform extracts detections associated with objects based on radar data. The radar system identifies fused detections associated with a specific object and determines whether the fused detections indicate that the specific object is moving. In response to determining that the fused detections indicate that the specific object is moving, the radar system increments the current motion counter for the specific object. In response to determining that the fused detections indicate that the specific object is moving perpendicular to the main vehicle, the radar system also increments the vertical motion counter. In response to determining that either the current motion counter or the vertical motion counter has a value greater than a threshold, the radar system then sets the current motion flag and / or the vertical motion flag to true. In response to setting either flag to true, the radar system increments the historical motion counter for the specific object to true. The main vehicle is then operated based on the current motion flag, the vertical motion flag, and the historical motion counter. In this way, the radar system introduces hysteresis in its motion classification to provide reliable and stable motion classification output to downstream vehicle-based systems. The described radar system can also introduce a flag reset mechanism to prevent false positives and erroneous vehicle operation responses.

[0004] This document also describes methods performed by the radar systems summarized above and other configurations of the radar systems described herein, as well as the apparatus for performing these methods.

[0005] This invention provides a simplified concept related to motion classification using low-level detection, which is further described in the detailed description and accompanying drawings. This invention is not intended to identify essential features of the claimed subject matter, nor is it intended to define the scope of the claimed subject matter. Attached Figure Description

[0006] This document describes in detail one or more aspects of motion classification using low-level detection with reference to the following figures. The same numbers are used throughout the figures to refer to similar features and components:

[0007] Figure 1 An example environment is shown in which a radar system capable of performing motion classification using low-level detection can be implemented according to the technology of this disclosure;

[0008] Figure 2 An example configuration of a vehicle is shown, featuring a radar system that performs motion classification using low-level detection.

[0009] Figure 3 An example method for performing motion classification using low-level detection is shown;

[0010] Figure 4-1 and Figure 4-2 An example concept diagram for performing motion classification using low-level detection is shown; and

[0011] Figure 5-1 and Figure 5-2 A conceptual diagram is shown in which motion classification is performed using low-level detection. Detailed Implementation

[0012] Overview

[0013] Many driver assistance and autonomous driving applications rely on information from sensors (e.g., radar sensors, cameras) to visualize the surrounding environment. Accurate object classification (e.g., stationary, moving) and motion heading estimation (e.g., intersecting traffic, reversing, oncoming traffic) are essential elements for many driver assistance and autonomous driving applications. These applications typically use multi-sensor architectures (e.g., a combination of radar sensors and cameras) to provide improved motion classification and heading for nearby objects, thereby increasing the confidence level of driving decisions.

[0014] Sensors in a multi-sensor architecture have relative advantages and disadvantages. For example, radar sensors can provide improved performance in the presence of different environmental conditions (such as low lighting and fog, or the presence of moving or overlapping objects). Radar sensors can also provide accurate distance and rate of change estimates of objects in the environment. In contrast, cameras can provide improved object identification (e.g., identifying vulnerable road users (VRUs) such as pedestrians and cyclists)) and azimuth estimation. Specifically, cameras can determine the motion and orientation state of intersecting moving objects with greater accuracy than radar sensors.

[0015] To accurately classify the motion and heading of objects, multi-sensor systems should utilize computationally efficient algorithms with high confidence. Some multi-sensor systems use velocity-based determination to classify objects. For example, these systems can determine whether any trajectory or fused object associated with a target meets threshold criteria and label it accordingly. If it is determined that an object is moving, the system can determine whether the object is moving away from or towards the main vehicle.

[0016] Because sensors can typically detect negligible velocity components associated with stationary objects, these systems may find it difficult to distinguish between slow-moving and stationary objects. Similarly, these systems may struggle to track the motion of objects in urban driving environments, where objects may repeatedly switch between moving and stationary states. These systems may also struggle to correctly classify objects across a range of object types, such as pedestrians and vehicles.

[0017] In contrast, this document describes techniques for more robust and accurate motion classification and heading determination. For example, the described system can directly utilize sensor data to reduce noise output and overprocessing. The system can use camera data to identify object types and dynamically apply corresponding thresholds to classify their motion more accurately. Hysteresis techniques are used to provide stable motion and heading classification, especially for urban environments and slow-moving objects. In this way, the described system provides more reliable and stable motion and heading classification for all types of objects with different headings and orientations. The system also utilizes a reset mechanism to reduce false positives and avoid erroneous responses in assisted driving and autonomous driving applications.

[0018] Example Environment

[0019] Figure 1An example environment 100 is shown in which a radar system 104 capable of performing motion classification using low-level detection can be implemented according to the technology of this disclosure. In the depicted environment 100, the radar system 104 is mounted to or integrated into a vehicle 102. The radar system 104 is capable of detecting one or more objects 112 in the vicinity of the vehicle 102 by emitting and receiving electromagnetic (EM) signals. The vehicle 102 may also include a camera 106 (or vision system) configured to capture visual data relating to the objects 112.

[0020] Although illustrated as a truck, vehicle 102 can represent other types of motorized vehicles (e.g., cars, motorcycles, buses, tractors, semi-trailers, or construction equipment), various types of non-motorized vehicles (e.g., bicycles), various types of rail vehicles (e.g., trains or trams), water vehicles (e.g., boats or ships), aircraft (e.g., airplanes or helicopters), or spacecraft (e.g., satellites). Typically, radar system 104 and camera 106 can be mounted on any type of mobile platform (including mobile machinery or robotic equipment).

[0021] In the depicted implementation, radar system 104 and camera 106 are mounted near the front of vehicle 102, each providing an instrument field of view 110. In other implementations, radar system 104 and camera 106 may be mounted on the rear, left, or right side of vehicle 102. In some cases, vehicle 102 includes multiple radar systems 104 (or cameras 106), such as a first front-facing radar system 104 positioned near the left side of vehicle 102 and a second front-facing radar system 104 positioned near the right side of vehicle 102. Typically, the positions of radar system 104 and camera 106 can be designed to provide a specific field of view 106 containing a region of interest in which object 112 may be present. Example fields of view 110 include 360-degree fields of view, one or more 180-degree fields of view, one or more 90-degree fields of view, etc., which may overlap (e.g., four 120-degree fields of view).

[0022] Radar system 104 may include at least one antenna array and at least one transceiver for transmitting and receiving radar signals. The antenna array includes at least one transmitting antenna element and at least one receiving antenna element. In some cases, the antenna array includes multiple transmitting antenna elements and multiple receiving antenna elements to achieve a multiple-input multiple-output (MIMO) radar capable of transmitting multiple different waveforms at a given time (e.g., each transmitting antenna element transmits a different waveform). The antenna elements may be circularly polarized, horizontally polarized, vertically polarized, or a combination thereof.

[0023] Using an antenna array, radar system 104 can form steered or non-steered beams, as well as wide or narrow beams. Steering and shaping can be achieved through analog beamforming or digital beamforming. The transmitting antenna elements can have, for example, a non-steered omnidirectional radiation mode, or can generate wide, steered beams to illuminate large spatial volumes. To obtain target angular accuracy and angular resolution, receiving antenna elements can be used to generate hundreds of narrow steered beams using digital beamforming. In this way, radar system 104 can effectively monitor the external environment and detect one or more objects 112 within the field of view 110.

[0024] A transceiver includes circuitry and logic for transmitting and receiving radar signals via an antenna array. Components of the transceiver may include amplifiers, mixers, switches, analog-to-digital converters, or filters for conditioning the radar signals. The transceiver also includes logic for performing in-phase / quadrature (I / Q) operations, such as modulation or demodulation. Various modulation methods can be used, including linear frequency modulation, triangular frequency modulation, stepped frequency modulation, or phase modulation. The transceiver can be configured to support continuous wave or pulse radar operation. The spectrum (e.g., frequency range) used by the transceiver to generate the radar signal may be contained in frequencies between 1 MHz and 400 GHz, between 4 GHz and 100 GHz, or between approximately 70 GHz and 80 GHz. The transceiver may employ spread spectrum techniques (such as Code Division Multiple Access (CDMA)) to support MIMO operation.

[0025] Generally, object 112 is made of one or more materials that reflect radar signals. Object 112 can be an inanimate object, such as a four-wheeled vehicle or a two-wheeled vehicle. In other cases, object 112 is a moving object, such as a pedestrian or animal. Object 112 can be moving or stationary. If it is moving, object 112 can move perpendicular to and / or parallel to the travel path of vehicle 102. Other types of object 112 may include continuous or discontinuous road obstacles, traffic cones, concrete barriers, guardrails, fences, or trees.

[0026] Information about the object type, its motion classification, and its heading detected by radar system 104 and / or camera 106 can enable driver assistance or autonomous driving applications of vehicle 102 to determine appropriate actions to avoid a potential collision with object 112. Example actions may include braking to stop, steering, or decelerating. Specifically, the object type may indicate an appropriate speed threshold for classifying the motion of object 112 (e.g., moving, stationary). Based on this information, driver assistance or autonomous driving applications can determine an appropriate safety margin (e.g., appropriate distance and / or speed) to maintain when approaching object 112 to allow sufficient time to avoid a potential collision.

[0027] Vehicle 102 also includes a processor 114 (e.g., a hardware processor, processing unit) and a computer-readable storage medium (CRM) 116 (e.g., a memory, long-term storage device, short-term storage device) storing computer-executable instructions for motion classifier 118 and heading classifier 120. Processor 114 may include multiple processing units or a single processing unit (e.g., a microprocessor). Processor 114 may also be a system-on-a-chip of a computing device, controller, or electronic control unit. Processor 114 executes the computer-executable instructions stored within CRM 116. As an example, processor 114 may execute motion classifier 118 to classify the motion of object 112 as moving or stationary and maintain a history of motion classifications for the object. Similarly, when processor 114 executes heading classifier 120, the heading of object 112 (e.g., crossing traffic, oncoming traffic, reversing traffic) relative to vehicle 102 can be determined. The motion classifier 118 can directly use data from the radar system 104 and the camera 106 to reduce overprocessing of the vehicle 102 and / or the processor 114, as well as noise from motion classification.

[0028] Example System

[0029] Figure 2 An example configuration of a vehicle 102 is shown, featuring a radar system 104 capable of performing motion classification using low-level detection. (See also: Regarding...) Figure 1 As described, vehicle 102 may include radar system 104, camera 106, processor 114, CRM 116, motion classifier 118, and heading classifier 120. Vehicle 102 may also include one or more communication devices 202 and one or more vehicle-based systems 208.

[0030] Communication device 202 may include sensor interfaces and vehicle-based system interfaces. For example, when the motion classifier 118 and heading classifier 120 are integrated as separate components within vehicle 102, the sensor interfaces and vehicle-based system interfaces can transmit data on the communication bus of vehicle 102. Communication device 202 can provide raw or processed sensor data from radar system 104 and camera 106 to motion classifier 118 and heading classifier 120.

[0031] The motion classifier 118 may include a sign tracker 204 and a motion threshold module 206. The sign tracker 204 can provide historical tracking of signs and counters associated with the movement and / or heading of a single object 112. For example, the sign tracker 204 can maintain instantaneous and historical sign data for different motion and heading classifications associated with the object 112 (e.g., moving, backward, movable, oncoming, and / or stationary signs). In this way, the motion classifier 118 can avoid inconsistent or flickering motion classifications for specific objects, especially in urban or congested driving environments, by introducing hysteresis in motion tracking. Furthermore, the described technique reduces the occurrence of false positives when identifying a specific object as moving or stationary.

[0032] The motion thresholding module 206 can use visual data from camera 106 to determine the object type of each object 112. Based on the object type, the motion thresholding module 206 can use different motion thresholds to classify a specific object 112 as moving or stationary. In this way, the motion classifier 118 can adjust the motion threshold based on the type of the detected object and provide improved motion and heading classification for the object 112. For example, the motion thresholding module 206 can use a reduced motion threshold for pedestrians to account for the inherent uncertainties and ambiguities in recognizing the motion patterns of pedestrians and other VRUs.

[0033] Vehicle 102 also includes vehicle-based systems 208 (such as driver assistance system 210 and autonomous driving system 212) that rely on data from motion classifier 118 and / or heading classifier 120 to control the operation of vehicle 102 (e.g., braking, lane changing). Typically, vehicle-based systems 208 can use data provided by motion classifier 118 and / or heading classifier 120 to control the operation of vehicle 102 and perform specific functions. For example, driver assistance system 210 may output information (e.g., messages, sounds, visual indicators) to alert the driver to stationary objects and perform evasive maneuvers to avoid collisions with stationary objects. As another example, autonomous driving system 212 may navigate vehicle 102 to a specific location to avoid collisions with moving objects.

[0034] Example Method

[0035] Figure 3 An example method 300 for performing motion classification using low-level detection is shown. Method 300 is shown as multiple sets of numbered operations (or actions) to be performed, but is not limited to the order or combination of operations shown herein. Furthermore, any one or more operations can be repeated, combined, or rearranged to provide additional functionality. References can be found in the sections discussed below. Figure 1Environment 100 and Figure 1 and Figure 2 The entities detailed herein are for illustrative purposes only. This technique is not limited to being performed by one or more entities.

[0036] At 302, radar signals are acquired from the radar system for one data cycle. The radar signals are reflected by one or more objects in the vicinity of the main vehicle. For example, radar system 104 may be installed in vehicle 102 and provide a field of view 110 of road 108. Radar system 104 transmits and receives radar signals, including radar signals reflected by objects 112 in the environment 100 surrounding vehicle 102.

[0037] At 304, radar signals are used to identify one or more detections associated with each of one or more objects. Detections associated with a specific object among the one or more objects represent fusion detections for that specific object. For example, processor 114 or radar system 104 can use radar signals to determine detections associated with object 112. Processor 114 or radar system 104 can associate detections associated with a specific object 112 as fusion detections for that specific object 112. In this way, processor 114 or radar system 104 can provide fusion detections to motion classifier 118 for each object in object 112.

[0038] At 306, for each fusion detection, it is determined whether the fusion detection indicates that a particular object is moving. For example, motion classifier 118 may determine for each fusion detection whether a particular fusion detection indicates that a particular object 112 associated with that fusion detection is moving. If at least half of the detections associated with a particular object 112 indicate movement, then motion classifier 118 may determine that the particular object 112 is moving. If less than half of the detections associated with a particular object 112 indicate movement, then motion classifier 118 may determine that the motion is ambiguous.

[0039] The motion classifier 118 can also determine for each fusion detection whether the fusion detection indicates that a particular object 112 is moving perpendicular to the vehicle 102 (e.g., having a velocity component orthogonal to the vehicle 102). The motion classifier 118 can determine vertical movement by determining that the orthogonal velocity component associated with the fusion detection is greater than zero.

[0040] Method 300 includes performing operations 306 to 314 for each fusion detection (associated with each object 112) in each data period (or fused data period). Although operations 306 to 314 may be written in singular form for a particular fusion detection or a particular object, it should be understood that operations 306 to 314 are performed for each fusion detection in each data period.

[0041] At 308, in response to determining that the fusion detection indicates a particular object is moving, the continuous motion counter associated with the particular object is incremented. For example, in response to determining that the fusion detection indicates object 112 is moving, motion classifier 118 may increment (e.g., increase the value by 1) the continuous motion counter associated with the particular object 112. Similarly, in response to determining that the movement of object 112 is fuzzy, motion classifier 118 may increment the continuous fuzzy counter associated with the particular object 112. In response to incrementing the continuous fuzzy counter associated with the particular object 112, motion classifier 118 may reset the value of the continuous motion counter associated with the particular object 112 to zero. Similarly, in response to incrementing the continuous motion counter associated with the particular object 112, motion classifier 118 may reset the value of the continuous fuzzy counter associated with the particular object 112 to zero. In response to determining that the fusion detection indicates object 112 is moving perpendicular to vehicle 102, motion classifier 118 may also increment the continuous vertical motion counter associated with the particular object 112.

[0042] At 310, in response to determining that the continuous motion counter associated with a specific object has a value greater than a threshold counter value, the continuous motion flag associated with the specific object is set to true. For example, in response to determining that the value of the continuous motion counter associated with specific object 112 is greater than a threshold counter value (e.g., a value of three), motion classifier 118 may set the continuous motion flag associated with specific object 112 to true. Similarly, in response to determining that the continuous vertical motion counter has a value greater than a threshold counter value, motion classifier 118 may set the continuous vertical motion flag associated with specific object 112 to true.

[0043] At 312, if the continuous motion flag associated with a specific object is set to true, and the fusion detection associated with the specific object has a velocity component greater than a motion threshold, then the movement flag associated with the specific object is set to true. For example, if the continuous motion flag associated with object 112 is set to true, and the fusion detection associated with object 112 has a velocity component greater than a motion threshold (e.g., a longitudinal velocity component), then motion classifier 118 can set the movement flag associated with specific object 112 to true. Motion classifier 118 can determine whether the fusion detection associated with specific object 112 has a longitudinal velocity component greater than a motion threshold by first determining the corresponding motion threshold. Specifically, motion classifier 118 can determine the object type of specific object 112 (e.g., pedestrian, vehicle, cyclist) for each fusion detection using visual data from camera 106. Then, motion classifier 118 can set the motion threshold based on the object type of specific object 112, as per the object type of specific object 112. Figure 4-1and Figure 4-2 A more detailed description.

[0044] Motion classifier 118 can compare the longitudinal velocity component associated with a specific object 112 to the motion threshold. This comparison can be performed by first comparing the radar longitudinal velocity component associated with the specific object 112 to the motion threshold. The radar longitudinal velocity component is obtained from the radar signal. If the radar data is insufficient for this comparison, motion classifier 118 can then compare the visual velocity component associated with the specific object 112 to the motion threshold. The visual longitudinal velocity component is obtained from the visual data. If neither the radar data nor the visual data supports the comparison, motion classifier 118 can use the previous state of the motion marker associated with the specific object 112 determined during the previous data period. To reduce processing requirements, motion classifier 118 can determine whether the fusion detection associated with the specific object 112 has a longitudinal velocity component greater than the clear motion threshold. If the velocity component is greater than the clear motion threshold, motion classifier 118 does not need to compare the radar or visual data with the motion threshold.

[0045] Similarly, if the continuous vertical motion flag associated with object 112 is set to true, and the fusion detection associated with object 112 has an orthogonal velocity component greater than the vertical motion threshold, then motion classifier 118 may set the motion flag associated with a specific object 112 to true. The vertical velocity threshold may be set based on the object type of the specific object 112. Motion classifier 118 may determine that the orthogonal velocity component of the fusion detection based on radar signals is greater than the vertical motion threshold by determining that it is greater than the vertical motion threshold. Alternatively, motion classifier 118 may determine that the visual orthogonal velocity component of the fusion detection based on visual data is greater than the vertical motion threshold. If motion classifier 118 determines that the orthogonal velocity component associated with the fusion detection is not greater than zero (e.g., no vertical motion was detected for that data period), it may decrement the value of the continuous vertical motion counter associated with the specific object 112 by one instead of resetting its value to zero.

[0046] If the movement flag associated with a particular object is set to true, the heading classifier 120 can then determine the heading classification associated with the particular object (e.g., cross, east, west, northeast, etc.) based on the heading associated with the fusion detection. If the heading classification associated with the particular object is a classification indicating vertical movement relative to the vehicle 102, the heading classifier 120 can set the cross flag to true.

[0047] At 314, in response to setting the motion flag associated with a specific object to true, the historical motion counter associated with the specific object is incremented for each fusion detection. For example, if the motion flag of specific object 112 is set to true, the motion classifier 118 may increment the historical motion counter associated with specific object 112. If the motion flag associated with specific object 112 is true, the motion classifier 118 may also set the movable flag associated with specific object 112 to true. The movable flag indicates whether the motion flag associated with specific object 112 has been set to true in the current data period or any previous data period. If the historical motion counter associated with object 112 has a value less than ten, the motion classifier 118 may also set the movable flag associated with specific object 112 to false.

[0048] At point 316, the primary vehicle is operated on the road based on the state of the moving signs and historical motion counters associated with each of one or more objects. For example, driver assistance system 210 or autonomous driving system 212 may operate vehicle 102 based on the state of the moving signs and historical motion counters associated with each of objects 112. Driver assistance system 210 or autonomous driving system 212 may also operate vehicle 102 based on the state of the cross signs and / or movable signs associated with each of objects 112.

[0049] Figure 4-1 and Figure 4-2 A sample concept diagram 400 for performing motion classification using low-level detection is shown. Specifically, concept diagram 400 illustrates the motion classification process of motion classifier 118 and the heading classification process of heading classifier 120. Concept diagram 400 shows example inputs, outputs, and operations of motion classifier 118 and heading classifier 120, but motion classifier 118 and heading classifier 120 are not necessarily limited to the order or combination of inputs, outputs, and operations shown herein. Furthermore, any one or more operations can be repeated, combined, or recombined to provide additional functionality.

[0050] At operation 404, motion classifier 118 determines the motion attributes of the object 112 associated with the fusion detection. Motion classifier 118 receives radar data 402 from one or more radar systems 104. Radar data 402 may be at the detection level and includes one or more detections associated with each of the objects 112 within the field of view 110 of the radar system 104. The multiple detections associated with a particular object 112 may represent fusion detections (or “fusion objects”). Motion classifier 118 determines the number or percentage of radar detections associated with each fusion detection that indicate that the particular object 112 is moving.

[0051] At operation 406, motion classifier 118 determines counter values ​​for continuous motion and continuous ambiguity. Specifically, motion classifier 118 determines the values ​​of continuous motion counter 408 (e.g., "cntConsecutiveMoving") and continuous ambiguity counter 410 (e.g., "cntConsecutiveAmbigious"). For each frame of radar data, motion classifier 118 can classify the fused detections associated with each object 112 as moving or ambiguous. For example, if more than fifty percent of the radar detections associated with the fused detections indicate that a particular object 112 is moving, the continuous motion counter 408 is linearly incremented. The continuous motion counter 408 can have a maximum value of one hundred. In other implementations, motion classifier 118 can use different percentage thresholds to classify each fused detection as moving or ambiguous. If the continuous motion counter 408 is incremented, motion classifier 118 reinitializes the continuous ambiguity counter 410 to zero.

[0052] Similarly, if less than 50 percent of the radar detections associated with the fusion detection indicate that a specific object 112 is moving, the continuous fuzzy counter 410 is linearly incremented. The continuous fuzzy counter 410 may also have a maximum value (e.g., one hundred). If the continuous fuzzy counter 410 is incremented, the motion classifier 118 reinitializes the continuous motion counter 408 to zero. The motion classifier 118 stores historical values ​​of the continuous motion counter 408 and the continuous fuzzy counter 410 for each object 112.

[0053] At operation 414, motion classifier 118 determines a motion threshold for fusion detection based on the object category of a specific object 112. Specifically, motion classifier 118 can determine the motion classification of VRU movement by using a lower threshold to account for the uncertainty in VRU movement and the difficulty of identifying their motion patterns. For example, motion classifier 118 can use a default minimum movement speed of 1.0 m / s as the motion threshold to classify object 112 as moving. A motion threshold of 0.5 m / s can be used to classify pedestrians as moving. Motion classifier 118 can use visual data 412 to identify a specific object 112 as a pedestrian. Visual data 412 can be image data from camera 106.

[0054] Similarly, the motion classifier 118 can use different cross-direction (e.g., vertical) motion thresholds based on the object type associated with each fusion detection. For example, a vertical motion threshold of 2.0 m / s can be used to classify vehicles as having vertical motion. In contrast, a vertical motion threshold of 1.0 m / s can be used to classify pedestrians as moving perpendicular to vehicle 102.

[0055] At operation 416, motion classifier 118 determines whether the fused detection can be classified as intersecting traffic. Motion classifier 118 can use radar data 402 to determine radial and orthogonal velocities for each fused detection and its heading; these radial and orthogonal velocities are then used for each data period to determine the intersecting traffic motion state of each object 112. Based on an appropriate vertical motion threshold, motion classifier 118 determines whether to set the vertical motion (radar) flag 418 (e.g., "f_cur_perp_motion") to true. The vertical motion (radar) flag 418 is a Boolean flag.

[0056] Similarly, motion classifier 118 can use visual data 412 to determine radial and orthogonal velocities for each fused detection and its heading. Based on an appropriate vertical motion threshold, motion classifier 118 determines whether to set the vertical motion (visual) flag 420 (e.g., "f_vis_data_perp_motion") to true. The vertical motion (visual) flag 420 is also a Boolean flag and allows motion classifier 118 to add redundancy to motion classification by discretely including visual data 412. This redundancy in vertical motion classification increases the robustness of the motion classification algorithm of motion classifier 118.

[0057] The motion classifier 118 also uses a continuous vertical motion counter 422 (e.g., "cntConsecutivePerp") to track the cyclic history of vertical motion classification for each fusion detection. For that data period, the continuous vertical motion counter 422 is incremented by one if the vertical motion (radar) flag 418 or the vertical motion (visual) flag 420 of the fusion detection is true. The continuous vertical motion counter 422 can have a maximum value of twenty. The motion classifier 118 stores the historical values ​​of the continuous vertical motion counter 422 for each object 112.

[0058] At operation 424, motion classifier 118 determines the historical motion classification for each fusion detection. If the continuous motion counter 408 of the fusion detection has a value greater than or equal to three, it is determined that a particular object 112 has motion in that data period, and the continuous motion flag 426 (e.g., "f_ConsecutiveMoving_ok") is marked as true for that data period. In other implementations, motion classifier 118 may use a value other than three to determine the state value of the continuous motion flag 426.

[0059] Similarly, if the fused detection of the continuous vertical motion counter 422 has a value greater than or equal to three, it is determined that the specific object 112 has vertical motion in that data period, and the continuous vertical motion flag 428 (e.g., "f_perp_motion") is marked as true for that data period. In other implementations, the motion classifier 118 can use a value other than three to determine the state value of the continuous vertical motion flag 428. Considering the difficulty of detecting vertical motion, the continuous vertical motion counter 422 can be linearly reduced. The linear reduction of the continuous vertical motion counter 422 can avoid unstable driving operations from the vehicle-based system 208, which would be caused by an immediate drop in the counter value.

[0060] At operation 430, motion classifier 118 determines the motion classification of potential motion fusion detection. If the speed of the fusion detection is greater than a clear motion threshold (e.g., 4 m / s or 8.95 miles per hour (mph)), motion classifier 118 can confidently classify object 112 as moving. In this way, computational power can be saved for fusion detections that are safely considered moving. The clear motion threshold can also be adjusted based on object type or driving environment.

[0061] For fusion detections below the clear movement threshold, motion classifier 118 can use the motion threshold and vertical motion threshold from operation 414 to determine whether radar data 402 indicates that the fusion detection is moving. If it is determined that the object is moving, the movement flag 432 (e.g., "f_moving") is set to true. For pedestrians and other VRUs with true or positive values ​​for continuous motion flag 426 and continuous vertical motion flag 428, motion classifier 118 sets the movement flag 432 to true if the corresponding motion threshold and vertical motion threshold are also met. Additional constraints or checks for VRUs help reduce false motion classifications of such objects.

[0062] If the motion classifier 118 is unable to perform motion classification using radar data 402 for a specific data period of fusion detection, then the motion classifier 118 can use visual data 412 to perform motion classification. The motion classifier 118 can only use visual data 412 to ensure its accuracy if the difference between velocities determined according to visual data 412 is less than 3 m / s (or some other threshold) in both the longitudinal and lateral directions. Visual data 412 can be used to update the continuous motion counter 408 and the continuous blur counter 410. Furthermore, the motion classifier 118 can use visual data 412 to recalculate the continuous motion flag 426 to determine whether the movement flag 432 should be set to true for fusion detection.

[0063] If motion classification cannot be performed using radar data 402 or visual data 412, the motion classifier 118 can retain the value of the motion flag 432 from a previous data period. This scenario may occur when updated sensor data is not available for a particular data period.

[0064] If the movement flag 432 is updated for fusion detection within a specific data period, the motion classifier 118 can also update the stationary flag, movable flag, fast-moving flag, slow-moving flag, oncoming flag, oncoming flag, backward flag, and backward flag for fusion detection. These updates to other motion flags ensure consistency among motion flags. The motion classifier 118 can use a threshold of 7.5 m / s to distinguish between fast-moving and slow-moving objects, regardless of the object type. If the continuous motion counter 408 is zero and the fusion detection has a speed below the corresponding threshold, the stationary flag can be set to true. If the stationary flag is not true, the movable flag is set to true. The movable flag represents a historical version of the movement flag 432 and defines whether a particular object has ever been classified as moving. If the fusion detection indicates that a particular object 112 is moving away from the vehicle 102, the oncoming flag is set to true. If the fusion detection indicates that a particular object 112 is moving towards the vehicle 102, the backward flag is set to true. Similar to movable signs, oncoming signs and backward signs are historical versions of oncoming and backward signs, respectively.

[0065] At operation 434, the motion classifier 118 can determine whether to increment the historical motion counter for each fusion detection. For example, if the motion flag of the fusion detection is set to true, the consecutive arbitrary motion counter 436 for a particular object 112 is incremented by one. The consecutive arbitrary motion counter 436 represents the number of data periods in which the motion flag 432 of a particular fusion detection is activated or set to true.

[0066] The motion classifier 118 can also update the continuous arbitrary motion past counter 438 based on the value of the continuous arbitrary motion counter 436. Specifically, the continuous arbitrary motion past counter 438 represents the maximum value of the continuous arbitrary motion counter 436 for a particular fusion detection and can be used to determine potentially erroneous movable classifications, as described with respect to operation 440. The motion classifier 118 stores historical values ​​of the continuous arbitrary motion counter 436 and the continuous arbitrary motion past counter 438 for each object 112.

[0067] At operation 440, motion classifier 118 determines whether to correct outdated information in flags and / or counters. For example, motion classifier 118 may determine whether to reset the movable flag, the oncoming flag, and the backward flag. If a particular object 112 is a recent fusion detection (e.g., it has been detected for less than ten data periods), is more than 150 meters away, has a maximum observed velocity of less than 2.5 m / s, and has consecutive arbitrary motion counter values ​​of less than ten, then the movable flag, the oncoming flag, and the backward flag are reset to zero. Other period, distance, and velocity thresholds may be used to determine whether to reset these flags. Considering the challenges of motion classification around VRUs, exceptions for resetting these flags are set for VRUs.

[0068] At operation 442, the heading classifier 120 can determine the heading classification of the fusion detection. The heading classifier 120 can classify the heading based on the detected heading of the fusion detection. The heading sector can be defined by the heading classifier 120 as north, northeast, east, southeast, south, southwest, west, and northwest headings. If the heading classification is set to east or west for the fusion detection, the heading classifier 120 can set the intersection flag 444 (e.g., "f_crossing") to true for a specific object 112. The intersection flag 444 can be provided to the vehicle-based system 208 to improve decision-making related to straight-through-intersection path scenarios.

[0069] Example of sports classification

[0070] Figure 5-1 and Figure 5-2 Conceptual diagrams 500-1 and 500-2, respectively, illustrate motion classification performed using low-level detection. Conceptual diagrams 500-1 and 500-2 demonstrate the motion classification process of motion classifier 118 for two different driving scenarios. Conceptual diagrams 500-1 and 500-2 show example motion signs 506 and 510 provided by motion classifier 118 to vehicle-based system 208; however, motion signs 506 and 510 are not necessarily limited to this order or combination of signs. In other implementations, fewer or more signs and counters may be provided to vehicle-based system 208 to improve the operation of vehicle 102.

[0071] exist Figure 5-1 In the scenario, vehicle 102 is driving along a road, while another vehicle 502 is driving in front of it. Vehicle 102 includes a radar system 104 and a camera 106, which have respective fields of view including the other vehicle 502. The radar system 104 can emit radar signals 504 reflected by the other vehicle 502 and processed by the radar system 104 and / or a motion classifier 118.

[0072] The motion classifier 118 can perform motion and heading classification processes based on the concept diagram 400 and set the corresponding value of the motion flag 506 for another vehicle 502. Specifically, the motion classifier 118 can set the motion flag 506-1 to true, which is determined by... Figure 5-1 The value "1" in the value indicates that the movement flag 506-1 indicates that another vehicle 502 is moving. Since the movement flag 506-1 is set to true, the motion classifier 118 sets the stationary flag 506-7 to false, which is determined by... Figure 5-1 The value "0" indicates this. Furthermore, since the movement flag 506-1 was set to true in this data cycle or a previous data cycle, the movable flag 506-4 is also set to true. The radar system 104 or motion classifier 118 also determines that another vehicle 502 is moving away from vehicle 102 and sets the reverse flag 506-2 to true. Since the reverse flag 506-2 was set to true in this data cycle or a previous data cycle, the reversible flag 506-3 is also set to true. Since another vehicle 502 is moving away from vehicle 102, the motion classifier 118 sets the oncoming flag 506-5 to false. Similar to the reversible flag 506-3, the motion classifier 118 also sets the oncoming flag 506-6 to false because the oncoming flag 506-5 was not true for this data cycle or a previous data cycle.

[0073] exist Figure 5-2 In the scenario, vehicle 102 is driving along the road, while pedestrian 508 is crossing the road in front of it. Radar system 104 can emit radar signal 504 reflected by pedestrian 508 and processed by radar system 104 and / or motion classifier 118.

[0074] Motion classifier 118 can perform motion and heading classification processes according to concept diagram 400 and set corresponding values ​​for motion markers 510 for pedestrians 508. Specifically, motion classifier 118 can set moving marker 510-1 to true. Moving marker 506-1 indicates that pedestrian 508 is moving (e.g., perpendicular to vehicle 102). Since moving marker 506-1 is set to true, motion classifier 118 sets stationary marker 510-7 to false. Furthermore, since moving marker 506-1 was set to true in this data cycle or a previous data cycle, movable marker 510-4 is also set to true. Radar system 104 or motion classifier 118 also determines that pedestrian 508 is moving toward vehicle 102 and sets oncoming marker 510-5 to true. Since oncoming marker 510-5 was set to true in this data cycle or a previous data cycle, oncoming marker 510-6 is also set to true. Since pedestrian 508 is moving toward vehicle 102, motion classifier 118 also sets the backward flag 510-2 to false. Similar to the oncoming flag 510-6, motion classifier 118 also sets the backward flag 510-3 to false, because the backward flag 510-2 was not true for this data period or a previous data period.

[0075] Example of sports classification

[0076] The following are some examples.

[0077] Example 1. A method comprising: for a data period, acquiring radar signals reflected from a radar system by one or more objects in an area near a primary vehicle; using the radar signals, identifying one or more detections associated with each of the one or more objects, wherein the one or more detections associated with a particular object represent a fusion detection of the particular object; for each fusion detection, determining whether the fusion detection indicates that the particular object is moving; in response to determining that the fusion detection indicates that the particular object is moving, incrementing a continuous motion counter associated with the particular object; in response to determining that the continuous motion counter associated with the particular object has a value greater than a threshold counter value, setting a continuous motion flag associated with the particular object to true; if the continuous motion flag associated with the particular object is set to true and the fusion detection associated with the particular object has a velocity component greater than a motion threshold, then setting a motion flag associated with the particular object to true; in response to setting the motion flag associated with the particular object to true, incrementing a historical motion counter associated with the particular object; and operating a primary vehicle in a road based on the state of the motion flag and the historical motion counter associated with each of the one or more objects.

[0078] Example 2. The method of Example 1, further comprising: for each fusion detection, determining whether the fusion detection indicates that a particular object is moving perpendicular to the main vehicle; in response to determining that the fusion detection indicates that a particular object is moving perpendicular to the main vehicle, incrementing a continuous vertical motion counter associated with the particular object; in response to determining that the continuous vertical motion counter has a value greater than a threshold counter value, setting a continuous vertical motion flag associated with the particular object to true; if the continuous vertical motion flag associated with the particular object is set to true and the fusion detection associated with the particular object has an orthogonal velocity component greater than a vertical motion threshold, then setting a movement flag associated with the particular object to true; in response to setting the movement flag associated with the particular object to true, determining a heading classification associated with the particular object based on the heading associated with the fusion detection; if the heading classification associated with the particular object is set to perpendicular to the main vehicle, then setting a cross sign to true; and operating the main vehicle in the road based on the state of the cross sign associated with each of one or more objects.

[0079] Example 3. The method of Example 2, wherein determining that the fusion detection indicates that a particular object is moving perpendicular to the main vehicle includes: determining that the orthogonal velocity component associated with the fusion detection is greater than zero.

[0080] Example 4. The method of Example 3, further comprising: in response to determining that the orthogonal velocity component associated with the fusion detection is not greater than zero, decrementing the value of the continuous vertical motion counter associated with the particular object by one.

[0081] Example 5. A method of any one of Examples 2 to 4, the method further comprising: for each fusion detection and using visual data from one or more cameras of the main vehicle, determining the object type of a particular object; and setting a vertical motion threshold associated with the particular object based on the object type of the particular object.

[0082] Example 6. The method of Example 5, wherein determining that the orthogonal velocity component of the fusion detection is greater than the vertical motion threshold includes: determining that the first orthogonal velocity component of the fusion detection based on radar signals is greater than the vertical motion threshold; or determining that the second orthogonal velocity component of the fusion detection based on visual data is greater than the vertical motion threshold.

[0083] Example 7. A method of any of the preceding examples, wherein determining whether a fusion detection indicates that a particular object is moving comprises: determining that the particular object is moving if at least half of one or more detections associated with the particular object indicate that the particular object is moving; and determining that the movement of the particular object is fuzzy if less than half of one or more detections associated with the particular object indicate that the movement is fuzzy.

[0084] Example 8. The method of Example 7, further comprising: in response to determining that the fusion detection indicates that the movement of a particular object is blurred, incrementing a continuous blur counter associated with the particular object; in response to incrementing the continuous blur counter associated with the particular object, resetting the value of a continuous motion counter associated with the particular object to zero; or in response to incrementing the continuous motion counter associated with the particular object, resetting the value of a continuous blur counter associated with the particular object to zero.

[0085] Example 9. A method for any of the previous examples, where the threshold counter value is three.

[0086] Example 10. A method of any of the preceding examples, the method further comprising: determining whether a fusion detection associated with a particular object has a longitudinal velocity component greater than a motion threshold by: determining an object type for each fusion detection and using visual data from one or more cameras of a primary vehicle; setting a motion threshold based on the object type of the particular object; and comparing the longitudinal velocity component associated with the particular object with the motion threshold.

[0087] Example 11. The method of Example 10, wherein comparing a longitudinal velocity component associated with a particular object with a motion threshold includes: comparing a first longitudinal velocity component associated with the particular object with a motion threshold, the first longitudinal velocity component being obtained from a radar signal; comparing a second longitudinal velocity component associated with the particular object with a motion threshold, the second longitudinal velocity component being obtained from visual data; or using a previous state of a motion marker associated with the particular object determined during a previous data period.

[0088] Example 12. The method of Example 10 or 11, determining whether the longitudinal velocity component of the fusion detection associated with a specific object is greater than a motion threshold, includes: determining whether the longitudinal velocity component is greater than a clear motion threshold.

[0089] Example 13. A method of any of the preceding examples, the method further comprising: in response to setting a movement flag associated with a particular object to true, setting a movable flag associated with the particular object to true, the movable flag associated with the particular object indicating whether the movement flag associated with the particular object has been set to true in a data cycle or any previous data cycle; operating a primary vehicle in the road based on the state of the movable flag associated with each of the one or more objects; and resetting the movable flag associated with the particular object to false if the historical movement counter associated with the particular object has a value less than ten.

[0090] Example 14. A system comprising: a radar system configured to: receive, for a data period, radar signals reflected by one or more objects in a region near a primary vehicle; and a processor configured to perform a method of any of the preceding examples.

[0091] Example 15. A computer-readable storage medium including computer-executable instructions that, when executed by a processor, cause the processor to perform a method of any one of Examples 1 to 13.

[0092] While various embodiments of the present disclosure have been described in the foregoing description and illustrated in the accompanying drawings, it should be understood that the present disclosure is not limited thereto, but can be practiced in various ways within the scope of the following claims. It will be apparent from the foregoing description that various modifications can be made without departing from the spirit and scope of the present disclosure as defined by the following claims.

Claims

1. A method for a means of transport, the method comprising: For a given data period, radar signals reflected by one or more objects in the vicinity of the main vehicle are obtained from the radar system. Using the radar signal, one or more detections associated with each of the one or more objects are identified, and the one or more detections associated with a particular object among the one or more objects represent the fusion detection of that particular object; For each fusion detection, determine whether the fusion detection indicates that the specific object is moving; In response to determining that the fusion detection indicates that the specific object is moving, the continuous motion counter associated with the specific object is incremented; In response to determining that the continuous motion counter associated with the specific object has a value greater than a threshold counter value, the continuous motion flag associated with the specific object is set to true; The following steps are taken to determine whether the fusion detection associated with the specific object has a longitudinal velocity component greater than a motion threshold: (i) for each fusion detection and using visual data from one or more cameras of the main vehicle, determine the object type of the specific object; (ii) setting the motion threshold based on the object type of the specific object; and (iii) comparing the longitudinal velocity component associated with the specific object with the motion threshold; If the continuous motion flag associated with the specific object is set to true, and the longitudinal velocity component of the fusion detection associated with the specific object has a velocity component greater than the motion threshold, then the motion flag associated with the specific object is set to true. In response to setting the movement flag associated with the specific object to true, increment the historical movement counter associated with the specific object; and The primary vehicle is operated on the road based on the status of the movement marker and the historical movement counter associated with each of the one or more objects.

2. The method as described in claim 1, characterized in that, The method further includes: For each fusion detection, determine whether the fusion detection indicates that the particular object is moving perpendicular to the main vehicle; In response to determining that the fusion detection indicates that the specific object is moving perpendicular to the main vehicle, the continuous vertical motion counter associated with the specific object is incremented; In response to determining that the continuous vertical motion counter has a value greater than the threshold counter value, the continuous vertical motion flag associated with the specific object is set to true; If the continuous vertical motion flag associated with the specific object is set to true, and the fusion detection associated with the specific object has an orthogonal velocity component greater than the vertical motion threshold, then the motion flag associated with the specific object is set to true. In response to setting the movement flag associated with the specific object to true, a heading classification associated with the specific object is determined based on the heading associated with the fusion detection; If the heading classification associated with the specific object is set to be perpendicular to the primary vehicle, then the cross sign is set to true; and The main vehicle operates on the road based on the state of the cross sign associated with each of the one or more objects.

3. The method as described in claim 2, characterized in that, Determining that the fusion detection indicates that the specific object is moving perpendicular to the main vehicle includes: determining that the orthogonal velocity component associated with the fusion detection is greater than zero.

4. The method as described in claim 3, characterized in that, The method further includes: In response to determining that the orthogonal velocity component associated with the fusion detection is not greater than zero, the value of the continuous vertical motion counter associated with the specific object is reduced by one.

5. The method as described in claim 2, characterized in that, The method further includes: For each fusion detection and using visual data from one or more cameras of the main vehicle, determine the object type of the specific object; and The vertical motion threshold associated with the specific object is set based on the object type of the specific object.

6. The method as described in claim 5, characterized in that, Determining that the orthogonal velocity component of the fused detection is greater than the vertical motion threshold includes: It is determined that the first orthogonal velocity component of the fused detection based on the radar signal is greater than the vertical motion threshold; or It is determined that the second orthogonal velocity component of the fusion detection based on the visual data is greater than the vertical motion threshold.

7. The method as described in claim 1, characterized in that, Determining whether the fusion detection indicates that the specific object is moving includes: If at least half of the one or more detections associated with the specific object indicate movement of the specific object, then it is determined that the specific object is moving; and If less than half of the one or more detections associated with the particular object indicate movement, then the determination of movement of the particular object is ambiguous.

8. The method as described in claim 7, characterized in that, The method further includes: In response to determining that the fusion detection indicates that the movement of the particular object is fuzzy, the continuous fuzziness counter associated with the particular object is incremented; In response to incrementing the continuous fuzzy counter associated with the specific object, the value of the continuous motion counter associated with the specific object is reset to zero; or In response to incrementing the continuous motion counter associated with the specific object, the value of the continuous fuzzy counter associated with the specific object is reset to zero.

9. The method as described in claim 1, characterized in that, The threshold counter value is three.

10. The method as described in claim 1, characterized in that, Comparing the longitudinal velocity component associated with the specific object with the motion threshold includes: A first longitudinal velocity component associated with the specific object is compared with the motion threshold, the first longitudinal velocity component being obtained from the radar signal; The second longitudinal velocity component associated with the specific object is compared with the motion threshold, the second longitudinal velocity component being obtained from the visual data; or Use the previous state of the movement flag associated with the specific object, which was determined during a previous data period.

11. The method as described in claim 1, characterized in that, Determining whether the longitudinal velocity component of the fusion detection associated with the specific object is greater than the motion threshold includes: determining whether the longitudinal velocity component is greater than a clear motion threshold.

12. The method as described in claim 1, characterized in that, The method further includes: In response to setting the movement flag associated with the specific object to true, the movable flag associated with the specific object is set to true, wherein the movable flag associated with the specific object indicates whether the movement flag associated with the specific object has been set to true in the data cycle or any previous data cycle; Based on the state of the movable sign associated with each of the one or more objects, the main vehicle operates in the road; and If the historical motion counter associated with the specific object has a value less than ten, the movable flag associated with the specific object is reset to false.

13. A system for a vehicle, the system comprising: A radar system configured to: receive radar signals reflected by one or more objects in the vicinity of the main vehicle for a data cycle; Processor, the processor being configured to: Using the radar signal, one or more detections associated with each of the one or more objects are identified, and the one or more detections associated with a particular object among the one or more objects represent the fusion detection of that particular object; For each fusion detection, determine whether the fusion detection indicates that the specific object is moving; In response to determining that the fusion detection indicates that the specific object is moving, the continuous motion counter associated with the specific object is incremented; In response to determining that the continuous motion counter associated with the specific object has a value greater than a threshold counter value, the continuous motion flag associated with the specific object is set to true; The following steps are taken to determine whether the fusion detection associated with the specific object has a longitudinal velocity component greater than a motion threshold: (i) for each fusion detection and using visual data from one or more cameras of the main vehicle, determine the object type of the specific object; (ii) Set the motion threshold based on the object type of the specific object; as well as (iii) Compare the longitudinal velocity component associated with the specific object with the motion threshold; If the continuous motion flag associated with the specific object is set to true, and the longitudinal velocity component of the fusion detection associated with the specific object has a velocity component greater than the motion threshold, then the motion flag associated with the specific object is set to true. In response to setting the movement flag associated with the specific object to true, increment the historical movement counter associated with the specific object; and The primary vehicle is operated on the road based on the status of the movement marker and the historical movement counter associated with each of the one or more objects.

14. The system as described in claim 13, characterized in that, The processor is further configured to: For each fusion detection, determine whether the fusion detection indicates that the particular object is moving perpendicular to the main vehicle; In response to determining that the fusion detection indicates that the specific object is moving perpendicular to the main vehicle, the continuous vertical motion counter associated with the specific object is incremented; In response to determining that the continuous vertical motion counter has a value greater than the threshold counter value, the continuous vertical motion flag associated with the specific object is set to true; If the continuous vertical motion flag associated with the specific object is set to true, and the fusion detection associated with the specific object has an orthogonal velocity component greater than the vertical motion threshold, then the motion flag associated with the specific object is set to true. In response to setting the movement flag associated with the specific object to true, a heading classification associated with the specific object is determined based on the heading associated with the fusion detection; If the heading classification associated with the specific object is set to be perpendicular to the primary vehicle, then the cross sign is set to true; as well as The main vehicle operates on the road based on the state of the cross sign associated with each of the one or more objects.

15. The system as described in claim 14, characterized in that, The processor is further configured to: The fusion detection is determined to indicate that the specific object is moving perpendicular to the main vehicle by determining that the orthogonal velocity component associated with the fusion detection is greater than zero; as well as In response to determining that the orthogonal velocity component associated with the fusion detection is not greater than zero, the value of the continuous vertical motion counter associated with the specific object is reduced by one.

16. The system as described in claim 14, characterized in that: The system further includes one or more cameras configured to capture visual data of the one or more objects in the vicinity of the main vehicle; and The processor is further configured to: For each fusion detection and using the visual data, determine the object type of the specific object; and The vertical motion threshold associated with the specific object is set based on the object type of the specific object.

17. The system as claimed in claim 13, characterized in that, The processor is configured to compare the longitudinal velocity component associated with the specific object with the motion threshold by: A first longitudinal velocity component associated with the specific object is compared with the motion threshold, the first longitudinal velocity component being obtained from the radar signal; The second longitudinal velocity component associated with the specific object is compared with the motion threshold, the second longitudinal velocity component being obtained from the visual data; or Use the previous state of the movement flag associated with the specific object, which was determined during a previous data period.

18. A computer-readable storage medium comprising computer-executable instructions, which, when executed by a processor, cause the processor to: For a data cycle, radar signals reflected by one or more objects in the vicinity of the main vehicle are obtained from the radar system. Using the radar signal, one or more detections associated with each of the one or more objects are identified, and the one or more detections associated with a particular object among the one or more objects represent the fusion detection of that particular object; For each fusion detection, determine whether the fusion detection indicates that the specific object is moving; In response to determining that the fusion detection indicates that the specific object is moving, the continuous motion counter associated with the specific object is incremented; In response to determining that the continuous motion counter associated with the specific object has a value greater than a threshold counter value, the continuous motion flag associated with the specific object is set to true; The following steps are taken to determine whether the fusion detection associated with the specific object has a longitudinal velocity component greater than a motion threshold: (i) for each fusion detection and using visual data from one or more cameras of the main vehicle, determine the object type of the specific object; (ii) setting the motion threshold based on the object type of the specific object; and (iii) comparing the longitudinal velocity component associated with the specific object with the motion threshold; If the continuous motion flag associated with the specific object is set to true, and the longitudinal velocity component of the fusion detection associated with the specific object has a velocity component greater than the motion threshold, then the motion flag associated with the specific object is set to true. In response to setting the movement flag associated with the specific object to true, increment the historical movement counter associated with the specific object; and The primary vehicle is operated on the road based on the status of the movement marker and the historical movement counter associated with each of the one or more objects.

Images