Roadside vehicle detection device

Abstract

To provide an on-street vehicle detection device capable of promptly detecting the presence of on-street parking.SOLUTION: An on-street vehicle detection device 100 comprises: a speed vector calculation section VC which calculates a speed vector of a vehicle in a detection range on a street; and a determination section JD (an on-street parking determination section PD) which determines the presence / absence of on-street parking from a relation of the speed vector, which is calculated by the speed vector calculation section VC, in the detection range. In the on-street vehicle detection device 100, the presence / absence of on-street parking is determined promptly based on the speed vector of the vehicle in the detection range.SELECTED DRAWING: Figure 2

Description

Technical Field

[0001] The present invention relates to a road vehicle detection device that detects vehicles present on the road.

Background Art

[0002] For example, as a road-vehicle communication device for assisting an autonomous vehicle in continuing autonomous driving, there is known one that provides information on the presence or absence of on-road parking to the autonomous vehicle (see Patent Document 1).

[0003] However, in Patent Document 1 above, when determining on-road parking, since the presence of on-road parking is determined based on the time of a red signal or the like, there is a possibility that a predetermined time is required until information on on-road parking is provided.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

[0005] The present invention has been made in view of the above points, and an object thereof is to provide a road vehicle detection device that can quickly detect the presence of on-road parking.

[0006] The road vehicle detection device for achieving the above object includes a speed vector calculation unit that calculates a speed vector of a vehicle in a detection range on the road, and a determination unit that determines the presence or absence of on-road parking based on the relationship of the speed vectors in the detection range calculated by the speed vector calculation unit.

[0007] In the above road vehicle detection device, by determining the presence or absence of on-road parking based on the speed vector of the vehicle in the detection range, it is possible to make a determination more quickly than in the case of determining the presence of on-road parking based on, for example, the time of a red signal or the like.

[0008] In a specific aspect of the present invention, the determination unit determines whether or not a vehicle is parked on the street based on a trained model relating to velocity vectors. In this case, a quick and accurate determination can be made.

[0009] In another aspect of the present invention, the determination unit determines whether a stopped vehicle is parked on the road or not based on the relationship between the speed vector of a stopped vehicle whose speed vector is below a predetermined value and the speed vectors of vehicles surrounding the stopped vehicle. In this case, the determination is made based on the relative relationship between the vehicles.

[0010] In yet another aspect of the present invention, the determination unit determines that a stopped vehicle is parked on the road if the speed vector calculation unit calculates changes in the speed vectors of other vehicles in front of and behind the stopped vehicle. In this case, road parking in specific situations can be accurately identified.

[0011] In yet another aspect of the present invention, the determination unit determines that a stopped vehicle is parked on the road if the velocity vector calculation unit determines that the velocity vectors of all vehicles in front of the stopped vehicle are greater than a predetermined value. In this case, road parking in specific situations can be accurately identified.

[0012] In yet another aspect of the present invention, the determination unit determines that a stopped vehicle is parked on the road if there is an empty space in front of the stopped vehicle. In this case, it is possible to accurately identify road parking in certain situations.

[0013] In yet another aspect of the present invention, an image extraction unit is provided that extracts a vehicle from a series of images captured within a detection range, and a velocity vector calculation unit calculates a velocity vector from the changes in the vehicle as a series of images extracted by the image extraction unit. In this case, a determination can be made based on the appropriately calculated velocity vector.

[0014] In yet another aspect of the present invention, the system includes a communication unit that, in response to driving information received from the autonomous vehicle, returns to the autonomous vehicle the result of the determination of the traffic conditions within the detection range by the determination unit. In this case, information can be provided to the autonomous vehicle quickly and accurately. [Brief explanation of the drawing]

[0015] [Figure 1] This is a conceptual diagram illustrating the overview of the road vehicle detection device according to the first embodiment. [Figure 2] This is a block diagram illustrating one example configuration of a road vehicle detection device. [Figure 3] (A) to (C) are conceptual diagrams to explain future location information. [Figure 4] (A) to (D) are conceptual diagrams illustrating data extraction for determining whether or not there is street parking. [Figure 5] (A) to (C) are conceptual diagrams illustrating the manner in which a stopped vehicle is determined to be parked on the street. [Figure 6] This is a flowchart illustrating the series of operations in a road vehicle detection system. [Figure 7] This is a conceptual diagram illustrating the overview of the road vehicle detection device according to the second embodiment. [Figure 8] (A) is a conceptual diagram illustrating the creation of a trained model used in the road vehicle detection device of the third embodiment, and (B) is a block diagram illustrating an example configuration of the road vehicle detection device. [Modes for carrying out the invention]

[0016] [First Embodiment] Hereinafter, an example of a road vehicle detection device according to the first embodiment will be described with reference to Figure 1, etc.

[0017] Figure 1 is a conceptual diagram illustrating the road vehicle detection device 100 according to this embodiment. The road vehicle detection device 100 is a device for detecting the driving status of vehicles GM on the road, and in the figure, the range of the road detected by the road vehicle detection device 100 is referred to as the detection range DR. In one example shown in the figure, the road vehicle detection device 100 is incorporated into a driver assistance device SS that provides information on road conditions to an autonomous vehicle VE on the infrastructure side (roadside) to assist driving. In other words, the driver assistance device SS uses the road vehicle detection device 100, which detects vehicles GM on the road within the detection range DR, to provide information on on-street parking to the autonomous vehicle VE traveling upstream of the detection range DR. Alternatively, it can be considered that the road vehicle detection device 100 also communicates with the autonomous vehicle VE, thereby functioning as a driver assistance device SS.

[0018] The road vehicle detection device 100 or driving assistance device SS of this embodiment comprises a sensor unit 10, a control device 50, and a communication unit 70 to achieve the above configuration. The sensor unit 10 detects vehicles GM that are located within a detection range DR, which is a predetermined area of ​​the road, and the control device 50 analyzes the driving state of the vehicles GM, enabling the understanding of the traffic situation throughout the detection range DR. In particular, in this embodiment, the presence or absence of on-street parking is determined based on the analysis by the control device 50. From the above, a typical example of the detection range DR is assumed to be a location where on-street parking of vehicles GM frequently occurs.

[0019] Also, in an example here, the on-road vehicle detection device 100 shall transmit the determination result in the control device 50 to the autonomous driving vehicle VE traveling upstream of the detection range DR via the communication unit 70. That is, by having the communication unit 70, the on-road vehicle detection device 100 functions as a driving support device SS installed on the roadside to support the driving of the autonomous driving vehicle VE. Alternatively, it can also be said that the driving support device SS includes the on-road vehicle detection device 100 and provides information on on-road parking in the detection range DR, particularly from the roadside, to the autonomous driving vehicle VE to support the driving of the autonomous driving vehicle VE. Also, as described above, the autonomous driving vehicle VE can smoothly travel at that location by receiving in advance from the on-road vehicle detection device 100 information such as the presence or absence of on-road parking in the detection range DR, which is the destination of travel, and the location of parked vehicles if there is on-road parking at the travel destination.

[0020] The sensor unit 10 is composed of, for example, various cameras and ranging devices such as LiDAR. The sensor unit 10 acquires various data such as continuous image data such as videos in order to capture the behavior of the vehicle GM existing in the detection range DR.

[0021] The control device 50 is composed of, for example, a CPU, a storage device, etc., and performs image processing on the image data of the detection range DR acquired by the sensor unit 10, so that, as conceptually shown in the figure, various data such as the position of the vehicle GM existing in the detection range DR, the displacement of the position, and the moving speed can be acquired. Specifically, as the vehicle GM, as exemplified by hatching in the figure, a stopped vehicle SV that is stopped can be captured, or by comparing continuous image data, a running vehicle RV that is moving over time, that is, running, as exemplified by the vehicle GM shown by a dashed line in the figure, can be captured. In the figure, the state of change of the position of the running vehicle RV is indicated by arrows, and these arrows are taken as the velocity vectors VV. In the present embodiment, as described above, each vehicle GM is individually extracted from various images obtained by imaging or ranging the detection range DR, and further, the state of each vehicle GM moving is captured as the velocity vector VV, and by comparing these, it is possible to determine the presence or absence of on-road parking. In the above example of the illustration, the change in the velocity vector VV of the running vehicle RV, which is another vehicle, is calculated before and after the stopped vehicle SV shown by hatching. From these analysis results, it is found that the running vehicle RV has overtaken and further passed the stopped vehicle SV. In such a case, the stopped vehicle SV that has been overtaken by the running vehicle RV is determined by the control device 50 to be parked on the road. Hereinafter, the vehicle GM (stopped vehicle SV) determined to be parked on the road is denoted as the parked vehicle PV. That is, in the illustrated case, the stopped vehicle SV shown by hatching is the parked vehicle PV that is parked on the road in the detection range DR.

[0022] In addition to the above, the road vehicle detection device 100 or the driver assistance device SS is equipped with a communication unit 70, as previously described. The communication unit 70 is a device that handles communication with the autonomous vehicle VE, and upon receiving future location information (information indicating the planned driving route of the autonomous vehicle VE) from the autonomous vehicle VE, it provides information about the traffic conditions at the planned driving destination accordingly. In this embodiment, when future location information is received from the autonomous vehicle VE, as described above, the communication unit 70 transmits to the autonomous vehicle VE the determination result regarding the status of on-road parking in the detection range DR located downstream of the autonomous vehicle VE's destination as information to be provided. As described above, the communication unit 70 returns to the autonomous vehicle VE the determination result of the traffic conditions in the detection range DR by the control device 50 (more precisely, the determination unit JD described later with reference to Figure 2) in response to the future location information as driving information received from the autonomous vehicle VE.

[0023] The following will provide a more detailed explanation of one example configuration of the road vehicle detection device 100, with reference to the block diagram shown in Figure 2.

[0024] As shown in the diagram and as previously described, the road vehicle detection device 100 or driving assistance device SS comprises a sensor unit 10, a control device 50, and a communication unit 70. Of these, the control device 50 comprises a control unit 51 which is the main body, a determination unit JD, and a data storage unit DS. Alternatively, the control unit 51 is composed of a determination unit JD and a data storage unit DS.

[0025] The control device 50 or control unit 51 is connected to, for example, the sensor unit 10 or the communication unit 70, and receives image data acquired by the sensor unit 10, communicates with an external device, the automated driving vehicle VE, via the communication unit 70, and is responsible for various operation controls.

[0026] The control device 50 (control unit 51) includes, or is composed of, an object detection unit OD, a velocity vector calculation unit VC, and a road parking determination unit PD.

[0027] The object detection unit OD extracts various objects from various data, such as image data acquired by the sensor unit 10. Here, the extracted information is referred to as target information. That is, the target information includes various information such as the operating status of various vehicles and pedestrians present in the detection range DR (see Figure 1), as well as information about the presence of obstacles. In particular, as a road vehicle detection device 100, information regarding vehicles GM (see Figure 1) from the target information is extracted as image information. Hereafter, both the vehicle GM as an object and the image of it detected will be referred to as vehicle GM. A typical example of the object detection unit OD is one in which it is composed of an image extraction unit GE that extracts vehicle GM from continuous images captured by the sensor unit 10 over the detection range DR.

[0028] The velocity vector calculation unit VC extracts the velocity vector VV from the extracted vehicle GM (see Figure 1). That is, the velocity vector calculation unit VC calculates the velocity vector VV of the vehicle GM within the detection range DR on the road. Furthermore, in the above, for vehicles that remain stationary, such as a stationary vehicle SV (see Figure 1), the value of the velocity vector VV will be calculated as zero. Here, we will refer to this situation as vector zero. One typical example of being treated as vector zero is when the calculated velocity vector VV value is less than or equal to a predetermined value. In other words, a vehicle GM whose value is less than or equal to a predetermined value is determined to be a stationary vehicle SV. When the object detection unit OD is configured with an image extraction unit GE, the velocity vector calculation unit VC calculates the velocity vector VV from the changes in the vehicle GM as a continuous image extracted by the image extraction unit GE.

[0029] The on-street parking detection unit PD determines whether or not there is on-street parking based on the relationship between the speed vectors VV within the detection range DR calculated by the speed vector calculation unit VC. In other words, the on-street parking detection unit PD functions as the main body of the detection unit JD.

[0030] The on-street parking determination unit PD, or determination unit JD, determines whether a stopped vehicle SV is parked on the street based on its relationship with the velocity vectors VV of other vehicles GM surrounding the stopped vehicle SV. A typical example is the case of a parked vehicle PV, as shown by the hatching in Figure 1. Specifically, the determination unit JD (on-street parking determination unit PD) determines that the stopped vehicle SV is a parked vehicle PV if the velocity vector calculation unit VC calculates a change in the velocity vectors VV of other vehicles GM, which are moving vehicles RV, in front of and behind the stopped vehicle SV.

[0031] The data storage unit DS is composed of various storage devices, for example, and stores various data and programs necessary for performing the various processing tasks such as the image processing and analysis processing described above, as well as temporarily storing various data acquired by the sensor unit 10, and analysis results and judgment results from the control device 50 (control unit 51).

[0032] The following conceptual explanation will describe an example of future position information transmitted from the autonomous vehicle VE to the road vehicle detection device 100 (driving assistance device SS), with reference to Figure 3. Of Figures 3(A) to 3(C), Figure 3(A) first shows the initial future position information transmitted from the autonomous vehicle VE. Assuming the position where the autonomous vehicle VE is depicted is its current position and the direction of travel of the autonomous vehicle VE is the Z direction, points FP1 to FP4 shown along the Z direction indicate the future position of the autonomous vehicle VE. More specifically, with time T at the current position (present moment) being 0 (T=0), point FP1 indicates the planned position of the autonomous vehicle VE t seconds later (T=t). Similarly, point FP2 indicates the planned position of the autonomous vehicle VE 2t seconds later (T=2t), point FP3 indicates the planned position of the autonomous vehicle VE 3t seconds later (T=3t), and point FP4 indicates the planned position of the autonomous vehicle VE 4t seconds later (T=4t).

[0033] Next, Figure 3(B) shows the second set of future position information transmitted from the autonomous vehicle VE to the road vehicle detection device 100 at the point when t seconds have actually elapsed from the state in Figure 3(A), i.e., t seconds after the state in Figure 3(A). In this case, since t seconds have elapsed from the state in Figure 3(A), the autonomous vehicle VE has moved to the position corresponding to point FP1 in Figure 3(A). At this point, the autonomous vehicle VE transmits its new future position information, i.e., information for new points FP1 to FP4, to the road vehicle detection device 100. Similarly, as shown in Figure 3(C), a third set of future position information (information for points FP1 to FP4) is transmitted after another t seconds. Note that the illustrated example shows the case where the autonomous vehicle VE is driving as planned, but if the plan changes, the new future position information will be different from the previous future position information.

[0034] Furthermore, the illustrated example only shows the case up to point FP4, and omits information beyond point FP4, but it is also possible to include information about future positions further ahead.

[0035] From the perspective of the infrastructure-side road vehicle detection device 100 (driving assistance device SS), as shown in Figures 3(A) to 3(C), as the autonomous vehicle VE continues to drive, new future location information will be transmitted from the autonomous vehicle VE at regular intervals (for example, every t=1 second), and the future location information will be updated accordingly each time. In response to this, the road vehicle detection device 100 can, for example, constantly monitor the traffic situation within the detection range DR, thereby providing the autonomous vehicle VE with the most recently updated information on on-street parking.

[0036] The following will provide a more detailed explanation of data extraction for determining the presence or absence of on-street parking in the road vehicle detection device 100, referring to the conceptual diagrams shown in Figures 4(A) to 4(D). First, as shown in the example in Figure 4(A), assume that two vehicles (vehicles as images), GM1 and GM2, exist as vehicles GM in the first image GG1 of a series of images GG capturing the detection range DR. In this case, as shown in Figure 4(B), each vehicle GM1 and GM2 is extracted using an existing image extraction method, enclosed in a rectangular frame FM, and they are distinguished and detected based on differences in vehicle color, shape, etc. Note that the illustration shows how they are distinguished by changing the hatching pattern (vehicle GM1 is distinguished by a dot pattern, and vehicle GM2 by a cross pattern). Through this image processing, each vehicle GM (vehicles GM1 and GM2) is extracted, and the identity (linking) of each vehicle GM among the series of images GG composed of multiple images is established. As a result, for example, as shown in Figure 4(C), by relating each vehicle GM1 and GM2 in the next image GG2 (where a portion of image GG1 is superimposed with a dashed line) with the previous image GG1 shown in Figure 4(B), and performing velocity vector calculation processing, it is determined that one vehicle GM1 is a moving vehicle RV, and that its velocity vector VV is velocity vector VV1 (not vector zero). On the other hand, it is determined that vehicle GM2 is a stationary vehicle SV, meaning its velocity vector VV is vector zero.

[0037] Furthermore, as shown in the example in Figure 4(D), by associating each vehicle GM1 and GM2 in successive images GG3 with images GG1 and GG2 and performing velocity vector calculation processing, it is found that the stationary vehicle SV (vehicle GM2) remains stationary, while for the moving vehicle RV, in addition to velocity vector VV1, velocity vector VV2 is calculated, revealing that the moving vehicle RV (vehicle GM1) overtakes and then passes the stationary vehicle SV (vehicle GM2). As a result, it is found that the stationary vehicle SV (vehicle GM2) is the parked vehicle PV.

[0038] Below, with reference to the conceptual diagrams shown in Figures 5(A) to 5(C), several examples of how a stopped vehicle SV is determined to be a parked vehicle PV parked on the street will be explained. Note that Figure 5(A) shows one example explained with reference to Figures 4(A) to 4(D), so the explanation will be omitted here, and other examples, namely the example shown in Figure 5(B) and the example shown in Figure 5(C), will be explained.

[0039] First, in the example shown in Figure 5(B), all vehicles GM in front of the stopped vehicle SV, indicated by hatching, start moving and become moving vehicles RV. This corresponds to the case where the velocity vector calculation unit VC shown in Figure 2 calculates that the velocity vector VV of all vehicles GM in front of the stopped vehicle SV is greater than a predetermined value (vector non-zero). In such a case, the determination unit JD (on-street parking determination unit PD) determines that the stopped vehicle SV, indicated by hatching in Figure 5(B), is a parked vehicle PV. Typical examples of such situations include cases where there is congestion or traffic jams further ahead of the detection range DR, or where the traffic light is red. In other words, when the congestion or traffic jams are resolved, or the traffic light turns green, the vehicles GM that were stopped in front of the stopped vehicle SV start moving. If the stopped vehicle SV remains stopped despite such a situation, it is treated as a parked vehicle PV parked on the street.

[0040] Next, Figure 5(C) shows an example where there is an empty space EP in front of the stopped vehicle SV. In other words, unlike the example shown in Figure 5(B), there is nothing in front of the stopped vehicle SV, and the stopped vehicle SV is not moving forward despite being able to move forward. In this case, the determination unit JD (street parking determination unit PD) shown in Figure 2 determines that the stopped vehicle SV is a parked vehicle PV that is parked on the street. As for the minimum size of the empty space EP shown in Figure 5(C), various options are possible, but for example, it could be set to be at least the length of one vehicle in the front-to-back direction.

[0041] Furthermore, regarding the determination patterns for whether or not a vehicle is parked on the street, various other configurations may exist depending on the characteristics of the detection range DR, etc. However, it is thought that by categorizing these according to the characteristics of the detection range DR, it will be possible to create patterns suitable for determining whether or not a vehicle is parked on the street.

[0042] The following describes an example of a series of operations in the road vehicle detection device 100, referring to the flowchart shown in Figure 6. For the sake of simplicity, this explanation will focus on the case where the presence or absence of a parked vehicle is determined based on the multiple judgment patterns exemplified in Figures 5(A) to 5(C).

[0043] First, the control device 50 (control unit 51) of the road vehicle detection device 100 acquires an image of the detection range DR from the sensor unit 10, which consists of a camera or the like (step S101). Then, the object detection unit OD (image extraction unit GE) detects (extracts) vehicles GM as targets to be detected (step S102). Furthermore, it commands the velocity vector calculation unit VC to perform a process to determine the velocity vector VV for each extracted vehicle GM (step S103).

[0044] The control device 50 (control unit 51) checks whether there is a vehicle GM whose velocity vector VV has been calculated by the velocity vector calculation unit VC in response to the command in step S103 (step S104). If it is determined in step S104 that a vehicle GM exists (step S104: Yes), it further checks whether there is a stationary vehicle SV (step S105). In other words, it checks whether there is a vehicle GM (stationary vehicle SV) whose vector is zero.

[0045] If it is determined in step S105 that a stationary vehicle SV exists (step S105: Yes), then it is further checked whether there is another vehicle GM besides the stationary vehicle SV (step S106). If such a vehicle GM exists (step S106: Yes), that is, if there is a moving vehicle RV in addition to the stationary vehicle SV, the determination unit JD (roadside parking determination unit PD) checks whether the moving vehicle RV has overtaken and passed the stationary vehicle SV (step S107). In other words, a determination is made as to whether the situation is as illustrated in Figure 5(A). If an overtaking / passing as illustrated in Figure 5(A) is confirmed in step S107 (step S107: Yes), the determination unit JD (roadside parking determination unit PD) determines that there is roadside parking (step S108).

[0046] On the other hand, if it is determined in step S104 that no vehicle GM exists (step S104: No), that is, if no vehicle GM exists in the detection range DR, or if it is determined in step S105 that no stationary vehicle SV exists (step S105: No), that is, if a vehicle GM exists in the detection range DR but there is no stationary vehicle like SV (vector zero), then the determination unit JD (on-street parking determination unit PD) determines that there is no on-street parking (step S109).

[0047] In addition to the above, for example, in step S107, if no overtaking or passing is confirmed (step S107: No), that is, if there are moving vehicles RV in addition to the stopped vehicle SV, but no overtaking or passing of the stopped vehicle SV by the moving vehicles RV is confirmed, then it is checked whether all vehicles GM other than the stopped vehicle SV have started moving (are detected as moving vehicles RV) or not (step S110). In other words, a determination is made as to whether or not the situation is as illustrated in Figure 5(B). In step S110, if all vehicles GM other than the stopped vehicle SV have started moving (are detected as moving vehicles RV) as illustrated in Figure 5(B), the determination unit JD (on-street parking determination unit PD) determines that there is on-street parking (step S111).

[0048] On the other hand, if it is not detected in step S111 that all vehicles GM other than the stopped vehicle SV have started moving, for example, if it is thought that a traffic jam or the like is occurring ahead of the detection range DR, the determination unit JD (on-street parking determination unit PD) makes a determination to that effect (step S112). In other words, in this case, the determination unit JD (on-street parking determination unit PD) does not determine that there is on-street parking.

[0049] Furthermore, for example, in step S106, if there are no vehicles GM other than the stopped vehicle SV in the detection range DR (step S106: No), that is, if there are no moving vehicles RV other than the stopped vehicle SV, it is checked whether or not there is an empty space EP in front of the stopped vehicle SV (step S113). In other words, a determination is made as to whether or not the situation is as illustrated in Figure 5(C). In step S113, if an empty space EP exists as illustrated in Figure 5(C), the determination unit JD (street parking determination unit PD) determines that there is street parking (step S114).

[0050] On the other hand, if there is no available space EP in step S113, for example, if there is a stopped vehicle SV at the very front of the detection range DR, or for example, if it is thought that there is a traffic jam or the like in front of the detection range DR, the determination unit JD (on-street parking determination unit PD) makes a determination to that effect (step S112). In other words, in this case, the determination unit JD (on-street parking determination unit PD) does not determine that there is on-street parking.

[0051] Furthermore, if the result is as described in step S112 above, the road vehicle detection device 100, which is a driver assistance device SS, may, for example, transmit this result to the autonomous vehicle VE as an indication of the current state of the detection range DR, instead of determining whether the vehicle is parked on the road.

[0052] As described above, the road vehicle detection device 100 of this embodiment includes a speed vector calculation unit VC that calculates the speed vector VV of the vehicle GM within the detection range DR on the road, and a determination unit JD (road parking determination unit PD) that determines the presence or absence of road parking based on the relationship of the speed vector VV within the detection range DR calculated by the speed vector calculation unit VC. In the road vehicle detection device 100, the presence or absence of road parking is determined based on the speed vector VV of the vehicle GM within the detection range DR, which allows for faster determination compared to cases where the presence or absence of road parking is determined based on, for example, the duration of a red light.

[0053] [Second Embodiment] Hereinafter, an example of a road vehicle detection device according to the second embodiment will be described with reference to Figure 7. Figure 7 is a conceptual diagram for illustrating the overview of the road vehicle detection device 200 of this embodiment, and corresponds to Figure 1. This embodiment differs from the first embodiment in that the detection range is further extended from the detection range DR that detects road parked vehicles. Specifically, in the example shown in Figure 7, in addition to the detection range DR, the extended range DRe, which is the area in front of the vehicle GM in the figure, is also checked. In the example shown, the extended range DRe includes the intersection CS, but it is not limited to this, and various areas can be the extended range DRe.

[0054] A typical example of a detection range DR is, as mentioned above, a location where on-street parking is common. On the other hand, areas where stopping or parking is prohibited under the Road Traffic Act, such as intersections CS, are unlikely to be selected as detection ranges DR because on-street parking is not expected to occur there. However, even if the area to be designated as a detection range DR is a no-stopping zone, depending on the situation further ahead, there is a possibility that a vehicle may be stopped in the detection range DR on the upstream side, even though it is not on-street parking. Therefore, the on-street vehicle detection device 200 of this embodiment checks the traffic conditions not only in the detection range DR, which is the target of on-street parking detection, but also in the extended range DRe, which is the area beyond the detection range DR that is not the target of on-street parking detection, and makes a determination of whether or not there is on-street parking in the detection range DR, taking into account the traffic conditions in the extended range DRe.

[0055] In other words, as shown in the figure, the area outside the detection range DR, including the intersection CS and its surroundings, is defined as the extended range DRe, and the sensor unit 10 of the road vehicle detection device 200 performs sensing, i.e., imaging and distance measurement, in addition to the detection range DR, up to the extended range DRe. That is, for example, it is possible to acquire information not only from the continuous image GG captured in the detection range DR, but also from the continuous image GGe captured in the extended range DRe. In the extended range DRe, for example, a traffic light TL is installed at the intersection CS, and the road vehicle detection device 200 can understand the color of the traffic light TL and the traffic situation within the intersection CS by performing analysis processing in the control device 50.

[0056] In this case, for example, as shown in the illustration, if a stationary vehicle SV is located at the very front of the detection range DR, it is possible to determine whether or not the stationary vehicle SV is parked on the road based on the analysis results of the traffic conditions in the extended range DRe, i.e., the continuous image GGe.

[0057] As described above, in this embodiment as well, the presence or absence of on-street parking can be determined quickly by determining the presence or absence of on-street parking based on the vehicle GM velocity vector VV in the detection range DR. In particular, in this embodiment, by detecting in the extended range DRe, which is an extension of the detection range DR, it becomes possible to make a more accurate determination regarding the presence or absence of on-street parking in the detection range DR.

[0058] [Third Embodiment] Hereinafter, with reference to Figure 8, an example of a road vehicle detection device according to the third embodiment will be described. In this embodiment, when determining whether or not there is a parked vehicle on the road, an AI (artificial intelligence) that has been trained on the presence or absence of parked vehicles on the road is used. This differs from the first embodiment, etc., in that the determination is made based on a classification of the determination pattern according to the characteristics of the detection range DR as needed. Specifically, as shown in Figure 8(A), an example of a configuration in the learning stage (learning phase) is shown, machine learning is performed to create a trained model for use in the road vehicle detection device 300, and as shown in Figure 8(B), the created trained model TM is incorporated into the road parking determination unit PD. Note that Figure 8(B) is a block diagram corresponding to Figure 2.

[0059] As an example of a machine learning model shown in Figure 8(A), existing forms such as convolutional neural networks (CNNs) can be used, and other methods such as ViT (Vision Transformers) can also be considered. In the illustrated example, various image data GD, such as continuous images GG (see Figure 4, etc.) captured around the detection range DR and partial image data such as vehicle GM extracted from them, as well as velocity vector data VD, which is data related to velocity vectors VV (see Figure 1, etc.), are prepared in advance. Based on these, a trained model TM that serves as the judgment criterion is created, with a human making appropriate adjustments (tune-ups) as needed. In other words, a criterion for determining the confidence level of whether or not the detection range DR is in a state of on-street parking is created through supervised learning. As a result, the judgment unit JD (on-street parking judgment unit PD) determines the presence or absence of on-street parking based on the continuous images GG and the trained model TM related to velocity vectors VV around the detection range DR. Although the above assumes that the trained model TM is created through supervised learning, it is not limited to this, and the trained model TM may also be created through unsupervised learning. Furthermore, the image data (GD) and velocity vector data (VD) used for training can also be configured in various ways other than those described above.

[0060] As described above, in this embodiment as well, the presence or absence of on-street parking can be determined quickly by determining the presence or absence of on-street parking based on the vehicle GM's velocity vector VV within the detection range DR. In particular, in this embodiment, by incorporating a trained model TM, the presence or absence of on-street parking can be determined more quickly and accurately.

[0061] 〔others〕 This invention is not limited to the embodiments described above, and can be implemented in various forms without departing from its spirit.

[0062] First, regarding the locations where the road vehicle detection device 100 is introduced, it is not limited to those mentioned above, but can be applied to various locations where the automated driving vehicle VE requires information. Furthermore, the driver assistance system SS may be configured to perform various detections, such as detecting pedestrians suddenly appearing or the movement of vehicles in the opposite lane, in addition to determining the presence or absence of parked vehicles as described above, as part of the driver assistance system SS, with the road vehicle detection device 100 determining the presence or absence of parked vehicles being incorporated as a part of the driver assistance system SS.

[0063] Furthermore, the road vehicle detection device 100 or the driver assistance device SS may also make judgments about the road surface conditions, in addition to determining whether or not there are parked vehicles on the road. In this case, for example, it is conceivable that the device may predict, based on changes in the velocity vector VV, the state of puddles on the road surface due to rain, the presence or absence of obstacles such as fallen trees, and whether or not it is necessary to avoid construction work.

[0064] Furthermore, in the above, the target vehicle for information provision by the road vehicle detection device 100 is an autonomous driving vehicle VE that transmits future location information, and the future location information is used as a trigger for providing information (driving assistance) regarding on-street parking. However, the trigger is not limited to future location information; various driving information received from the autonomous driving vehicle can also be used as a trigger. Moreover, any vehicle other than an autonomous driving vehicle can receive information (driving assistance) as long as it is able to communicate with the road vehicle detection device 100. In addition, the road vehicle detection device 100 may also be configured to constantly transmit information indicating the current traffic conditions in the detection range DR to a certain area upstream of the detection range DR, so that autonomous driving vehicles and the like can receive it.

[0065] Furthermore, while the above description mainly focused on an example of situation assessment performed on image data (sequential images GG), a variety of data can be expected to be acquired by the sensor unit 10. For example, in addition to the LiDAR for distance measurement described above, the sensor unit 10 could also employ millimeter-wave sensors or radar. For instance, it could be configured to acquire the vehicle GM's position by performing distance measurement and generating distance measurement data.

[0066] Furthermore, the sensor unit 10 in the above example may consist of a camera and a distance measuring device, but for example, the sensor unit 10 may consist of either one of these.

[0067] Furthermore, although the above description assumes that one road vehicle detection device 100 provides information (driving assistance) to one autonomous vehicle VE that is the target of support, information may be provided by two or more road vehicle detection devices 100. Moreover, it is conceivable that two or more road vehicle detection devices 100 may work in cooperation with each other. For example, in the second embodiment, one road vehicle detection device may detect within the detection range DR, another road vehicle detection device may detect within the extended range DRe, and the other road vehicle detection device may transmit the detection results to the first road vehicle detection device.

[0068] Furthermore, the timing of the calculation of the vehicle GM's velocity vector VV in the velocity vector calculation unit VC can also be set in various ways. For example, it could be performed continuously, or only when the displacement of the vehicle GM is confirmed. [Explanation of symbols]

[0069] 10...Sensor unit, 50...Control device, 51...Control unit, 70...Communication unit, 100, 200, 300...Road vehicle detection device, CS...Intersection, DR...Detection range, DRe...Extended range, DS...Data storage unit, EP...Empty space, FM...Frame, FP1~FP4...Point, GD...Image data, GE...Image extraction unit, GG...Continuous image, GG1,GG2,GG3...Image, GGe...Continuous image, GM,GM1,GM2...Vehicle, JD...Determination unit, OD...Object detection unit, PD...Road parking determination unit, PV...Parked vehicle, RV...Moving vehicle, SS...Driving assistance device, SV...Stopped vehicle, T...Time, TL...Traffic light, TM...Trained model, VC...Velocity vector calculation unit, VD...Velocity vector data, VE...Autonomous driving vehicle, VV,VV1,VV2...Velocity vector

Claims

1. A velocity vector calculation unit calculates the vehicle's velocity vector within the detection range on the road, Based on the relationship of the velocity vectors within the detection range calculated by the velocity vector calculation unit, a determination unit determines whether or not there is a parked car on the road. Equipped with, The aforementioned determination unit, upon confirming the presence of another vehicle that has overtaken a stopped vehicle whose velocity vector is below a predetermined value, based on the change in the velocity vector of the other vehicle, determines that the stopped vehicle is parked on the road.

2. The road vehicle detection device according to claim 1, wherein the determination unit determines whether or not there is a parked vehicle on the road based on a learned model relating to velocity vectors.

3. The roadside vehicle detection device according to claim 1, wherein the determination unit determines, in the speed vector calculation unit, that the stopped vehicle is parked on the road if the speed vectors of all vehicles in front of the stopped vehicle are greater than a predetermined value.

4. The roadside vehicle detection device according to claim 1, wherein the determination unit determines that the stopped vehicle is parked on the road if there is an empty space in front of the stopped vehicle.

5. The system includes an image extraction unit that extracts a vehicle from a series of images captured within the aforementioned detection range. The road vehicle detection device according to claim 1, wherein the velocity vector calculation unit calculates a velocity vector from the changes in the vehicle as a series of images extracted by the image extraction unit.

6. The road vehicle detection device according to claim 1, further comprising a communication unit that returns to the autonomous vehicle the determination result of the determination unit regarding the traffic conditions within the detection range, in accordance with the driving information received from the autonomous vehicle.

Images