Roadside sensing method and roadside sensing equipment

Abstract

The roadside sensing method comprises the steps that whether a target vehicle with the volume exceeding a preset threshold value exists in a road or not is determined; if it is determined that the target vehicle exists in the road, detecting a second sensing object caused by pseudo reflection in a predetermined range around a first sensing object corresponding to the target vehicle in a plurality of sensing objects, the plurality of sensing objects are sensing results generated based on point cloud data from a millimeter wave radar sensor; and processing the detected second sensing object so as to filter a sensing result caused by the pseudo reflection. Therefore, the influence of pseudo reflection is removed more robustly, and the perception performance based on millimeter wave radar point cloud data is improved.

Description

Technical Field

[0001] The present invention relates to the field of data security, and in particular to the processing of image data. Background Art

[0002] Roadside perception uses a variety of sensors such as visual sensors, millimeter-wave radars, and lidars, combined with computing devices, to achieve real-time acquisition of current road traffic participants and road conditions. Radar, especially millimeter-wave radar, has multipath transmission, which is prone to false reflections, resulting in erroneous detection results.

[0003] There are several ways to remove false detections due to false reflections when using mmWave radar on the vehicle side. One method is that the false detections are identified as objects at the perception layer, but are removed at the function layer after combining the vehicle / road conditions. Another method is to move the mmWave radar sensor itself to limit its detection range.

[0004] However, these methods cannot obtain robust detection results when the millimeter wave radar sensor is installed on the roadside or are not applicable to the scenario where the millimeter wave radar sensor is installed on the roadside. Summary of the invention

[0005] It is desired to provide a roadside perception method and a corresponding roadside perception device, which can more robustly remove the influence of false reflections and improve the perception performance based on radar point cloud data.

[0006] According to one aspect, a roadside perception method is provided, comprising: determining whether there is a target vehicle whose volume exceeds a predetermined threshold in a road; if it is determined that the target vehicle exists in the road, detecting a second perception object caused by pseudo reflection within a predetermined range around a first perception object corresponding to the target vehicle among multiple perception objects, wherein the multiple perception objects are perception results generated based on point cloud data from a radar sensor; and processing the detected second perception object to filter out the perception results caused by the pseudo reflection.

[0007] According to another aspect, a roadside perception device is provided, including: a target vehicle determination unit, used to determine whether there is a target vehicle whose volume exceeds a predetermined threshold in the road; a pseudo-reflection detection unit, if the target vehicle determination unit determines that the target vehicle exists in the road, the pseudo-reflection detection unit is used to detect a second perception object caused by pseudo-reflection within a predetermined range around a first perception object corresponding to the target vehicle among multiple perception objects, wherein the multiple perception objects are perception results generated based on point cloud data from a radar sensor; and a pseudo-reflection filtering unit, used to process the second perception object to filter out the perception results caused by the pseudo-reflection.

[0008] According to another aspect, a roadside perception system is provided, comprising: a roadside perception device according to various embodiments of the present invention; an image sensor for acquiring an image of a road; and a radar sensor for acquiring point cloud data.

[0009] According to still another aspect, a computer-readable medium is provided, storing a computer program, wherein when the computer program is executed by a computer or a processor, the computer or the processor executes the method according to various embodiments of the present invention.

[0010] According to various embodiments of various aspects of the present invention, it has been found that large vehicles (such as trucks or vans) on the road are prone to pseudo reflections, thereby producing erroneous perception results. Therefore, large vehicles on the road are first detected, and then objects that are erroneously perceived due to pseudo reflections are detected within a predetermined range around the detected large vehicles. After such objects are detected, the erroneously perceived objects are processed to filter out the influence of pseudo reflections. This can remove pseudo reflections more robustly, improve perception performance based on radar point cloud data, and also reduce computational costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Embodiments are illustrated by way of example only and not by way of limitation in the figures of the accompanying drawings in which like reference numerals refer to similar elements.

[0012] Figure 1 An example of a spurious reflection is shown;

[0013] Figure 2 and Figure 3 The perception results based on point cloud data from the millimeter-wave radar sensor in similar scenes are shown respectively;

[0014] Figure 4 A flow chart of a roadside perception method according to an embodiment of the present invention is shown;

[0015] Figure 5 A block diagram of a roadside sensing device according to an embodiment of the present invention is shown;

[0016] Figure 6 shows the perception results according to one embodiment of the present invention;

[0017] Figure 7 A block diagram of an electronic device for executing a method for processing image data according to an embodiment of the present invention is shown.

[0018] Various aspects and features of various embodiments of the present invention are described with reference to the above drawings. The above drawings are only illustrative and not restrictive. In addition, not all parts of the device of the embodiment of the present invention are marked with reference numerals in the above drawings. Only relevant parts are marked in some drawings, which does not limit the parts to those shown in the drawings of the specification. DETAILED DESCRIPTION

[0019] Millimeter-wave radar can be set at the end of the road to take advantage of its relatively accurate measurement properties for information such as speed. When the millimeter-wave radar is set at the end of the road, such as at a high place, its detection range is wider and farther than that of the vehicle-side millimeter-wave radar. Objects are perceived based on point cloud data (e.g., after clustering) from the millimeter-wave radar sensor, and each perceived object corresponds to a group of points in the point cloud data (e.g., clustered points). These perceived objects can then be semantically segmented to achieve target detection functions. Typically, these perceived objects correspond to real traffic participants on the road, such as vehicles.

[0020] However, the transmission of millimeter-wave radar is multipath, which is prone to pseudo-reflection, which may lead to erroneous perception results. When pseudo-reflection occurs, the point cloud data from the millimeter-wave radar sensor consists of two parts, one part represents the real reflection from the traffic participants, and the other part is generated due to pseudo-reflection. Figure 1 An example of a spurious reflection is shown in Figure 1 As shown in (a), in the actual scene, above the truck A is the small truck B, and there is no other vehicle between the truck A and the small truck B. Figure 1 As shown in (b), there are perceived objects as shown in two blocks between the perceived object A' corresponding to the truck A and the perceived object B' corresponding to the pickup truck B. The perceived objects in these two blocks are generated due to pseudo-reflections and are not real traffic participants on the road.

[0021] Although objects that are incorrectly perceived in the perception layer due to pseudo-reflections may be removed in the function layer, this is not generally effective. In addition, pseudo-reflections may also lead to incorrect judgments of spatial occupancy information, resulting in possible misjudgments in the downstream path planning module. Figure 2 and Figure 3 The perception results based on point cloud data from a millimeter-wave radar sensor in similar scenarios according to the current perception method are shown respectively. Figure 2 (a) and Figure 3 (a) shows a very similar scenario. Figure 2 (b) and Figure 3 (b) in FIG. 4 shows the perception result based on point cloud data for the corresponding scene. Figure 2As shown by the arrows in (b), although there are point data caused by pseudo reflection in the point cloud data, these point data are not perceived as objects by the current perception method, so no erroneous perception results will be generated. Figure 3 As shown by the arrow in (b), the point data caused by the pseudo-reflection is perceived as an object by the current perception method, resulting in an erroneous perception result. It should be noted that in actual annotation, the point data caused by the pseudo-reflection that is not perceived and the point data caused by the pseudo-reflection that is perceived are Figure 2 and Figure 3 can be distinguished in different ways. Figure 2 and Figure 3 It can be seen that the current perception method is not robust to remove the influence of pseudo reflections. If the influence of pseudo reflections cannot be filtered out, not only will the wrong perception results be obtained, leading to perception failure, but also the semantic segmentation of radar point cloud data will fail, which will lead to the failure of target detection or the failure of occupancy information judgment, resulting in wrong results in downstream modules.

[0022] The inventors of the present invention have found that large vehicles (such as trucks or vans) on the road are prone to pseudo-reflection, thereby producing erroneous perception results; in particular, large vehicles with a relatively high number of reflection points are prone to pseudo-reflection. Based on various embodiments of the present invention, large vehicles on the road are first detected, and then objects that are mistakenly perceived due to pseudo-reflection are detected within a predetermined range around the detected large vehicles. After such objects are detected, the mistakenly perceived objects are processed to filter out pseudo-reflections. This can more robustly remove the influence of pseudo-reflections, improve the perception performance based on millimeter-wave radar point cloud data, and also reduce computational costs.

[0023] Figure 4 FIG. 1 is a flow chart of a roadside perception method 100 for a millimeter wave radar sensor according to an embodiment of the present invention. Figure 4 As shown, in step 110, point cloud data of the road detected by the millimeter wave radar sensor is obtained. In step 120, perception processing is performed based on the point cloud data. Specifically, for example, each point in the point cloud data is clustered to generate multiple groups of clustered point data, and each group of clustered point data can be correspondingly perceived as a perception object. As mentioned above Figure 1 As shown, the point cloud data is sensed to generate sensed objects A′ and B′ corresponding to vehicles A and B. Step 120 is not necessary. Step 130 may also be performed directly after step 110 .

[0024] In step 130, it is determined whether there is a target vehicle whose volume exceeds a predetermined threshold, that is, a large target vehicle that easily causes false reflection. Whether there is a target vehicle on the road can be detected from at least the following two aspects.

[0025] In one aspect, a large vehicle has a large volume and relatively more reflection points, so the number of points detected for the large vehicle in the point cloud data is greater, and it is possible to determine whether the perceived object represents a large vehicle on the road based on the number of points corresponding to each perceived object.

[0026] Specifically, the number of points corresponding to each of the plurality of perceived objects as the perception result in the point cloud data is determined; the number of points is compared with a first predetermined threshold; if the number of points corresponding to a specific perceived object exceeds the first predetermined threshold, the specific perceived object is determined to be a target vehicle, that is, it is determined that there is a target vehicle whose volume exceeds the predetermined threshold in the road. According to this aspect, the above step 120 needs to be performed.

[0027] Because when detecting vehicles on the road, the number of points detected by the millimeter-wave radar is not only related to the size of the vehicle, but also to the distance between the vehicle and the millimeter-wave radar sensor and the installation and working parameters of the millimeter-wave radar sensor, the first predetermined threshold can be set based on human experience, and various parameters that affect the number of points need to be considered during the setting; in addition, the first predetermined threshold can also be obtained by using a machine learning model based on supervised or unsupervised training for the point cloud data collected by the current millimeter-wave radar sensor.

[0028] In one embodiment, the first predetermined threshold may be set as a threshold of the number of points that varies with the distance from the millimeter-wave radar sensor. Specifically, if the currently detected vehicle is at a first distance from the millimeter-wave radar sensor, the first predetermined threshold set for the first distance may be used; the number of points for the currently detected vehicle in the point cloud data is compared with the first predetermined threshold for the first distance. Different first predetermined thresholds are set in consideration of the distance of the millimeter-wave radar sensor, and the number of points corresponding to each perceived object is compared with the first predetermined threshold at the corresponding distance, which can accurately identify perceived objects with a large number of points, thereby correspondingly identifying the presence of large vehicles.

[0029] In another embodiment, the number of points corresponding to each sensed object determined is the number of points within a predetermined area, and correspondingly, the first predetermined threshold is also a threshold of the number of points set for different predetermined areas. The more points there are within the same area, the more reflection points the vehicle corresponding to the sensed object has, and the predetermined area can be set with reference to the area of ​​a large vehicle. Generally, traffic participants with a large number of reflection points are more likely to cause pseudo-reflections, and considering the density of reflection points when setting the first predetermined threshold may more accurately identify traffic participants that cause pseudo-reflections.

[0030] Although the first predetermined threshold is described as variable with reference to the above two embodiments, this does not exclude setting the first predetermined threshold to a constant value. In fact, the first predetermined threshold can also be set to a constant. In addition, the above two embodiments can also be used in combination.

[0031] In another aspect, an image sensor may be used to determine whether there is a target vehicle whose volume exceeds a predetermined threshold on the road. Specifically, an image sensor is used to obtain an image of the road; then, based on the image, it is determined whether there is a target vehicle whose volume exceeds a predetermined threshold on the road. This may be achieved using a multimodal sensor including the functions of a millimeter wave radar sensor and an image sensor. According to this aspect, the above step 120 may not be performed first, and the perception process based on point cloud data may be performed after the target vehicle is detected in step 130.

[0032] The above describes how to determine whether there is a target vehicle whose volume exceeds a predetermined threshold in step 130 with reference to two aspects, that is, on the one hand, determining whether the perceived object represents a large vehicle on the road according to the number of points corresponding to each perceived object, and on the other hand, determining whether there is a target vehicle whose volume exceeds the predetermined threshold on the road with the help of an image sensor. Either of the two methods can be used to determine whether there is a target vehicle whose volume exceeds the predetermined threshold on the road, or a combination of the two methods can be used to determine whether there is a target vehicle whose volume exceeds the predetermined threshold on the road, so as to improve the robustness of the target vehicle determination.

[0033] If it is determined in step 130 that there is no target vehicle ("N") on the road, that is, there is no large vehicle that may cause false reflections, then the perception result obtained in step 120 (if executed) is reliable, and further processing (not shown) can be performed based on the perception result, and at the same time, the process can return to step 110 to continue acquiring point cloud data.

[0034] If it is determined in step 130 that there is a target vehicle ("Y") on the road, then in step 140, a second perceived object caused by pseudo-reflection is detected within a predetermined range around the first perceived object corresponding to the target vehicle among the multiple perceived objects obtained in step 120. Specifically, after the target vehicle is determined, the first perceived object corresponding to the detected target vehicle can be determined from the multiple perceived objects perceived based on the point cloud data, and the predetermined range around the first perceived object can be determined, and then the second perceived object caused by pseudo-reflection can be detected within the predetermined range. The predetermined range can correspond to a range within 20 meters around the target vehicle, for example. Different predetermined ranges can be determined for different roads.

[0035] In one embodiment, an image sensor can be used to assist in detecting a second perceived object caused by a pseudo-reflection. An image of the road is acquired using an image sensor; then, based on the image, the second perceived object caused by the pseudo-reflection is detected within a predetermined range around the first perceived object corresponding to the target vehicle. Specifically, for example, the image sensor of a camera can generate an image showing the surroundings of the target vehicle (e.g., a bird's-eye view), and based on the image, an image target is detected in a detection area around the target vehicle, and the detection area corresponds to a predetermined range around the first perceived object of the corresponding target vehicle; accordingly, a radar target (e.g., a perceived object or a point cloud cluster detected based on the point cloud data) can be detected within the above-mentioned predetermined range based on the point cloud data. If the detected image target and the radar target match each other, it indicates that the corresponding radar target is not caused by a pseudo-reflection; on the contrary, it indicates that the corresponding radar target is caused by a pseudo-reflection. Figure 1 For example, by comparing images (a) and (b), there is no image target in the image (a) that matches the perceived object indicated by the two squares in (b), so it is easy to determine that the perceived object indicated by the two squares in (b) is caused by pseudo reflection. When using image sensors to assist in detecting perceived objects caused by pseudo reflection, the image sensors and millimeter wave radar sensors need to be well calibrated.

[0036] When a second perceived object caused by a false reflection within a predetermined range is detected, in step 150, the second perceived object is processed to filter out the false reflection. Specifically, in one embodiment, the second perceived object can be removed from the plurality of perceived objects as the perception result. Figure 1 For example, the perceived object indicated by two squares in (b) will no longer be recognized as a perceived object.

[0037] Although the description is made with reference to using an image to assist in detecting the second perceived object caused by the pseudo reflection, this is not limiting. In one embodiment, after determining in step 130 that there is a target vehicle on the road, an algorithm can be directly used to detect and process the second perceived object caused by the pseudo reflection from the perception result of step 120 to filter out the pseudo reflection.

[0038] In the field of robotics and automobiles, occupancy grid maps are a recognized perception method, in which radar occupancy grid maps are maps that describe where traffic participants occupy. The classic occupancy grid map method divides the environment into multiple grid cells and estimates the occupancy probability of each grid cell. In the occupancy grid map method, each point in the point cloud data is represented by multiple particles. The multiple particles can be obtained by randomly scattering points. The dynamic state of the grid cell can be represented based on all the particles in each grid cell. The perception method according to an embodiment of the present invention can also be a method based on the occupancy grid map. In other words, the perception method according to an embodiment of the present invention can be obtained by improving the current occupancy grid map method. The specific perception method of the occupancy grid map will not be repeated here.

[0039] According to an embodiment of the present invention, the occupancy grid map based perception method described below may be performed to detect and process the second perception object caused by the reflection artifact to filter out the pseudo reflection.

[0040] Specifically, a specific perceived object within a predetermined range of a first perceived object corresponding to the target vehicle is determined among a plurality of perceived objects perceived based on point cloud data, and then a grid unit cluster corresponding to the specific perceived object is determined.

[0041] Next, for each grid cell in the grid cell cluster of the specific perceived object, the following operations are performed: determining a parameter representing a speed change of a plurality of particles in the grid cell; if the parameter determined for the current grid cell represents a speed change greater than a predetermined speed-related threshold, determining the state of the grid cell as unoccupied; if the parameter determined for the current grid cell represents a speed change not greater than a predetermined speed-related threshold, determining the state of the grid cell as occupied.

[0042] The parameters representing the speed change may include multiple parameters, such as the covariance of the speed, the variance of the speed, or the covariance of the speed and other parameters (such as distance). In one embodiment, for each grid cell, the variance of the speed component along the x direction is calculated and used as the parameter representing the speed change. Correspondingly, for example, the predetermined speed-related threshold value may be set to 0.4, and when the calculated parameter is greater than the predetermined speed-related threshold value, the corresponding grid cell is set to be unoccupied.

[0043] Although the variance of the velocity component in the x direction is used as an example above, this is only an example, and parameters representing speed changes in other directions are also feasible. For example, the variance of the velocity in the y direction can be used. Among them, the x direction is the direction along the road, and the y direction is the direction perpendicular to the road. For another example, the covariance of the velocity in the x direction and the velocity in the y direction can be used as a parameter representing speed changes. After obtaining the parameter representing speed changes, the parameter can be compared with the corresponding speed-related threshold for judgment.

[0044] After determining whether the state of each grid cell is unoccupied or occupied, it is possible to determine whether the multiple points corresponding to the specific perceived object in the point cloud data are in the unoccupied grid cell or in the occupied grid cell. If it is determined that the multiple points of the specific perceived object are in the unoccupied grid cell, it is determined that the specific perceived object is caused by pseudo reflection, otherwise it is not caused by pseudo reflection. Therefore, the second perceived object caused by pseudo reflection can also be detected based on the occupied grid map algorithm.

[0045] There is a possibility that some of the points in a set of points corresponding to a specific perceived object are in unoccupied grid cells, while another part of the points are in occupied grid cells. In this case, according to the above processing, it is still possible to determine that the specific perceived object is caused by pseudo reflection.

[0046] Those points determined to be caused by spurious reflections can be deleted, thereby changing the perception of the point cloud data in the perception process to filter out the effects of the spurious reflections.

[0047] In the above embodiments, reference is made to improvements to the occupancy grid map method to describe how to detect and process the second perceived object caused by the pseudo reflection. It is also conceivable to directly perform detection based on point cloud data without using the occupancy grid map method.

[0048] In this case, for a specific sensing object within a predetermined range around the target vehicle among multiple sensing objects, a parameter representing the speed change between the points corresponding to the specific sensing object is determined. The parameter of the speed change can be selected as discussed above. For point cloud data, each point corresponds to a detected reflection point. Each point in the point cloud data detected by the millimeter wave radar includes corresponding characteristics. These characteristics include timestamp, distance (variance, divergence), speed (variance, divergence), distance and speed covariance (quantity, diffusion direction), elevation (quantity, variance), azimuth (quantity, variance), radar cross section (radar cross section, rcs, which represents the reflection area of ​​the object), received signal strength index (received signal strength index, RSSI, which represents the strength of the received signal), etc. The parameter representing the speed change can be determined based on one or more of these characteristics.

[0049] If the determined parameter representing the speed change indicates that the speed change between the points corresponding to the specific perceived object in the point cloud data is greater than the predetermined speed-related threshold, the specific perceived object is determined to be a second perceived object caused by pseudo-reflection, otherwise, the specific perceived object is determined to be a real target. This is because there is usually a large speed change between the reflection points caused by pseudo-reflection. By detecting the speed change between the points in the point cloud data, those reflection points with large speed changes can be found, and then it can be determined whether the corresponding perceived object is caused by pseudo-reflection.

[0050] References Figure 1-4 The present invention describes a roadside sensing method for a millimeter wave radar sensor, and in particular, how to filter out the influence of false reflections. Figure 4 The steps shown are only for illustration and are not required. In addition, the order of some of the steps may be changed as needed. In addition, the perception method of the present invention may also be applicable to point cloud data generated by other radar sensors such as laser radar.

[0051] Figure 5 FIG. 1 shows a block diagram of a roadside sensing device 10 according to an embodiment of the present invention. Figure 5 As shown, the roadside perception device 10 includes: a target vehicle determination unit 11, a false reflection detection unit 12 and a false reflection filtering unit 13.

[0052] The target vehicle determination unit 11 is used to determine whether there is a target vehicle on the road with a volume exceeding a predetermined threshold.

[0053] In one embodiment, it is possible to determine whether a target vehicle exists based on the number of points corresponding to each perceived object in the point cloud data. Specifically, the target vehicle determination unit 11 determines the number of points corresponding to each perceived object in the point cloud data; compares the number of points with a first predetermined threshold; if the number of points corresponding to a specific perceived object exceeds the first predetermined threshold, the specific perceived object is determined as a target vehicle. In one example, the number of points may be the number of points within a predetermined area, and correspondingly, the first predetermined threshold is a threshold for the number of points in the predetermined area. In another example, the first predetermined threshold is a threshold for the number of points that varies with the distance to the millimeter-wave radar sensor. The corresponding first predetermined threshold can be determined based on the distance to the millimeter-wave radar sensor corresponding to each point.

[0054] In another embodiment, it is possible to determine whether a target vehicle exists based on an image. Specifically, an image sensor is used to acquire an image of a road, and the target vehicle determination unit 11 determines whether a target vehicle having a volume exceeding a predetermined threshold value exists on the road based on the acquired image. In this case, the image sensor and the millimeter wave radar sensor should be calibrated in advance so that the mapping relationship between the image acquired by the image sensor and the point cloud data acquired by the millimeter wave radar sensor is known.

[0055] If the target vehicle determination unit 11 determines that there is a target vehicle on the road, the pseudo-reflection detection unit 12 detects a second perception object caused by pseudo-reflection within a predetermined range around a first perception object corresponding to the target vehicle among multiple perception objects, wherein the multiple perception objects are generated based on point cloud data from a millimeter-wave radar sensor.

[0056] The detection of the second perceived object caused by the false reflection may be performed at least by the following embodiments.

[0057] According to one embodiment, a second perceived object caused by a pseudo reflection among a plurality of perceived objects can be detected based on an image. Specifically, if the target vehicle determination unit 11 determines that there is a target vehicle in the road, the pseudo reflection detection unit 12 detects the second perceived object caused by the pseudo reflection within a predetermined range around the first perceived object corresponding to the target vehicle based on an image, the image being an image of the road acquired using an image sensor.

[0058] According to another embodiment, a second perceived object caused by pseudo-reflection among multiple perceived objects can be detected based on an occupied grid map. Specifically, the pseudo-reflection detection unit 12 determines a grid cell cluster corresponding to a specific perceived object within a predetermined range; for each grid cell in the grid cell cluster of the specific perceived object: determining a parameter representing a speed change, if the parameter indicates that the speed change is greater than a predetermined speed-related threshold, setting the state of the grid cell to unoccupied, if the parameter indicates that the speed change is not greater than a predetermined speed-related threshold, setting the state of the grid cell to occupied; determining whether multiple points corresponding to the specific perceived object are in unoccupied grid cells or in occupied grid cells, if it is determined that multiple points corresponding to the specific perceived object are in unoccupied grid cells, determining that multiple points of the specific perceived object are caused by the pseudo-reflection.

[0059] According to another embodiment, it is possible to determine whether the corresponding perceived object is caused by pseudo reflection based on the fact that there is usually a large speed change between each reflection point caused by pseudo reflection. Specifically, for a specific perceived object in a predetermined range around the target vehicle among the multiple perceived objects, a parameter representing the speed change between each point corresponding to the specific perceived object is determined. If the parameter represents that the speed change between each point corresponding to the specific perceived object in the point cloud data is greater than a predetermined speed-related threshold, the specific perceived object is determined as a second perceived object caused by pseudo reflection.

[0060] After detecting the second perceived object caused by the pseudo reflection, the pseudo reflection filtering unit 13 processes the second perceived object to filter out the pseudo reflection. For example, the point caused by the pseudo reflection may be deleted. Or the perceived object caused by the pseudo reflection may be removed from the perception result.

[0061] Figure 6 The roadside sensing method according to the present invention is shown. Figure 1 The result of perceiving the scene in (a). Figure 6 As shown, Figure 1 Compared with (b), the point cloud cluster between the perceived object A' corresponding to truck A and the perceived object B' corresponding to pickup truck B is no longer perceived as an object. Thus, the perceived objects caused by the pseudo reflection are removed.

[0062] Although the reference Figure 5A roadside perception device 10 according to the present invention is shown, which is not restrictive, and one or more components therein may be omitted, merged, split and / or changed, as long as the technical solution under the concept of the present invention can be implemented. The above-mentioned units can be implemented by computer programs and / or corresponding hardware. In a scenario, the corresponding computer program is executed by the same or different processors to implement the functions of the corresponding units. Alternatively, the above-mentioned units may also be implemented by hardware, such as a dedicated processor, controller, etc. Any setting and arrangement of the components of the various units that can implement the concept of the present invention should be included in the equivalent scope of the technical solution defined in the claims of the present invention. The roadside perception device 10 may also include an acquisition unit (not shown), which acquires point cloud data of the road generated by the radar sensor.

[0063] In one embodiment, a roadside perception system may be designed, which includes the roadside perception device 10 according to the above embodiments, a millimeter wave radar sensor, and an image sensor. The millimeter wave radar sensor and the image sensor may be implemented using a multimodal sensor including the functions of both the millimeter wave radar sensor and the image sensor.

[0064] Figure 7 FIG. 5 is a block diagram of an electronic device 500 for executing a roadside sensing method according to an embodiment of the present invention. Figure 7 As shown, the electronic device 500 may include at least one processor 510 (which may include a central processing unit (CPU) and an image processing unit (GPU)), an image sensor 520, a memory 540, a communication interface 550, a storage (e.g., a non-volatile memory) 560, and a millimeter wave radar sensor, and the above-mentioned components are connected together via a bus 530. At least one processor 510 executes at least one computer-readable instruction (i.e., an element implemented in the form of software) stored or encoded in the memory 560, and the computer-readable instruction can be executed by at least one processor to implement the roadside perception method according to various embodiments of the present invention.

[0065] References Figure 1-7 Various embodiments of the roadside sensing method and device of the present invention are described, and these embodiments can be combined with each other to obtain different effects without being limited by the subject type. Furthermore, the various units / steps / processes mentioned above are not restrictive, and the functions of the various units / steps / processes mentioned above can be merged / combined / changed / modified to obtain corresponding effects.

[0066] The functions of each unit of the method and system of the present invention can be implemented by software or corresponding hardware, and the functions of the above-mentioned units are implemented by means of a processor such as a GPU and / or a CPU. For example, the processor can read the computer program in the memory and run these computer programs to implement the functions of the above-mentioned units. In one embodiment, the above-mentioned electronic device of the present invention can also be implemented by a memory and a processor.

[0067] It can be understood that the methods of various embodiments of the present disclosure can be implemented by computer programs / software. These software can be loaded into the working memory of the processor and used to execute the methods according to various embodiments of the present disclosure when running.

[0068] Exemplary embodiments of the present disclosure cover both creating / using the computer program / software of the present disclosure from the outset and converting an existing program / software to use the computer program / software of the present disclosure by means of an update.

[0069] According to another embodiment of the present disclosure, a computer program product is provided, such as a machine (e.g., computer) readable medium, such as a CD-ROM, including a computer program code, which, when executed, causes a computer or a processor to perform a method according to each embodiment of the present disclosure. The machine readable medium is, for example, an optical storage medium or a solid-state medium supplied with or as part of other hardware.

[0070] The above describes specific embodiments of the present disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0071] The present disclosure is described above with reference to specific embodiments. It should be understood by those skilled in the art that the technical solution of the present disclosure can be implemented in various ways without departing from the spirit and basic features of the present disclosure. The specific embodiments are merely illustrative and not restrictive. In addition, these embodiments can be arbitrarily combined to achieve the purpose of the present disclosure. The scope of protection of the present disclosure is defined by the attached claims.

[0072] The word "comprising" in the specification and claims does not exclude the existence of other elements or steps. The expressions "first", "second" and the order of the steps shown in the figures do not limit their order or quantity. The functions of the various elements described in the specification or recorded in the claims can also be split or combined and implemented by corresponding multiple elements or a single element.

Claims

1. A roadside perception method, comprising: Determining whether there is a target vehicle on the road whose volume exceeds a predetermined threshold; If it is determined that the target vehicle exists in the road, detecting a second perception object caused by a pseudo reflection within a predetermined range around a first perception object corresponding to the target vehicle among a plurality of perception objects, wherein the plurality of perception objects are perception results generated based on point cloud data from a radar sensor; and The detected second perception object is processed to filter out the perception result caused by the false reflection.

2. The roadside sensing method according to claim 1, further comprising: determining a number of points in the point cloud data corresponding to each of the plurality of perceived objects; comparing the number of points to a first predetermined threshold; If the number of points corresponding to the specific perception object exceeds the first predetermined threshold, the specific perception object is determined to be the target vehicle.

3. The roadside sensing method according to claim 2, wherein: The number of points is the number of points within a predetermined area, and the first predetermined threshold is a threshold for the number of points in the predetermined area; or, The first predetermined threshold is a threshold of the number of points that varies with the distance from the radar sensor.

4. The roadside sensing method according to claim 1, further comprising: The point cloud data of the road detected by the radar sensor is acquired.

5. According to the roadside sensing method of claim 1, determining whether there is a target vehicle with a volume exceeding a predetermined threshold on the road comprises: Acquiring an image of the road using an image sensor; and It is determined based on the image whether the target vehicle having a volume exceeding a predetermined threshold value exists on the road.

6. The roadside sensing method according to claim 1, further comprising: Acquiring an image of the road using an image sensor; If it is determined that the target vehicle exists on the road, the second perceived object caused by the false reflection is detected within the predetermined range around the first perceived object corresponding to the target vehicle based on the image.

7. The roadside sensing method according to claim 1, wherein: The roadside perception method is based on an occupancy grid map, and the roadside perception method further comprises: Determine a grid cell cluster corresponding to a specific perceived object within the predetermined range; For each grid cell in the grid cell cluster of the specific perceived object, determining a parameter representing a change in velocity of a plurality of particles in the grid cell; if the parameter indicates that the speed change is greater than a predetermined speed-related threshold, determining the state of the grid cell as unoccupied; if the parameter indicates that the speed change is not greater than a predetermined speed-related threshold, determining the state of the grid cell as occupied; determining whether the plurality of points corresponding to the specific sensed object in the point cloud data are in unoccupied grid cells or in occupied grid cells, If it is determined that at least a portion of the multiple points corresponding to the specific perceived object are in unoccupied grid cells, it is determined that the specific perceived object is the second perceived object caused by the false reflection.

8. The roadside sensing method according to claim 1, further comprising: For a specific sensing object within the predetermined range among the plurality of sensing objects, Determining a parameter representing a change in velocity between points corresponding to the specific perceived object; If the parameter indicates that the speed change between the points corresponding to the specific perceived object is greater than a predetermined speed-related threshold, the specific perceived object is determined to be the second perceived object caused by the false reflection.

9. A roadside sensing device, comprising: A target vehicle determination unit, used to determine whether there is a target vehicle with a volume exceeding a predetermined threshold on the road; a pseudo reflection detection unit, if the target vehicle determination unit determines that the target vehicle exists on the road, the pseudo reflection detection unit is used to detect a second perception object caused by pseudo reflection within a predetermined range around a first perception object corresponding to the target vehicle among a plurality of perception objects, wherein the plurality of perception objects are perception results generated based on point cloud data from a radar sensor; and A pseudo-reflection filtering unit is used to process the second perception object to filter out the perception result caused by the pseudo-reflection.

10. The roadside sensing device according to claim 9, wherein: The target vehicle determination unit is also used for: determining a number of points in the point cloud data corresponding to each of the plurality of perceived objects; comparing the number of points to a first predetermined threshold; If the number of points corresponding to the specific perception object exceeds the first predetermined threshold, the specific perception object is determined to be the target vehicle.

11. The roadside sensing device according to claim 10, wherein: The number of points is the number of points within a predetermined area, and the first predetermined threshold is a threshold for the number of points in the predetermined area; or, The first predetermined threshold is a threshold of the number of points that varies with the distance from the radar sensor.

12. The roadside sensing device according to claim 9, further comprising: An acquisition unit is used to acquire the point cloud data of the road detected by the radar sensor.

13. The roadside sensing device according to claim 9, wherein: The target vehicle determination unit is also used for: It is determined whether the target vehicle having a volume exceeding a predetermined threshold value exists on the road based on an image, wherein the image is an image of the road acquired by using an image sensor.

14. The roadside sensing device according to claim 9, wherein: If the target vehicle determination unit determines that the target vehicle exists in the road, the pseudo-reflection detection unit detects the second perceived object caused by the pseudo-reflection within the predetermined range around the first perceived object corresponding to the target vehicle based on an image, wherein the image is an image of the road acquired using an image sensor.

15. The roadside sensing device according to claim 9, wherein: The roadside sensing device performs a roadside sensing method based on an occupancy grid map, and the false reflection detection unit is used to: Determine a grid cell cluster corresponding to a specific perceived object within the predetermined range; For each grid cell in the grid cell cluster of the specific perceived object, determining a parameter representing a change in velocity of a plurality of particles in the grid cell; if the parameter indicates that the speed change is greater than a predetermined speed-related threshold, determining the state of the grid cell as unoccupied; if the parameter indicates that the speed change is not greater than a predetermined speed-related threshold, determining the state of the grid cell as occupied; determining whether the plurality of points corresponding to the specific sensed object in the point cloud data are in unoccupied grid cells or in occupied grid cells, If it is determined that the multiple points corresponding to the specific perceived object are in unoccupied grid cells, it is determined that the specific perceived object is the second perceived object caused by the false reflection.

16. The roadside sensing device according to claim 9, wherein the false reflection detection unit is used for: For a specific sensing object within the predetermined range among the plurality of sensing objects, Determining a parameter representing a change in velocity between points corresponding to the specific perceived object; If the parameter indicates that the speed change between the points corresponding to the specific perceived object is greater than a predetermined speed-related threshold, the specific perceived object is determined to be the second perceived object caused by the false reflection.

17. A roadside perception system, comprising: The roadside sensing device according to any one of claims 9 to 16; An image sensor for acquiring an image of the road; and A radar sensor is used to obtain point cloud data of the road.

18. A computer readable medium storing a computer program which, when executed by a processor or a computer, causes the processor or the computer to perform the method according to any one of claims 1 to 8.

Images