Abnormal point cloud determination method and device, storage medium and vehicle-mounted electronic equipment

Abstract

The invention discloses an abnormal point cloud determination method and device, a storage medium and vehicle-mounted electronic equipment. The method comprises the steps that points, except for the ground, in a to-be-processed point cloud are clustered, clustered point cloud is obtained, and the to-be-processed point cloud is the point cloud of a current frame obtained by measuring laser emitted by a laser radar in the vehicle driving process; point cloud features are extracted from the clustering point cloud according to target feature dimensions, the target feature dimensions at least comprise feature dimensions used for representing relative motion features of points in the clustering point cloud between the current frame and a historical frame, and the historical frame is obtained by measuring laser emitted by the laser radar before the current frame; and when the point cloud features of the clustering point cloud on the target feature dimension meet a preset condition, determining that the clustering point cloud is an abnormal point cloud which is a point cloud generated when the emitted laser passes through water drops on an acquisition window of the laser radar and hits the ground. The technical problem that abnormal point clouds cannot be detected in the prior art is solved.

Description

Technical Field

[0001] The present application relates to the field of autonomous driving, and more specifically, to a method and device for determining an abnormal point cloud, a storage medium, and an on-vehicle electronic device. Background Art

[0002] This section is intended to provide a background or context to what is stated in the claims or specification, and no admission is made that what is described herein is prior art by inclusion in this section.

[0003] There are two main problems when using LiDAR in rainy weather. One is that for raindrops in the air, when the laser emitted by the radar hits the raindrops, it will be refracted and reflected by the target in front. At this time, it is equivalent to a distance measurement for LiDAR, but in fact, it is a noise point that interferes with the distance measurement. The noise points generated by raindrops in the air are usually more discrete, so they can be filtered out from the point cloud by removing outliers. Another is that when raindrops fall on the radar window, a water droplet will be formed on the radar window. In this case, when the LiDAR detects, the emitted laser will be refracted by the water droplets before it can reach the object to be detected, resulting in a change in the energy optical path. The LiDAR solves according to the situation without water drop refraction, resulting in wrong distance measurement, and then generating abnormal point clouds. This point cloud cannot be removed by removing outliers, which will affect the subsequent application of point clouds, and there is currently no solution for its detection.

[0004] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0005] The embodiments of the present application provide a method and device for determining an abnormal point cloud, a storage medium, and an in-vehicle electronic device to at least solve the technical problem in the related art that abnormal point clouds cannot be detected.

[0006] According to one aspect of an embodiment of the present application, a method for determining an abnormal point cloud is provided, comprising: clustering points other than the ground in a point cloud to be processed to obtain a clustered point cloud, wherein the point cloud to be processed is a point cloud of a current frame obtained by measuring by emitting laser by a laser radar during vehicle driving; extracting point cloud features from the clustered point cloud according to a target feature dimension, wherein the target feature dimension at least includes a feature dimension for representing the relative motion features of the points in the clustered point cloud between the current frame and a historical frame, and the historical frame is obtained by measuring by emitting laser by a laser radar before the current frame; when the point cloud features of the clustered point cloud in the target feature dimension meet preset conditions, determining the clustered point cloud as an abnormal point cloud, wherein the abnormal point cloud is a point cloud generated by the emitted laser passing through water droplets on the acquisition window of the laser radar and hitting the ground.

[0007] Optionally, points other than the ground in the point cloud to be processed are clustered to obtain a clustered point cloud, including: eliminating points belonging to the ground from the point cloud to be processed; clustering the remaining points in the point cloud to be processed to obtain a clustered point cloud, and saving the index of each point in the clustered point cloud to a mapping table, wherein the mapping table uses the index of the point as a key and the queue as a value, and each queue is used to save the ranging value of the point with the same index in the current frame and the ranging value in the historical frame.

[0008] Optionally, the point cloud features of the feature dimension used to represent the relative motion characteristics of points in the clustered point cloud between the current frame and the historical frame in the target feature dimension include: the proportion of stationary points in the clustered point cloud determined according to the mapping table and the queue.

[0009] Optionally, before determining the proportion of stationary points in the clustered point cloud according to the mapping table queue, the method also includes: for each point in the clustered point cloud, determining whether the point is a stationary point in the following manner: determining a first point in the clustered point cloud that has not been processed; using the index of the first point to determine a corresponding target queue from the mapping table, wherein the target queue stores the ranging value of the first point in the current frame and the ranging values ​​of multiple second points with the same index as the first point in historical frames, and the multiple second points are points in multiple adjacent historical frames; calculating the average of the absolute values ​​of the differences between the ranging value of the first point and the ranging values ​​of the multiple second points; when the average of the absolute values ​​of the differences is less than a preset distance threshold, determining that the first point is a stationary point; when the average of the absolute values ​​of the differences is not less than the preset distance threshold, determining that the first point is not a stationary point.

[0010] Optionally, the point cloud features of the target feature dimension also include: the average grayscale of the clustered point cloud; and the total number of points in the clustered point cloud.

[0011] Optionally, when the point cloud features of the clustered point cloud in the target feature dimension meet preset conditions, the clustered point cloud is determined to be an abnormal point cloud, including: when the average grayscale of the clustered point cloud is less than a preset grayscale threshold, the total number of points in the clustered point cloud is less than a preset point count threshold, and the proportion of static points in the clustered point cloud is greater than a preset proportion threshold, the clustered point cloud is determined to be an abnormal point cloud.

[0012] Optionally, before clustering points outside the ground in the point cloud to be processed to obtain the clustered point cloud, the method also includes: selecting the point cloud to be processed located in the target scanning area of ​​the laser radar from the original point cloud collected by the laser radar, wherein the target scanning area is a pre-set area of ​​interest.

[0013] Optionally, after determining that the clustered point cloud is an abnormal point cloud, the method further includes: deleting the abnormal point cloud from the original point cloud to facilitate subsequent processing of the remaining point cloud.

[0014] According to another aspect of an embodiment of the present application, a device for determining an abnormal point cloud is also provided, including: a clustering unit, used to cluster points other than the ground in a point cloud to be processed to obtain a clustered point cloud, wherein the point cloud to be processed is a point cloud of a current frame obtained by measuring by emitting laser by a laser radar during vehicle driving; an extraction unit, used to extract point cloud features from the clustered point cloud according to a target feature dimension, wherein the target feature dimension at least includes a feature dimension for representing the relative motion features of the points in the clustered point cloud between the current frame and a historical frame, and the historical frame is obtained by measuring by emitting laser by a laser radar before the current frame; a determination unit, used to determine that the clustered point cloud is an abnormal point cloud when the point cloud features of the clustered point cloud in the target feature dimension meet preset conditions, wherein the abnormal point cloud is a point cloud generated when the emitted laser passes through the water droplets on the acquisition window of the laser radar and hits the ground.

[0015] According to another aspect of an embodiment of the present application, a storage medium (specifically, a memory on an onboard electronic device) is also provided. The storage medium includes a stored program, and the above method is executed when the program is run.

[0016] According to another aspect of an embodiment of the present application, there is also provided an in-vehicle electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the above method through the computer program.

[0017] According to one aspect of the present application, a computer program product or a computer program is provided, the computer program product or the computer program comprising computer instructions, the computer instructions being stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the steps of any one of the embodiments of the above method.

[0018] In an embodiment of the present application, points other than the ground in a point cloud to be processed are clustered to obtain a clustered point cloud, where the point cloud to be processed is a point cloud of a current frame obtained by measuring by emitting laser light from a laser radar during vehicle driving; point cloud features are extracted from the clustered point cloud according to a target feature dimension, where the target feature dimension at least includes a feature dimension for representing relative motion features of points in the clustered point cloud between a current frame and a historical frame, where the historical frame is obtained by measuring by emitting laser light from a laser radar before the current frame; when the point cloud features of the clustered point cloud in the target feature dimension meet preset conditions, the clustered point cloud is determined to be an abnormal point cloud, which is a point cloud generated by the emitted laser passing through water droplets on the acquisition window of the laser radar and hitting the ground, thereby solving the technical problem in the related art that such abnormal point clouds cannot be detected. Since the ground is originally a drivable area, abnormal point clouds will be generated after refraction by water droplets. For vehicles, such abnormal point clouds represent obstacles, which means that the road surface that was originally drivable is recognized by the vehicle as an impassable area. The adoption of this solution can identify the noise point cloud generated by water droplets hitting the ground, thereby avoiding the impact of such abnormal point clouds with erroneous ranging on vehicle operation and improving the stability and safety of vehicle operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0020] Figure 1 is a schematic diagram of an optional generation principle of an abnormal point cloud according to an embodiment of the present application;

[0021] Figure 2 is a schematic diagram of an optional abnormal point cloud according to an embodiment of the present application;

[0022] Figure 3 is a flow chart of an optional method for determining an abnormal point cloud according to an embodiment of the present application;

[0023] Figure 4 is a flow chart of an optional method for determining an abnormal point cloud according to an embodiment of the present application;

[0024] Figure 5 is a schematic diagram of an optional abnormal point cloud determination device according to an embodiment of the present application; and,

[0025] Figure 6 It is a structural block diagram of a vehicle-mounted electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0026] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0027] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0028] First, some nouns or terms that appear in the description of the embodiments of the present application are subject to the following interpretation:

[0029] Laser radar (full name Light Detection and Ranging, referred to as LiDAR) is a radar system that emits a detection signal (laser beam) to the target, and then measures the arrival time, strength and other parameters of the reflected or scattered signal to determine the target's distance, direction, motion state and surface optical properties. Laser radar has excellent performance due to its short laser wavelength and high collimation: high angular resolution and distance resolution, strong anti-interference ability, ability to obtain a variety of target image information (depth, reflectivity, etc.), small size and light weight.

[0030] In the related technology, when raindrops fall on the radar window, a water droplet will be formed on the radar window. In this case, the laser emitted by the radar will be refracted through the water droplet, causing the energy light path to change and produce an abnormal point cloud.

[0031] refer to Figure 1 As shown, the inventors analyzed the laser passing through the water droplets on the radar window and realized that:

[0032] 1) When the ideal light is refracted upward by water droplets, the actual point cloud distance measurement will become longer, that is, points that are sunken compared to the ground will be produced. When the ideal light is refracted downward by water droplets, the actual point cloud distance measurement will become shorter, that is, points that are raised compared to the ground will be produced.

[0033] When water accumulates on the laser radar window, the originally flat point cloud will have convex and concave point clouds representing obstacles, such as Figure 2 The actual scene shown is a point cloud of the ground, where the white ones are abnormal point clouds. Due to the refraction caused by hanging water, the point clouds appear to protrude and sink into the ground, causing the detection algorithm to believe that there is an obstacle at that location, which in turn affects the normal driving of the autonomous vehicle.

[0034] It should be noted that although the amount of rain can be reduced by installing a rain cover, this method will incur a certain cost, and there will still be a small amount of rain hanging on the front window. Some small raindrops will remain on the front window, resulting in a fixed noise point in the point cloud. Even if there is ground in front of the vehicle, it will be misjudged as an obstacle in front, making it impossible for the vehicle to drive normally.

[0035] 2) Since water droplets are generally small, the point cloud area of ​​the ground depression or ground protrusion formed is smaller (that is, the number of points is smaller).

[0036] Based on the above cognition of the inventors, according to one aspect of the embodiments of the present application, a method embodiment of a method for determining an abnormal point cloud is provided. Figure 3 is a flowchart of an optional method for determining an abnormal point cloud according to an embodiment of the present application, such as Figure 3 As shown, the method may include the following steps:

[0037] Step S302, clustering the points other than the ground in the point cloud to be processed to obtain a clustered point cloud, where the point cloud to be processed is a point cloud of the current frame obtained by measuring the laser radar emitting laser during the vehicle's driving process.

[0038] Optionally, the original point cloud collected by the laser radar can be directly used as the point cloud to be processed; considering that the data volume of a complete point cloud frame of the laser radar is large, in order to improve the processing data, the point cloud to be processed located in the target scanning area of ​​the laser radar can be selected from the original point cloud collected by the laser radar. The target scanning area here is a pre-set area of ​​interest, which can be determined, for example, based on factors to be considered (such as the position of water droplets, the effective measurement distance of the light emitted by the laser radar after passing through the water droplets, and the perception range determined by the scope of the subsequent planning and control algorithm).

[0039] In the above embodiment, taking into account the possible interference of the ground point cloud, the points other than the ground in the point cloud to be processed are clustered to obtain the clustered point cloud. The points belonging to the ground can be first removed from the point cloud to be processed, and then the remaining points in the point cloud to be processed are clustered to obtain the clustered point cloud, and the index of each point in the clustered point cloud is saved to a mapping table, wherein the mapping table uses the index of the point as a key and the queue as a value, and each queue is used to save the ranging value of the point with the same index in the current frame and the ranging value in the historical frame.

[0040] The above mapping table adopts a key-value data structure, where the key is the index of each point in the clustered point cloud, and the value of the mapping table is a queue, which stores multiple frame points corresponding to the same index. For example, the non-ground point cloud is clustered to obtain m clustered point clouds. If the lth clustered point cloud (other clustered point clouds can be processed in the same way) contains a total of n scan points, then the key of the mapping table is the index of the n scan points in the original point cloud. For each scan point, there is a queue (i.e. the corresponding value, which can store the ranging values ​​of N points) to store the ranging values ​​of the previous (N-1) frames and the current frame of the scan point. The value of N can be set as needed, such as 5.

[0041] Step S304, extracting point cloud features from the clustered point cloud according to the target feature dimension, where the target feature dimension at least includes a feature dimension used to represent the relative motion characteristics of points in the clustered point cloud between the current frame and the historical frame, where the historical frame is obtained by measuring by emitting laser by the laser radar before the current frame.

[0042] In this embodiment, extracting point cloud features from the clustered point cloud according to the target feature dimension includes: determining the proportion of static points in the clustered point cloud according to the mapping table and the queue, that is, the number of static points divided by the total number of points.

[0043] Since the positions of the ground and the lidar are relatively stationary, that is, the distance measured by the light emitted from the radar at a fixed angle hitting the ground will remain basically unchanged (actually it hits the water droplets on the front window), and since the refraction of the water droplets is also relatively consistent, the position of the noise point is also stationary relative to the radar. This solution targets the water noise points that hit the ground and seriously affect the driving of the vehicle. It combines the features of the previous and next frames and judges the overall relative motion (i.e., whether it is a stationary point) by counting the relative motion features of each point in each clustered point cloud, which can better solve this problem.

[0044] It should be noted that if only the overall features are calculated, such as the center of gravity of the clustered point cloud, the previous and next frames need to be tracked, and the point cloud will be missing and jittering due to the change of the ground. In addition, it is easy to be affected by the over-segmentation and under-segmentation of the clustering algorithm. Therefore, the above features are used for judgment.

[0045] For each point in the clustered point cloud, determine whether the point is a stationary point as follows:

[0046] 3.1) Determine the first unprocessed point in the clustered point cloud.

[0047] 3.2) Using the index of the first point, determine the corresponding target queue from the mapping table. The target queue stores the ranging value of the first point in the current frame and the ranging values ​​of multiple second points with the same index as the first point in the historical frame. The multiple second points are points in multiple adjacent historical frames.

[0048] 3.3) Calculate the average of the absolute values ​​of the differences between the distance value of the first point and the distance values ​​of the (N-1) second points.

[0049] 3.4) When the average value of the absolute value of the difference is less than a preset distance threshold (eg, 0.2 meters), it is determined that the first point is a stationary point.

[0050] 3.5) When the average value of the absolute values ​​of the differences is not less than the preset distance threshold, it is determined that the first point is not a stationary point.

[0051] In this embodiment, the target feature dimensions may also include:

[0052] 1) The average grayscale of the clustered point cloud, that is, the sum of the grayscale values ​​of all points divided by the total number of points;

[0053] 2) The total number of points in the clustered point cloud.

[0054] Step S306, when the point cloud features of the clustered point cloud in the target feature dimension meet the preset conditions, the clustered point cloud is determined to be an abnormal point cloud, which is a point cloud generated when the emitted laser passes through the water droplets on the collection window of the laser radar and hits a relatively static object such as the ground.

[0055] After step S306, abnormal point clouds may be deleted from the original point clouds to facilitate subsequent processing of the remaining point clouds.

[0056] In the above embodiment, when the target feature dimension only includes the feature dimension used to represent the relative motion characteristics of the points in the clustered point cloud between the current frame and the historical frame, it can be judged whether the clustered point cloud is an abnormal point cloud based only on this dimension. For example, whether the clustered point cloud is an abnormal point cloud is determined based on the proportion of stationary points in the clustered point cloud. When the proportion of stationary points in the clustered point cloud is greater than a preset proportion threshold, the clustered point cloud is determined to be an abnormal point cloud. When the target feature dimension also includes other features, the other features are combined to jointly determine whether the clustered point cloud is an abnormal point cloud.

[0057] For example, when the target feature dimensions also include: the average grayscale of the clustered point cloud and the total number of points in the clustered point cloud, the clustered point cloud can be determined to be an abnormal point cloud when the average grayscale of the clustered point cloud is less than a preset grayscale threshold, the total number of points in the clustered point cloud is less than a preset point number threshold, and the proportion of static points in the clustered point cloud is greater than a preset proportion threshold.

[0058] By adopting the technical solution of the present application, points other than the ground in the point cloud to be processed are clustered to obtain a clustered point cloud, wherein the point cloud to be processed is a point cloud of the current frame obtained by measuring by emitting laser by a laser radar during vehicle driving; point cloud features are extracted from the clustered point cloud according to a target feature dimension, wherein the target feature dimension at least includes a feature dimension for representing the relative motion features of the points in the clustered point cloud between the current frame and a historical frame, and the historical frame is obtained by measuring by emitting laser by a laser radar before the current frame; when the point cloud features of the clustered point cloud in the target feature dimension meet preset conditions, the clustered point cloud is determined to be an abnormal point cloud, which is a point cloud generated by the emitted laser passing through the water droplets on the acquisition window of the laser radar and hitting the ground, thereby solving the technical problem that such abnormal point clouds cannot be detected in the related art. Since the ground is originally a drivable area, abnormal point clouds will be generated after refraction by water droplets. For vehicles, such abnormal point clouds represent obstacles, which means that the road surface that was originally drivable is recognized by the vehicle as an impassable area. The adoption of this solution can identify the noise point cloud generated by water droplets hitting the ground, thereby avoiding the impact of such abnormal point clouds with erroneous ranging on vehicle operation and improving the stability and safety of vehicle operation.

[0059] As an optional embodiment, the technical solution of the present application is further described in detail below in combination with a specific implementation method. The water hanging abnormal point cloud detection solution proposed in the present application mainly involves the following steps (refer to Figure 4 ):

[0060] Step 1: Set a certain region of interest ROI (full name: region of interest, in machine vision and image processing, the area to be processed is outlined in the processed image in the form of a box, circle, ellipse, irregular polygon, etc.) for the input point cloud (i.e., the original point cloud), and further process the point cloud within the ROI.

[0061] The ROI can generally take a point cloud within a certain distance range, which is determined by the range of abnormal point clouds that are actually prone to occur and the range required to affect the subsequent planning and control algorithm. For example:

[0062] 1) Water droplets are often located in specific areas of the laser radar acquisition window (such as the edges of the acquisition window, especially its lower edge), so these areas are the target scanning areas at this time;

[0063] 2) Water droplets will refract the light emitted by the LiDAR when passing through the water droplets, resulting in changes in the ranging optical path and energy attenuation (for example, assuming that the LiDAR has a detection range of 10 meters for ground points, after refraction, due to energy reduction, there may only be normal point clouds within 5 meters, and signals beyond 5 meters cannot be detected due to energy attenuation). At this time, the target scanning area is the area that can generate valid point clouds after the optical path changes and energy attenuation;

[0064] 3) The target scanning area can be the perception range determined by the scope of the subsequent planning and control algorithm. For example, if there is a ground noise point at a closer distance, such as 1m in front, the vehicle will be unable to avoid it and will be completely unable to drive. However, there is a ground noise point 10m away, which may have little impact on the subsequent planning and control algorithm. In this case, the area within 10m is the target scanning area.

[0065] Of course, the target scanning area can also be obtained by finding the intersection of two or three of the above three factors (the specific area of ​​the acquisition window, the area that can generate a valid point cloud after the optical path changes and energy attenuation, and the perception range determined by the scope of the planning and control algorithm).

[0066] Step 2: Remove ground points through ground point detection algorithm.

[0067] For example, direct calculation method: directly calculate the pitch angle of the point cloud of two adjacent laser lines, and the points with pitch angle changes within a certain range are considered to be ground points;

[0068] Plane fitting method: After the ground plane equation is initially fitted using multiple points, other points are checked and those that meet a certain range are considered to be ground points;

[0069] Conditional screening method: the points above and below the ground points are filtered out, and the remaining points are ground points. For example, a height range is defined based on the installation height of the sensor from the ground, and the points within this range are considered to be ground points.

[0070] Step 3: Cluster the remaining point cloud after removing the ground points.

[0071] The clustering algorithm here can use the Euclidean clustering algorithm. The basic idea of ​​this algorithm is to use the Euclidean distance to measure the similarity between points and classify points with high similarity into the same category. First, select an initial point as the seed point of the cluster, then calculate the Euclidean distance between other points and the seed point, and add points with a distance less than the set threshold to the category. Then, select an unmarked point from the cluster as a new seed point, and repeat the above process until all points are clustered. In the specific implementation, all points can be traversed and the Euclidean distance between them and other points can be calculated, and points with a distance less than the threshold can be added to the same category.

[0072] Step 4, save the index of each cluster point cloud to a mapping table of a queue with a length of n (for example, n=5 (range n>=2)) for subsequent algorithm search.

[0073] Step 5: For each clustered point cloud, extract the following point cloud features (such as average grayscale, statistical point count), and then count the average error (i.e., the average of the absolute value of the difference) between the ranging value corresponding to each scanning angle and the previous n-1 frames according to the mapping table in step 3.

[0074] Step 6, determine whether the average grayscale is less than the grayscale threshold (such as 4, usually 3-10, usually after the water droplets are refracted and hit the ground, the grayscale value of the noise is lower), determine whether the number of points is less than the point threshold (such as 150, by taking the value of 0-500, the area affected by each water droplet is small), determine whether the average error is less than the distance threshold (such as 0.2m, usually between 0-1), if less than, the point is a stationary point, determine whether the proportion of the number of stationary points to the total number of points is greater than the proportion threshold (such as 0.7, the value range is 0-1).

[0075] Step 6: The clustered point clouds that meet the above three conditions are noise points generated by water hanging, and these point clouds are deleted from the original point cloud.

[0076] For the abnormal point cloud generated by water on the window, the obstacles to be processed are clustered by removing ground points and clustering algorithms, and then the features of each obstacle point cloud are extracted and the abnormal point cloud is identified by combining the features of the previous and next frames of each point, so as to avoid the abnormal point cloud generated by water on the lidar being treated as an obstacle, thereby affecting the normal driving of the autonomous driving vehicle. This solution can effectively detect the abnormal point cloud caused by static water on the lidar window on rainy days, thereby eliminating the problems caused by this abnormal point cloud, so that the lidar can be used normally under certain conditions on rainy days.

[0077] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the described order of actions, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

[0078] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0079] According to another aspect of an embodiment of the present application, a device for determining an abnormal point cloud for implementing the above-mentioned method for determining an abnormal point cloud is also provided. Figure 5 is a schematic diagram of an optional abnormal point cloud determination device according to an embodiment of the present application, such as Figure 5 As shown, the device may include:

[0080] A clustering unit 51 is used to cluster points other than the ground in the point cloud to be processed to obtain a clustered point cloud, wherein the point cloud to be processed is a point cloud of a current frame obtained by measuring the laser radar emitting laser during the vehicle's driving process;

[0081] An extraction unit 53 is used to extract point cloud features from the clustered point cloud according to a target feature dimension, wherein the target feature dimension at least includes a feature dimension for representing a relative motion feature of a point in the clustered point cloud between a current frame and a historical frame, and the historical frame is obtained by measuring by emitting laser light by a laser radar before the current frame;

[0082] The determination unit 55 is used to determine that the clustered point cloud is an abnormal point cloud when the point cloud features of the clustered point cloud in the target feature dimension meet the preset conditions, wherein the abnormal point cloud is a point cloud generated when the emitted laser passes through the water droplets on the collection window of the laser radar and hits the ground.

[0083] Through the above module, points other than the ground in the point cloud to be processed are clustered to obtain a clustered point cloud, where the point cloud to be processed is the point cloud of the current frame obtained by measuring by emitting laser by the laser radar during the driving of the vehicle; point cloud features are extracted from the clustered point cloud according to the target feature dimension, where the target feature dimension at least includes a feature dimension for representing the relative motion features of the points in the clustered point cloud between the current frame and the historical frame, where the historical frame is obtained by measuring by emitting laser by the laser radar before the current frame; when the point cloud features of the clustered point cloud in the target feature dimension meet the preset conditions, the clustered point cloud is determined to be an abnormal point cloud, where the abnormal point cloud is the point cloud generated by the emitted laser passing through the water droplets on the acquisition window of the laser radar and hitting the ground, thereby solving the technical problem that the abnormal point cloud cannot be detected in the related technology. Since the ground is originally a drivable area, abnormal point clouds will be generated after refraction by water droplets. For vehicles, such abnormal point clouds represent obstacles, which means that the road surface that was originally drivable is recognized by the vehicle as an impassable area. The adoption of this solution can identify the noise point clouds generated by water droplets hitting the ground, thereby avoiding the impact of such abnormal point clouds on vehicle ranging and improving the stability and safety of vehicle operation.

[0084] Optionally, the clustering unit is also used to: remove points belonging to the ground from the point cloud to be processed; cluster the remaining points in the point cloud to be processed to obtain a clustered point cloud, and save the index of each point in the clustered point cloud to a mapping table, wherein the mapping table uses the index of the point as a key and the queue as a value, and each queue is used to save the ranging value of the point with the same index in the current frame and the ranging value in the historical frame.

[0085] Optionally, the point cloud features of the feature dimension used to represent the relative motion characteristics of points in the clustered point cloud between the current frame and the historical frame in the target feature dimension include: the proportion of stationary points in the clustered point cloud determined according to the mapping table and the queue.

[0086] Optionally, the point cloud features of the target feature dimension also include: the average grayscale of the clustered point cloud; and the total number of points in the clustered point cloud.

[0087] Optionally, the extraction unit is also used to: before determining the proportion of stationary points in the clustered point cloud according to the mapping table, for each point in the clustered point cloud, determine whether the point is a stationary point in the following manner: determine a first point in the clustered point cloud that has not been processed; use the index of the first point to determine a corresponding target queue from the mapping table, wherein the target queue stores the ranging value of the first point in the current frame and the ranging values ​​of multiple second points with the same index as the first point in historical frames, and the multiple second points are points in multiple adjacent historical frames; calculate the average of the absolute values ​​of the differences between the ranging value of the first point and the ranging values ​​of the multiple second points; when the average value of the absolute values ​​of the differences is less than a preset distance threshold, determine that the first point is a stationary point; when the average value of the absolute values ​​of the differences is not less than the preset distance threshold, determine that the first point is not a stationary point.

[0088] Optionally, the determination unit is also used to determine that the clustered point cloud is an abnormal point cloud when the average grayscale of the clustered point cloud is less than a preset grayscale threshold, the total number of points in the clustered point cloud is less than a preset point number threshold, and the proportion of static points in the clustered point cloud is greater than a preset proportion threshold.

[0089] Optionally, the clustering unit is also used to: before clustering points outside the ground in the point cloud to be processed and obtaining the clustered point cloud, select the point cloud to be processed located in the target scanning area of ​​the laser radar from the original point cloud collected by the laser radar, and the target scanning area is a pre-set area of ​​interest.

[0090] It should be noted that the examples and application scenarios implemented by the above modules and corresponding steps are the same, but are not limited to the contents disclosed in the above embodiments. It should be noted that the above modules, as part of the device, can run in the hardware environment of the vehicle and can be implemented by software or hardware, wherein the hardware environment includes a network environment.

[0091] According to another aspect of an embodiment of the present application, a vehicle-mounted electronic device for implementing the above-mentioned method for determining an abnormal point cloud is also provided.

[0092] Figure 6 is a structural block diagram of a vehicle-mounted electronic device according to an embodiment of the present application, such as Figure 6 As shown, the vehicle-mounted electronic device may include: one or more ( Figure 6 (only one is shown) processor 601, memory 603, and transmission device 605, as shown in Figure 6 As shown, input and output devices 607 may also be included.

[0093] Among them, the memory 603 can be used to store software programs and modules, such as the program instructions / modules corresponding to the method and device for determining the abnormal point cloud in the embodiment of the present application. The processor 601 executes various functional applications and data processing by running the software programs and modules stored in the memory 603, that is, the above-mentioned method for determining the abnormal point cloud is realized. The memory 603 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 603 may further include a memory remotely arranged relative to the processor 601, and these remote memories can be connected to the terminal via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0094] The above-mentioned transmission device 605 is used to receive or send data via a network, and can also be used for data transmission between a processor and a memory. Specific examples of the above-mentioned network may include a wired network and a wireless network. In one example, the transmission device 605 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices and routers via a network cable so as to communicate with the Internet or a local area network. In one example, the transmission device 605 is a radio frequency (Radio Frequency, RF) module, which is used to communicate with the Internet wirelessly.

[0095] Specifically, the memory 603 is used to store application programs.

[0096] The processor 601 may call the application stored in the memory 603 through the transmission device 605 to perform the following steps:

[0097] Points other than the ground in the point cloud to be processed are clustered to obtain a clustered point cloud, wherein the point cloud to be processed is a point cloud of the current frame obtained by measuring by emitting laser from a laser radar during vehicle driving; point cloud features are extracted from the clustered point cloud according to the target feature dimension, wherein the target feature dimension at least includes a feature dimension for representing the relative motion features of the points in the clustered point cloud between the current frame and a historical frame, and the historical frame is obtained by measuring by emitting laser from a laser radar before the current frame; when the point cloud features of the clustered point cloud in the target feature dimension meet preset conditions, the clustered point cloud is determined to be an abnormal point cloud, wherein the abnormal point cloud is a point cloud generated by the emitted laser passing through the water droplets on the acquisition window of the laser radar and hitting a relatively static obstacle such as the ground.

[0098] Optionally, the specific examples in this embodiment may refer to the examples described in the above embodiments, and this embodiment will not be described in detail here.

[0099] It can be understood by those skilled in the art that Figure 6 The structure shown is for illustration only. Figure 6 It does not limit the structure of the above electronic device. For example, it may also include Figure 6 More or fewer components (such as network interfaces, display devices, etc.) shown in, or having Figure 6 Different configurations shown.

[0100] A person of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the hardware related to the terminal device through a program, and the program can be stored in a computer-readable storage medium, and the storage medium may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0101] The embodiment of the present application further provides a storage medium. Optionally, in this embodiment, the storage medium can be used to execute the program code of the method for determining an abnormal point cloud.

[0102] Optionally, in this embodiment, the storage medium may be located on at least one network device among a plurality of network devices in the network shown in the above embodiment.

[0103] Optionally, in this embodiment, the storage medium is configured to store program codes for executing the following steps:

[0104] Points other than the ground in the point cloud to be processed are clustered to obtain a clustered point cloud, wherein the point cloud to be processed is a point cloud of the current frame obtained by measuring by emitting laser from a laser radar during vehicle driving; point cloud features are extracted from the clustered point cloud according to the target feature dimension, wherein the target feature dimension at least includes a feature dimension for representing the relative motion features of the points in the clustered point cloud between the current frame and a historical frame, and the historical frame is obtained by measuring by emitting laser from a laser radar before the current frame; when the point cloud features of the clustered point cloud in the target feature dimension meet preset conditions, the clustered point cloud is determined to be an abnormal point cloud, wherein the abnormal point cloud is a point cloud generated by the emitted laser passing through the water droplets on the acquisition window of the laser radar and hitting a relatively static obstacle such as the ground.

[0105] Optionally, the specific examples in this embodiment may refer to the examples described in the above embodiments, and this embodiment will not be described in detail here.

[0106] Optionally, in this embodiment, the above-mentioned storage medium may include but is not limited to: a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and other media that can store program codes.

[0107] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0108] If the integrated units in the above embodiments are implemented in the form of software functional units and sold or used as independent products, they can be stored in the above computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling one or more computer devices (which may be personal computers, servers, or network devices, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application.

[0109] In the above embodiments of the present application, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0110] In the several embodiments provided in the present application, it should be understood that the disclosed client can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0111] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0112] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0113] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for determining an abnormal point cloud, characterized in that: include: Clustering points other than the ground in the point cloud to be processed to obtain a clustered point cloud, wherein the point cloud to be processed is a point cloud of the current frame obtained by measuring the laser radar emitting laser during the vehicle's driving process; Extracting point cloud features from the clustered point cloud according to target feature dimensions, wherein the target feature dimensions at least include feature dimensions for representing relative motion features of points in the clustered point cloud between a current frame and a historical frame, wherein the historical frame is obtained by the laser radar emitting laser for measurement before the current frame; When the point cloud features of the clustered point cloud in the target feature dimension meet preset conditions, the clustered point cloud is determined to be an abnormal point cloud, wherein the abnormal point cloud is a point cloud generated when the emitted laser passes through the water droplets on the acquisition window of the laser radar and hits the ground.

2. The method according to claim 1, characterized in that The step of clustering points other than the ground in the point cloud to be processed to obtain a clustered point cloud includes: Eliminate points belonging to the ground from the point cloud to be processed; The remaining points in the point cloud to be processed are clustered to obtain the clustered point cloud, and the index of each point in the clustered point cloud is saved in a mapping table, wherein the mapping table uses the index of the point as a key and the queue as a value, and each queue is used to save the ranging value of the point with the same index in the current frame and the ranging value in the historical frame.

3. The method according to claim 2, characterized in that The point cloud features of the feature dimension used to represent the relative motion features of the points in the clustered point cloud between the current frame and the historical frame in the target feature dimension include: The proportion of stationary points in the clustered point cloud is determined according to the mapping table and the queue.

4. The method according to claim 3, characterized in that Before determining the proportion of stationary points in the clustered point cloud according to the mapping table and the queue, the method further includes: for each point in the clustered point cloud, determining whether the point is a stationary point in the following manner: Determining a first unprocessed point in the clustered point cloud; Determine a corresponding target queue from the mapping table using the index of the first point, wherein the target queue stores a ranging value of the first point in the current frame and ranging values ​​of a plurality of second points having the same index as the first point in the historical frame, wherein the plurality of second points are points in a plurality of adjacent historical frames; Calculate an average of absolute values ​​of differences between the distance measurement value of the first point and the distance measurement values ​​of the plurality of second points; When the average value of the absolute values ​​of the differences is less than a preset distance threshold, determining that the first point is a stationary point; When the average value of the absolute values ​​of the differences is not less than the preset distance threshold, it is determined that the first point is not a stationary point.

5. The method according to claim 3 or 4, characterized in that: The point cloud features of the target feature dimension also include: The average grayscale of the clustered point cloud; The total number of points in the clustered point cloud.

6. The method according to claim 5, characterized in that When the point cloud feature of the clustered point cloud in the target feature dimension meets a preset condition, determining that the clustered point cloud is an abnormal point cloud includes: When the average grayscale of the clustered point cloud is less than a preset grayscale threshold, the total number of points of the clustered point cloud is less than a preset point number threshold, and the proportion of static points in the clustered point cloud is greater than a preset proportion threshold, the clustered point cloud is determined to be an abnormal point cloud.

7. The method according to any one of claims 1 to 4, characterized in that Before clustering points other than the ground in the point cloud to be processed to obtain the clustered point cloud, the method further includes: From the original point cloud collected by the laser radar, the point cloud to be processed located in the target scanning area of ​​the laser radar is selected, wherein the target scanning area is a pre-set area of ​​interest.

8. A device for determining an abnormal point cloud, characterized in that: include: A clustering unit, used for clustering points other than the ground in the point cloud to be processed to obtain a clustered point cloud, wherein the point cloud to be processed is a point cloud of a current frame obtained by measuring the laser radar emitting laser during the vehicle's driving process; an extraction unit, configured to extract point cloud features from the clustered point cloud according to a target feature dimension, wherein the target feature dimension at least includes a feature dimension for representing a relative motion feature of a point in the clustered point cloud between a current frame and a historical frame, wherein the historical frame is obtained by the laser radar emitting laser for measurement before the current frame; A determination unit is used to determine that the clustered point cloud is an abnormal point cloud when the point cloud features of the clustered point cloud in the target feature dimension meet preset conditions, wherein the abnormal point cloud is a point cloud generated when the emitted laser passes through the water droplets on the acquisition window of the laser radar and hits the ground.

9. A storage medium, characterized in that: The storage medium includes a stored program, wherein the program executes the method described in any one of claims 1 to 7 when executed.

10. An in-vehicle electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: The processor executes the method described in any one of claims 1 to 7 through the computer program.

Images