Tow position detection method and device, electronic equipment and storage medium

Abstract

Embodiments of the present disclosure disclose a dragline pose detection method and device, electronic equipment and a storage medium. The method comprises: obtaining a pose of a first dragline, determining a target detection box corresponding to a second dragline according to the pose of the first dragline; projecting each point cloud point in the target detection box onto a horizontal plane to obtain each projection point; determining an edge point from each projection point according to a preset reference point, and determining a fitted edge of the second dragline according to the edge point; determining the pose of the second dragline according to the fitted edge and the size of the second dragline, taking the second dragline as a new first dragline, taking a dragline adjacent to the second dragline and located after the second dragline as a new second dragline, and repeating the step of determining the pose of the second dragline until the second dragline is the last dragline. In the case that the rear dragline is blocked and far away from the laser radar, the dragline pose can still be accurately determined.

Description

Technical Field

[0001] This disclosure relates to the field of target perception and tracking technology, and in particular to a method, apparatus, electronic device and storage medium for detecting the pose of a tow bucket. Background Technology

[0002] When autonomous driving is applied in unmanned logistics scenarios, it is usually necessary to attach trailers for carrying goods. For example, an autonomous vehicle is used as the tractor unit, and one or more trailers are connected together to form a "trailer convoy".

[0003] Because the trailer convoy is mounted behind the autonomous vehicle, the LiDAR installed on the vehicle can only accurately detect the pose of the first trailer. The trailers behind the first trailer are usually obscured by the first trailer, so their poses are generally inferred from the poses of the trailers in front. However, if the trailer convoy is long, the cumulative error in the poses of the trailers further back increases, deviating significantly from the true pose. Furthermore, the LiDAR point clouds on the later trailers become sparser, leading to a significant decrease in the signal-to-noise ratio and affecting the accuracy of pose detection. Summary of the Invention

[0004] To solve the above-mentioned technical problems, or at least partially solve them, this disclosure provides a method, apparatus, electronic device, and storage medium for detecting the position of a trailer, which can still accurately determine the position of the trailer even when the trailer is obstructed and the distance between it and the lidar is far.

[0005] In a first aspect, embodiments of this disclosure provide a method for detecting the pose of a tow bucket, the method comprising:

[0006] The pose of the first trailer is obtained, and a target detection box corresponding to the second trailer is determined based on the pose of the first trailer; wherein the second trailer is located after the first trailer and adjacent to the first trailer.

[0007] Project each point cloud within the target detection box onto a horizontal plane to obtain each projection point;

[0008] Based on the preset reference points, edge points are determined from each projection point, and the fitting edge of the second bucket is determined based on the edge points;

[0009] Based on the fitted edge and the size of the second trailer, the pose of the second trailer is determined, the second trailer is designated as the new first trailer, and the trailer adjacent to and following the second trailer is designated as the new second trailer. The step of determining the pose of the second trailer is repeated until the second trailer is the last trailer.

[0010] Secondly, embodiments of this disclosure also provide a bucket position detection device, the device comprising:

[0011] The target detection box determination module is used to obtain the pose of the first trailer and determine the target detection box corresponding to the second trailer based on the pose of the first trailer; wherein the second trailer is located after the first trailer and adjacent to the first trailer.

[0012] The projection point determination module is used to project each point cloud within the target detection box onto a horizontal plane to obtain each projection point.

[0013] The fitting edge determination module is used to determine edge points from the projection points based on preset reference points, and to determine the fitting edge of the second trailer based on the edge points.

[0014] The pose determination module is used to determine the pose of the second trailer based on the fitted edge and the size of the second trailer, to designate the second trailer as the new first trailer, and to designate the trailer adjacent to and following the second trailer as the new second trailer, and to repeat the step of determining the pose of the second trailer until the second trailer is the last trailer.

[0015] Thirdly, this disclosure also provides an electronic device, which includes: one or more processors; a storage device for storing one or more programs; and when the one or more programs are executed by the one or more processors, the one or more processors implement the tow bucket pose detection method as described above.

[0016] Fourthly, embodiments of this disclosure also provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the bucket pose detection method as described above.

[0017] This disclosure provides a method for detecting the pose of a trailer bucket. By acquiring the pose of a first trailer bucket, a target detection box corresponding to a second trailer bucket is determined based on the pose of the first trailer bucket. This more effectively limits the approximate range of the point cloud corresponding to the second trailer bucket, thereby improving detection accuracy. Furthermore, each point cloud point within the target detection box is projected onto a horizontal plane to obtain projection points. Edge points are determined from each projection point based on a preset reference point, and the fitted edge of the second trailer bucket is determined based on the edge points. This method compresses the three-dimensional point cloud to a two-dimensional plane and, based on the preset reference point, further... By employing a noise filtering method, the point cloud density is increased, noise interference is reduced, and the signal-to-noise ratio is improved. Based on the fitted edge and the size of the second bucket, the pose of the second bucket is determined. The second bucket is then used as the new first bucket, and the bucket adjacent to and following the second bucket is used as the new second bucket. This process of determining the pose of the second bucket is repeated until the second bucket becomes the last bucket. This method achieves the effect of increasing point cloud density, improving the signal-to-noise ratio, and thus improving the detection accuracy of the bucket pose even when the bucket behind is obstructed and is far from the lidar. Attached Figure Description

[0018] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and the originals and elements are not necessarily drawn to scale.

[0019] Figure 1 A schematic diagram illustrating the accumulation of errors in bucket position estimation;

[0020] Figure 2 This is a flowchart of a bucket pose detection method according to an embodiment of the present disclosure;

[0021] Figure 3 This is a schematic diagram of a target detection frame and the coverage of the lidar over the second trailer in an embodiment of this disclosure;

[0022] Figure 4 This is a flowchart of another bucket pose detection method in this embodiment of the present disclosure;

[0023] Figure 5 This is a schematic diagram of a preset reference point in an embodiment of this disclosure;

[0024] Figure 6 This is a schematic diagram of a target sector in an embodiment of the present disclosure;

[0025] Figure 7 This is a schematic diagram of an edge point in an embodiment of this disclosure;

[0026] Figure 8This is a flowchart of another bucket pose detection method in this embodiment of the present disclosure;

[0027] Figure 9 This is a schematic diagram of the structure of a bucket posture detection device according to an embodiment of the present disclosure;

[0028] Figure 10 This is a schematic diagram of the structure of an electronic device according to an embodiment of this disclosure. Detailed Implementation

[0029] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0030] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0031] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0032] Commonly used trailer pose detection techniques employ LiDAR sensors to detect the pose of the first trailer section. The poses of subsequent trailer sections are typically obscured by the first section; therefore, the poses of the later trailer sections are generally inferred from the poses of the preceding trailer sections or estimated from their trajectories. However, if there are many trailer sections, the cumulative error of these inference methods increases significantly as the number of trailer sections increases, eventually leading to a large discrepancy between the inferred and actual poses, which can pose safety risks. Figure 1 A schematic diagram illustrating the accumulation of bucket attitude estimation errors, as shown below. Figure 1As shown, the rounded rectangle represents the autonomous vehicle, i.e., the trailer, with four trailers attached to its rear. Solid rectangles represent the actual poses of the trailers, while dashed rectangles represent the inferred poses. The autonomous vehicle's sensors can detect the pose of the first trailer well and then infer the poses of the trailers behind it. However, due to accumulated errors, the pose of the rear trailer deviates significantly. Another method, based on a fixed detection ROI (Region of Interest), detects trailer poses at curves. Since the rear trailers have a larger range of motion, this type of method requires increasingly larger ROIs for each rear trailer, increasing the detection range and the false detection rate. Furthermore, the closer to the rear trailer, the sparser the 3D point cloud of the LiDAR becomes when projecting onto the trailer. With each rear trailer corresponding to a larger ROI, noise increases, leading to a significant decrease in the signal-to-noise ratio and affecting detection performance. Therefore, a noise filtering step is needed. Typically, the detection target during bucket inspection is the bucket edge. Generally, the 3D point cloud from the LiDAR forms a straight line at the bucket edge. However, the LiDAR may also hit cargo on the bucket. This cargo could be a cargo box (with a regular straight point cloud, similar to the bucket edge) or miscellaneous items (irregular shape and variable density). Therefore, general noise filtering algorithms are not suitable. For example, density-based filtering algorithms typically filter out low-density outliers, but not all noise points conform to the low-density characteristic, leading to poor filtering results. Downsampling-based filtering algorithms generally smooth and sparse the point cloud, but they do not filter out point clouds outside the bucket edge, resulting in poor point cloud filtering performance.

[0033] To address the aforementioned issues, this disclosure provides a method for detecting the position and orientation of a trailer, which can still accurately determine the position and orientation of the trailer even when the trailer is obstructed from behind and is far from the radar.

[0034] Figure 2 This is a flowchart illustrating a bucket pose detection method according to an embodiment of this disclosure. The bucket pose detection method can be executed by a bucket pose detection device, which can be implemented in software and / or hardware, and can be configured in an electronic device. Figure 2 As shown, the method may specifically include the following steps:

[0035] S110. Obtain the pose of the first trailer and determine the target detection box corresponding to the second trailer based on the pose of the first trailer.

[0036] The second trailer is located after and adjacent to the first trailer. It can be understood that the first trailer is the Nth trailer section, and the second trailer is the (N+1)th trailer section, where 1 ≤ N < n, N is a positive integer, and n is the total number of trailers. Pose includes position and orientation. The target detection box corresponding to the second trailer is the predicted detection box surrounding the second trailer, which can be understood as the ROI corresponding to the second trailer.

[0037] Specifically, the pose of the first trailer can be obtained through detection or prediction. Then, given the pose of the first trailer, the target detection box corresponding to the second trailer can be predicted based on the direction of the autonomous vehicle and the size of the second trailer, so that the pose of the second trailer can be determined in the target detection box.

[0038] Building upon the above example, when the autonomous vehicle enters a curve, it can be pre-determined whether subsequent detection is needed to reduce time consumption and avoid unnecessary false detections. Specifically, before determining the target detection box corresponding to the second trailer based on the pose of the first trailer, the following steps can be taken:

[0039] Based on the heading angle of the unmanned vehicle, the driving distance of the unmanned vehicle, and the pose of the first trailer, determine the coverage of the second trailer by the lidar on the unmanned vehicle; if the coverage covers the second trailer, proceed to determine the pose of the second trailer; if the coverage does not cover the second trailer, stop proceeding to determine the pose of the second trailer.

[0040] The heading angle indicates the direction of movement of the autonomous vehicle. The travel distance indicates the distance the autonomous vehicle has traveled. The lidar can be installed at a preset location on the autonomous vehicle, such as behind or in front. Coverage describes whether the lidar's detection range covers the side of the second trailer.

[0041] Specifically, based on the autonomous vehicle's heading angle, travel distance, and the pose of the first trailer, the detection range of the LiDAR can be determined. This allows us to determine whether the second trailer will be excluded from the detection range due to obstruction by the first trailer. Specifically, we can determine if the side of the second trailer is obstructed by the first trailer, thus determining the coverage of the second trailer by the LiDAR on the autonomous vehicle. If the second trailer is covered, it indicates that its side is not obstructed by the first trailer, and subsequent pose detection can be performed using the LiDAR. Therefore, the subsequent step of determining the pose of the second trailer can proceed. If the second trailer is not covered, it indicates that its side is obstructed by the first trailer, and the LiDAR cannot detect its side. Therefore, the step of determining the pose of the second trailer can be stopped to save time and reduce false detections.

[0042] For example, such as Figure 3 The diagram shows the target detection bounding box and the coverage of the lidar on the second trailer. The dashed box represents the pose of the second trailer, the solid box surrounding the dashed box represents the target detection bounding box, the two circles on the autonomous vehicle represent the lidar, and the two dashed lines emitted by the lidar represent its detection range. Figure 3 It can be seen that in this case, the coverage extends to the second trailer.

[0043] S120. Project each point cloud within the target detection box onto the horizontal plane to obtain each projection point.

[0044] The point cloud points within the target detection bounding box can be point cloud points located within the target detection bounding box detected by LiDAR. The projection points can be points obtained by projecting point cloud points at various heights onto a horizontal plane.

[0045] Specifically, the system uses lidar to detect the area behind the autonomous vehicle, identifies the point cloud points that fall within the target detection frame, projects these point cloud points onto a horizontal plane, and uses the projected points as projection points.

[0046] S130. Based on the preset reference points, determine the edge points from each projection point, and based on the edge points, determine the fitting edge of the second trailer.

[0047] The preset reference point can be a reference point pre-set outside the target detection box to determine the edge points in the projection points. The fitted edge can be an edge obtained by fitting each edge point, used to represent an edge of the predicted second bucket.

[0048] Specifically, based on a preset reference point and the distance between each projection point and the preset reference point, projection points closer to the preset reference point are determined as edge points. Then, by fitting each edge point, a straight line segment is obtained, and this straight line segment is used as the fitted edge of the second trailer.

[0049] S140. Based on the fitted edge and the size of the second trailer, determine the pose of the second trailer, use the second trailer as the new first trailer, and use the trailer adjacent to and following the second trailer as the new second trailer. Repeat the step of determining the pose of the second trailer until the second trailer is the last trailer.

[0050] The dimensions of the second trailer may include its length, width, and height.

[0051] Specifically, the pose of the second trailer can be expanded and recovered based on the fitted edge and the size of the second trailer. Then, to predict the poses of subsequent trailers, the second trailer can be used as the new first trailer, and the trailer adjacent to and following the second trailer can be used as the new second trailer. This process of determining the pose of the second trailer is repeated until the second trailer becomes the last trailer, meaning the poses of all trailers have been detected.

[0052] Based on the above example, if the first trailer is adjacent to and behind the autonomous vehicle, then the pose of the first trailer can be obtained in any of the following ways:

[0053] Method 1: Obtain the pose of the unmanned vehicle; determine the target detection box corresponding to the first trailer based on the pose of the unmanned vehicle; project the point cloud points in the target detection box onto the horizontal plane to obtain the projection points; determine the edge points from the projection points based on the preset reference points, and determine the fitted edge of the first trailer based on the edge points; determine the pose of the first trailer based on the fitted edge and the size of the first trailer.

[0054] Method 1 can be understood as follows: the unmanned vehicle is likened to the first trailer in the present invention, and the first trailer is likened to the second trailer in the present invention. Accordingly, the position and orientation of the first trailer can be determined in a manner similar to that in the present invention.

[0055] Method 2: Determine the detection edge of the first trailer based on the lidar on the unmanned vehicle, and determine the pose of the first trailer based on the detection edge and the size of the first trailer.

[0056] The detection edge can be at least one edge that the lidar can directly cover. The dimensions of the first trailer can include its length, width, and height.

[0057] Specifically, the first trailer is detected using the lidar on the autonomous vehicle, and the detected edges are used as the detected edges of the first trailer. Then, based on the known dimensions of the first trailer and the detected edges, the pose of the first trailer is reconstructed.

[0058] The first trailer in both Method 1 and Method 2 can be understood as the first trailer section suspended at the rear of the unmanned vehicle.

[0059] The bucket pose detection method provided in this embodiment obtains the pose of the first bucket, determines the target detection box corresponding to the second bucket based on the pose of the first bucket, so as to more effectively limit the approximate range of the point cloud corresponding to the second bucket, thereby improving the detection accuracy. Furthermore, the point cloud points within the target detection box are projected onto a horizontal plane to obtain projection points. Edge points are determined from each projection point based on a preset reference point, and the fitting edge of the second bucket is determined based on the edge points. This method compresses the three-dimensional point cloud to a two-dimensional plane and performs noise reduction based on the preset reference point. By using point filtering, the point cloud density is increased, noise interference is reduced, and the signal-to-noise ratio is improved. Based on the fitted edge and the size of the second bucket, the pose of the second bucket is determined. The second bucket is then used as the new first bucket, and the bucket adjacent to and following the second bucket is used as the new second bucket. This process of determining the pose of the second bucket is repeated until the second bucket is the last bucket. This method improves the point cloud density, signal-to-noise ratio, and thus the accuracy of bucket pose detection even when the bucket behind is obscured and the distance from the lidar is far.

[0060] Figure 4 This is a flowchart of another bucket pose detection method according to an embodiment of this disclosure. Based on the above embodiments, the methods for determining the target detection box corresponding to the second bucket, determining edge points, and determining fitted edges can be found in the detailed description of this technical solution. Furthermore, the method for determining preset reference points can also be found in the detailed description of this technical solution. The explanations of terms that are the same as or corresponding to those in the above embodiments will not be repeated here. Figure 4 As shown, the method may specifically include the following steps:

[0061] S210. Obtain the position of the first trailer bucket, and determine the target angle between the second trailer bucket and the first trailer bucket based on the reference angle between the first trailer bucket and the reference trailer bucket.

[0062] In this context, the reference tractor is the tractor or unmanned vehicle adjacent to and preceding the first tractor. It can be understood that the first tractor is the Nth tractor section, the second tractor is the (N+1)th tractor section, and the reference tractor is the (N-1)th tractor section, where 1 ≤ N < n, N is a positive integer, n is the total number of tractors, and the 0th tractor section can be understood as the unmanned vehicle. The reference angle can be the angle between the heading angle of the first tractor and the heading angle of the reference tractor. The target angle can be the angle between the heading angle of the second tractor and the heading angle of the first tractor.

[0063] Specifically, the position and orientation of the first trailer can be obtained, and the reference angle between the first trailer and the reference trailer can be obtained. Then, based on the reference angle and the preset trailer angle calculation method, the target angle between the second trailer and the first trailer can be estimated and determined.

[0064] It should be noted that the preset method for calculating the trailer angle can be to establish a motion model based on the unmanned vehicle and the convoy of trailers, analyze its turning process, and determine the angle change relationship between each trailer section during the turning process.

[0065] S220. Based on the pose of the first trailer, the target angle, and the size of the second trailer, determine the target detection box corresponding to the second trailer.

[0066] The size of the target detection box is larger than the size of the second trailer.

[0067] Specifically, based on the pose of the first trailer and the target angle, the approximate location of the center of the second trailer can be determined. Then, based on the determined approximate location of the center of the second trailer and the size of the second trailer, a detection box larger than the size of the second trailer can be expanded, that is, the target detection box corresponding to the second trailer.

[0068] It is understandable that the approximate location of the center of the second trailer is an estimated location. Therefore, the target detection box that is exactly the same size as the second trailer may have errors. Thus, selecting a target detection box that is larger than the size of the second trailer can basically ensure that the second trailer is located within the target detection box and avoid problems such as false detection.

[0069] S230. Project each point cloud within the target detection box onto the horizontal plane to obtain each projection point.

[0070] S240. Determine the target area where the preset reference point is located based on the heading angle of the unmanned vehicle; determine the preset reference point in the target area based on the reference side of the target detection box and the preset distance.

[0071] The preset distance is determined based on the size of the target detection box. The target area can be the area to one side pointed to by the autonomous vehicle's heading angle. The side of the target detection box facing the forward trailer can be used as the front side, the side opposite the front side as the rear side, and of the remaining two sides, the side closer to the target area is the reference side.

[0072] Specifically, the autonomous vehicle's heading angle can be used to determine its direction of travel. Furthermore, the autonomous vehicle and the tow truck convoy can be used as a dividing line to separate two areas, and the area on the side facing the autonomous vehicle's direction of travel can be designated as the target area. Then, a reference side can be determined based on the target detection bounding box and the target area. Starting from the center point of the reference side, along a direction perpendicular to the reference side and pointing towards the target area, a point at a preset distance from the center point of the reference side is designated as a preset reference point.

[0073] An example, a schematic diagram of the preset reference point is shown below. Figure 5As shown. Based on the target detection box and the heading angle of the autonomous vehicle, a preset reference point is set on the inside of the turn. The position of this preset reference point can be related to the size of the target detection box. On the inside of the turn, the distance between the preset reference point and the center point of the reference side in the target detection box is half the length of the reference side. That is, the preset distance is half the length of the reference side. Based on this, the preset reference point can be determined.

[0074] S250: Determine multiple target sectors based on preset reference points and preset angular resolution.

[0075] The preset angular resolution can be a pre-defined angle used to divide the sectors, such as 1°, 2°, etc., which can be set according to actual needs and is not specifically limited in this example. The target sector can be any sector obtained by dividing the sector with a preset reference point as the starting point and a preset angular resolution as the resolution.

[0076] Specifically, starting from a preset reference point, rays are emitted in all directions according to a preset angular resolution. The area between each pair of adjacent rays can be considered as a target sector. Based on this, multiple target sectors can be determined.

[0077] S260. For each target sector, if there is at least one projection point in the target sector, the projection point in the target sector that is closest to the preset reference point is taken as the edge point in the target sector. If there is no projection point in the target sector, there is no edge point in the target sector.

[0078] Specifically, the same judgment and processing can be performed for each target sector. Taking one target sector as an example, if there is at least one projection point in the target sector, it is considered that the sector covers the second trailer. Therefore, the projection point closest to the preset reference point is determined from these projection points as the edge point corresponding to the target sector, thus filtering out noise points that interfere with edge detection. If there is no projection point in the target sector, it can be intuitively shown that there are no edge points in the target sector. Based on the above processing, multiple edge points can be obtained for subsequent fitting of the side of the second trailer, i.e., the fitting edge.

[0079] For example, a schematic diagram of the target sector is shown below. Figure 6 As shown. The projection points within the target detection box are divided into target sectors. Each target sector may or may not contain projection points. For target sectors with projection points, the method for determining the projection point in the target sector that is closest to the preset reference point is as follows:

[0080] Establish a rectangular coordinate system with a preset reference point as the origin. Calculate the distance between the preset reference point and each projection point in the target sector. Select the projection point in the target sector that is closest to the preset reference point as the edge point within the target sector.

[0081] A rectangular coordinate system is established with the preset reference point as the origin, and a polar coordinate system is established with the preset reference point as the pole. The rectangular coordinates of each projection point in the target sector are determined, and each rectangular coordinate is transformed into a polar coordinate system to obtain each polar coordinate. The polar radius in each polar coordinate is used as the distance between the preset reference point and each projection point in the target sector. The projection point with the smallest polar radius in the target sector is used as the edge point in the target sector.

[0082] A schematic diagram of the edge points is shown below. Figure 7 As shown, the projection point (black point) that is closest to the preset reference point in each target sector is retained, while the projection points (gray points) that are farther away are removed. This filters out the noise that interferes with the detection of the side of the trailer, and the remaining projection points are the edge points.

[0083] S270. According to the preset straight line fitting algorithm, the edge points are fitted, and the straight line segment obtained by the fitting process is used as the fitting edge of the second trailer.

[0084] The preset line fitting algorithm can be the least squares method, gradient descent method, or Hough transform method, etc., which are used to fit discrete points to a straight line.

[0085] S280. Based on the fitted edge and the size of the second trailer, determine the pose of the second trailer, use the second trailer as the new first trailer, and use the trailer adjacent to and following the second trailer as the new second trailer. Repeat the step of determining the pose of the second trailer until the second trailer is the last trailer.

[0086] The tow truck pose detection method provided in this embodiment determines the target angle between the second tow truck and the first tow truck based on the reference angle between the first tow truck and the reference tow truck. Based on the pose of the first tow truck, the target angle, and the size of the second tow truck, a target detection box corresponding to the second tow truck is determined to increase the range of the point cloud corresponding to the second tow truck and avoid omissions. Furthermore, based on the heading angle of the autonomous vehicle, a target area containing a preset reference point is determined. Based on the reference side of the target detection box and a preset distance, a preset reference point is determined within the target area to facilitate more accurate edge detection of the second tow truck. Finally, based on the preset reference point and... By setting a preset angular resolution and determining multiple target sectors, for each target sector, if there is at least one projection point in the target sector, the projection point closest to the preset reference point in the target sector is taken as the edge point in the target sector. If there is no projection point in the target sector, there is no edge point in the target sector, in order to reduce noise interference and improve the signal-to-noise ratio. According to the preset straight line fitting algorithm, the edge points are fitted, and the straight line segment obtained by the fitting process is used as the fitting edge of the second tow bucket. This achieves the effect of improving point cloud density, improving signal-to-noise ratio, and thus improving the detection accuracy of tow bucket pose when the rear tow bucket is blocked and the distance from the lidar is far.

[0087] Figure 8 This is a flowchart of another bucket pose detection method according to an embodiment of this disclosure. Figure 8 As shown, the method may specifically include the following steps:

[0088] Step 1: Determine the pose of the first trailer section: Using the lidar installed on the unmanned vehicle, detect the side of the first trailer section facing the unmanned vehicle, and then, based on the known size parameters of the first trailer section, recover the pose of the first trailer section.

[0089] Of course, the position of the first trailer can also be restored based on the position of the unmanned vehicle, similar to steps 2-7, which will not be elaborated here.

[0090] Step 2: Determine the coverage of the second trailer section by the field of view: After the autonomous vehicle enters the curve, based on the vehicle's heading angle, driving distance, and the pose of the first trailer section detected in Step 1, determine whether the field of view (FoV) of the vehicle's LiDAR can cover the trailer section to be detected behind it (the second trailer section). If yes, proceed to Step 3; otherwise, skip the subsequent detection to reduce time consumption and unnecessary false detections.

[0091] Step 3: Generate target detection boxes: Using the pose of the first trailer section, generate the target detection box corresponding to the second trailer section (if the current target is the Nth (N>1) trailer section, then use the pose of the (N-1)th section to generate the target detection box corresponding to the Nth trailer section). This target detection box is placed behind the first trailer section according to the connection relationship between the trailer sections, and the deflection angle is determined according to the angle relationship between the first trailer section and the unmanned vehicle in front (if the current target is the Nth (N>2) trailer section, then determine the deflection angle by judging the angle relationship between the (N-1)th and (N-2)th trailer sections). This target detection box needs to encompass the second trailer section, so its size is generally larger than the actual size of the trailer section (if the current target is the Nth (N>2) trailer section, then the target detection box needs to encompass the Nth trailer section).

[0092] Step 4: Point cloud projection compression within the target detection box: Compress the 3D points (point cloud points) within the target detection box into 2D. Since the vertical height difference of the autonomous vehicle is not significant, directly treating the 3D points as 2D points (projection points) makes the error on the horizontal plane negligible. Furthermore, this increases the density, making the outline of the trailer clearer on the horizontal plane and overcoming the problem of sparse point cloud corresponding to the trailer behind.

[0093] Step 5: Set a preset reference point: Based on the pose of the target detection box and the heading angle of the autonomous vehicle, set a preset reference point on the inside of the turn. The position of this preset reference point is related to the size of the second trailer section (if the current detection is of the Nth (N>1) trailer section, then the position of the preset reference point is related to the size of the Nth trailer section). The preset reference point should be located approximately in the middle of one side of the target detection box, and the distance between the preset reference point and the reference side of the target detection box should be approximately half the length of the reference side of the target detection box. Figure 6 As shown.

[0094] Step 6: Projection point filtering: Using a preset reference point as the center, convert the projection points within the target detection box from two-dimensional Cartesian coordinates to polar coordinates. Then, divide the angle with a certain angular resolution, discard points with larger distances (noise points), and retain the closest points (edge ​​points).

[0095] The specific method is as follows: the target detection box is divided into multiple small sectors (target sectors) with a certain angular resolution, using a preset reference point as the origin. Each small sector may or may not contain a projection point. For small sectors with projection points, the projection point closest to the preset reference point is retained, while the projection points that are farther away are removed. This filters out the noise that interferes with the detection of the side of the trailer.

[0096] Step 7: Bucket pose detection: Use the remaining valid points (edge ​​points) in the target detection box from step 6 to fit a straight line segment (fitted edge), and then reconstruct the bucket frame based on the known bucket size in the system, thus obtaining the bucket pose.

[0097] Step 8: Loop Inspection: Continue inspecting backwards until all trailers have been inspected.

[0098] The bucket pose detection method provided in this embodiment does not rely entirely on estimation to obtain the bucket pose, nor does it simply detect the bucket pose based on a fixed detection box. Instead, it dynamically obtains the target detection box of the rear bucket based on the pose and connection relationship of the bucket in front, thus detecting the bucket pose more effectively. This more effectively limits the approximate range of the bucket point cloud and improves detection accuracy. The method of compressing the 3D point cloud into 2D and then using preset reference points to transform these 2D points to polar coordinates for noise filtering increases the point cloud density while reducing noise interference, significantly improving the signal-to-noise ratio and detection accuracy. Furthermore, it can dynamically select whether to skip unnecessary detections based on the coverage of the LiDAR, avoiding false detections and reducing algorithm processing time.

[0099] Figure 9 This is a schematic diagram of the structure of a bucket posture detection device according to an embodiment of this disclosure. Figure 9 As shown: The device includes: a target detection box determination module 310, a projection point determination module 320, a fitting edge determination module 330, and a pose determination module 340.

[0100] The target detection box determination module 310 is used to acquire the pose of the first trailer and determine the target detection box corresponding to the second trailer based on the pose of the first trailer; wherein the second trailer is located after and adjacent to the first trailer; the projection point determination module 320 is used to project each point cloud in the target detection box onto a horizontal plane to obtain each projection point; the fitting edge determination module 330 is used to determine edge points from each projection point based on a preset reference point, and determine the fitting edge of the second trailer based on the edge points; the pose determination module 340 is used to determine the pose of the second trailer based on the fitting edge and the size of the second trailer, take the second trailer as the new first trailer, and take the trailer adjacent to and located after the second trailer as the new second trailer, and repeat the step of determining the pose of the second trailer until the second trailer is the last trailer.

[0101] Based on the above example, the first trailer is a trailer adjacent to and located behind the unmanned vehicle. The target detection box determination module 310 is further used to acquire the pose of the unmanned vehicle, determine the target detection box corresponding to the first trailer based on the pose of the unmanned vehicle; project each point cloud in the target detection box onto a horizontal plane to obtain each projection point; determine edge points from each projection point based on a preset reference point, and determine the fitted edge of the first trailer based on the edge points; determine the pose of the first trailer based on the fitted edge and the size of the first trailer; or, determine the detection edge of the first trailer based on the lidar on the unmanned vehicle, and determine the pose of the first trailer based on the detection edge and the size of the first trailer.

[0102] Based on the above example, before determining the target detection box corresponding to the second trailer based on the pose of the first trailer, the device further includes: a coverage detection module, used to determine the coverage of the lidar on the unmanned vehicle covering the second trailer based on the heading angle of the unmanned vehicle, the travel distance of the unmanned vehicle, and the pose of the first trailer; if the coverage covers the second trailer, then the step of determining the pose of the second trailer is executed; if the coverage does not cover the second trailer, then the step of determining the pose of the second trailer is stopped.

[0103] Based on the above example, the target detection box determination module 310 is further configured to determine the target angle between the second trailer and the first trailer based on the reference angle between the first trailer and the reference trailer; wherein the reference trailer is a trailer or unmanned vehicle adjacent to and located in front of the first trailer; and to determine a target detection box corresponding to the second trailer based on the pose of the first trailer, the target angle, and the size of the second trailer; wherein the size of the target detection box is larger than the size of the second trailer.

[0104] Based on the above example, before determining the edge point from the projection points according to the preset reference point, the device further includes: a preset reference point determination module, used to determine the target area where the preset reference point is located according to the heading angle of the unmanned vehicle; and to determine the preset reference point in the target area according to the reference side of the target detection box and a preset distance; wherein the preset distance is determined according to the size of the target detection box.

[0105] Based on the above example, the fitting edge determination module 330 is further configured to determine multiple target sectors according to the preset reference point and the preset angular resolution; for each target sector, if there is at least one projection point in the target sector, the projection point in the target sector that is closest to the preset reference point is taken as the edge point in the target sector; if there is no projection point in the target sector, there is no edge point in the target sector.

[0106] Based on the above example, the fitting edge determination module 330 is further configured to perform fitting processing on the edge points according to a preset straight line fitting algorithm, and use the straight line segment obtained by the fitting processing as the fitting edge of the second trailer.

[0107] The bucket pose detection device provided in this embodiment can execute the steps in the bucket pose detection method provided in this embodiment, and has the execution steps and beneficial effects, which will not be repeated here.

[0108] Figure 10 This is a schematic diagram of the structure of an electronic device according to an embodiment of this disclosure. See below for details. Figure 10 It shows a schematic diagram of a structure suitable for implementing the electronic device 400 in the embodiments of this disclosure. Figure 10 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments disclosed herein.

[0109] like Figure 10 As shown, the electronic device 400 may include a processing device (e.g., a central processing unit, a graphics processor, etc.) 401, which can perform various appropriate actions and processes to implement the methods of the embodiments described herein, based on a program stored in a read-only memory (ROM) 402 or a program loaded from a storage device 408 into a random access memory (RAM) 403. The RAM 403 also stores various programs and data required for the operation of the electronic device 400. The processing device 401, ROM 402, and RAM 403 are interconnected via a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0110] In particular, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts, thereby implementing the bucket pose detection method as described above. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 409, or installed from a storage device 408, or installed from a ROM 402. When the computer program is executed by the processing device 401, it performs the functions defined in the methods of embodiments of this disclosure.

[0111] It should be noted that the computer-readable medium described in this disclosure can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this disclosure, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus, or device. In this disclosure, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0112] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device. The aforementioned computer-readable medium carries one or more programs that, when executed by the electronic device, cause the electronic device to:

[0113] The pose of the first trailer is obtained, and a target detection box corresponding to the second trailer is determined based on the pose of the first trailer; wherein the second trailer is located after the first trailer and adjacent to the first trailer.

[0114] Project each point cloud within the target detection box onto a horizontal plane to obtain each projection point;

[0115] Based on the preset reference points, edge points are determined from each projection point, and the fitting edge of the second bucket is determined based on the edge points;

[0116] Based on the fitted edge and the size of the second trailer, the pose of the second trailer is determined, the second trailer is designated as the new first trailer, and the trailer adjacent to and following the second trailer is designated as the new second trailer. The step of determining the pose of the second trailer is repeated until the second trailer is the last trailer.

[0117] Optionally, when one or more of the above-described procedures are executed by the electronic device, the electronic device may also perform other steps described in the above embodiments.

[0118] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0119] The above description is merely a preferred embodiment of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features disclosed in this disclosure that have similar functions.

Claims

1. A method for detecting the position and posture of a tow bucket, characterized in that, The method includes: The pose of the first trailer is obtained, and a target detection box corresponding to the second trailer is determined based on the pose of the first trailer; wherein the second trailer is located after the first trailer and adjacent to the first trailer. Project each point cloud within the target detection box onto a horizontal plane to obtain each projection point; Multiple target sectors are determined based on preset reference points and preset angular resolution; For each target sector, if there is at least one projection point in the target sector, then the projection point in the target sector that is closest to the preset reference point is taken as the edge point in the target sector; if there is no projection point in the target sector, then there is no edge point in the target sector. Based on the edge points, the fitting edge of the second trailer is determined; wherein, the preset reference point is a pre-set reference point outside the target detection frame and inside the turn, used to determine the edge points in the projection points; Based on the fitted edge and the size of the second trailer, the pose of the second trailer is determined, the second trailer is designated as the new first trailer, and the trailer adjacent to and following the second trailer is designated as the new second trailer. The step of determining the pose of the second trailer is repeated until the second trailer is the last trailer.

2. The method according to claim 1, characterized in that, The first trailer is adjacent to and located behind the unmanned vehicle. Obtaining the pose of the first trailer includes: The pose of the unmanned vehicle is obtained; based on the pose of the unmanned vehicle, a target detection bounding box corresponding to the first trailer is determined; the point cloud points within the target detection bounding box are projected onto a horizontal plane to obtain projection points; edge points are determined from the projection points based on preset reference points, and the fitted edge of the first trailer is determined based on the edge points; the pose of the first trailer is determined based on the fitted edge and the size of the first trailer; or, The detection edge of the first trailer is determined based on the lidar on the unmanned vehicle, and the pose of the first trailer is determined based on the detection edge and the size of the first trailer.

3. The method according to claim 1, characterized in that, Before determining the target detection box corresponding to the second trailer based on the pose of the first trailer, the method further includes: Based on the heading angle of the unmanned vehicle, the travel distance of the unmanned vehicle, and the position of the first trailer, the coverage of the lidar on the unmanned vehicle to the second trailer is determined. If the coverage extends to the second trailer, then the step of determining the pose of the second trailer is performed; If the coverage situation is that the second trailer is not covered, then the step of determining the pose of the second trailer is stopped.

4. The method according to claim 1, characterized in that, The step of determining the target detection box corresponding to the second trailer based on the pose of the first trailer includes: The target angle between the second trailer and the first trailer is determined based on the reference angle between the first trailer and the reference trailer; wherein the reference trailer is a trailer or unmanned vehicle that is adjacent to and in front of the first trailer. Based on the pose of the first trailer, the target angle, and the size of the second trailer, a target detection frame corresponding to the second trailer is determined; wherein the size of the target detection frame is larger than the size of the second trailer.

5. The method according to claim 1, characterized in that, Before determining the edge points from the projection points based on the preset reference points, the method further includes: The target area where the preset reference point is located is determined based on the heading angle of the unmanned vehicle; Based on the reference side of the target detection box and a preset distance, a preset reference point is determined in the target area; wherein, the preset distance is determined according to the size of the target detection box.

6. The method according to claim 1, characterized in that, Determining the fitted edge of the second bucket based on the edge points includes: According to the preset straight line fitting algorithm, the edge points are fitted, and the straight line segment obtained by the fitting process is used as the fitted edge of the second trailer.

7. A bucket position detection device, characterized in that, include: The target detection box determination module is used to obtain the pose of the first trailer and determine the target detection box corresponding to the second trailer based on the pose of the first trailer; wherein the second trailer is located after the first trailer and adjacent to the first trailer. The projection point determination module is used to project each point cloud within the target detection box onto a horizontal plane to obtain each projection point. The fitting edge determination module is used to determine multiple target sectors based on a preset reference point and a preset angular resolution. For each target sector, if there is at least one projection point in the target sector, the projection point in the target sector that is closest to the preset reference point is taken as the edge point in the target sector. If there is no projection point in the target sector, there is no edge point in the target sector, and the fitting edge of the second trailer is determined based on the edge point. The preset reference point is a pre-set reference point outside the target detection frame and inside the turn, used to determine the edge point among the projection points. The pose determination module is used to determine the pose of the second trailer based on the fitted edge and the size of the second trailer, to designate the second trailer as the new first trailer, and to designate the trailer adjacent to and following the second trailer as the new second trailer, and to repeat the step of determining the pose of the second trailer until the second trailer is the last trailer.

8. An electronic device, characterized in that, The electronic device includes: One or more processors; Storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the bucket pose detection method as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the bucket pose detection method as described in any one of claims 1-6.

Images