Vehicle chassis position positioning method and device, electronic equipment and storage medium

Abstract

The invention relates to a vehicle chassis position positioning method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining a first 2D detection frame and a second 2D detection frame corresponding to a to-be-detected vehicle, determining the target orientation and the target offset of the to-be-detected vehicle based on the first 2D detection frame and the second 2D detection frame, and determining the target orientation and the target offset of the to-be-detected vehicle based on the target orientation and the target offset. And based on the target orientation and the target offset, determining a bottom edge quadrilateral detection frame corresponding to the to-be-detected vehicle, so as to determine the chassis position of the to-be-detected vehicle according to the bottom edge quadrilateral detection frame. According to the method, the chassis position can be accurately positioned based on the target orientation and the target offset, and compared with a scheme of a 3D detection frame, the data processing speed can be increased, and deployment of the edge end is facilitated.

Description

Technical Field

[0001] The present invention relates to the field of machine learning and positioning technology, and in particular to a vehicle chassis positioning method, device, electronic device and storage medium. Background Art

[0002] Traffic incident detection systems need to accurately locate the vehicle when determining whether illegal incidents such as occupying the emergency lane, illegal lane change, and driving in a non-compliant lane have occurred. For trucks with a long body, the method of locating the vehicle chassis based on the bounding box cannot describe the vehicle's position well due to the tilt of the imaging perspective.

[0003] In the prior art, the solutions to the above problems are:

[0004] Post-processing logic correction: Neither the center of gravity of the bounding box nor the center point of the bottom edge can describe the vehicle position well. The post-processing logic takes the middle position between these two points to roughly represent the vehicle position. The disadvantage of this approach is that the error is large and it is not applicable to all situations.

[0005] Use 3D object detection: 3D object detection can output eight fixed points representing a cube. Only four key points on the bottom surface of the cube need to be obtained to accurately locate the vehicle chassis. The disadvantage of this solution is that the 3D object detection model is generally slower inference and more computationally intensive than the single-stage object detection model, which is not conducive to edge deployment. Summary of the invention

[0006] The technical problem to be solved by the present invention is to provide a vehicle chassis position positioning method, device, electronic equipment and storage medium, aiming to solve at least one of the above technical problems.

[0007] In the first aspect, the technical solution of the present invention to solve the above technical problem is as follows: a vehicle chassis position positioning method, the method comprising:

[0008] Obtain a first 2D detection frame and a second 2D detection frame corresponding to the vehicle to be detected, wherein the first 2D detection frame is a 2D detection frame corresponding to the front of the vehicle to be detected perpendicular to the ground, and the second 2D detection frame is a 2D detection frame corresponding to the rear of the vehicle to be detected perpendicular to the ground;

[0009] Based on the first 2D detection frame and the second 2D detection frame, determine the target orientation and target offset of the vehicle to be detected, the target offset includes an x-axis offset and a y-axis offset, the target offset is the side length of the 2D detection frame, and the target orientation represents the relative position of the first 2D detection frame and the second 2D detection frame of the vehicle to be detected;

[0010] Based on the target orientation and the target offset, a bottom quadrilateral detection frame corresponding to the vehicle to be detected is determined, so as to determine the chassis position of the vehicle to be detected according to the bottom quadrilateral detection frame.

[0011] The beneficial effect of the present invention is that: on the basis of the first 2D detection frame corresponding to the vehicle to be detected and the second 2D detection frame corresponding to the vehicle to be detected, the characteristics that can characterize the chassis position of the vehicle to be detected, that is, the target orientation and the target offset, are determined, so that the chassis position can be accurately located based on the target orientation and the target offset, and compared with the 3D detection frame solution, the data processing speed can be accelerated, which is convenient for edge deployment.

[0012] Based on the above technical solution, the present invention can also be improved as follows.

[0013] Furthermore, the above-mentioned obtaining of the first 2D detection frame and the second 2D detection frame corresponding to the vehicle to be detected includes:

[0014] Acquire an image corresponding to the vehicle to be detected;

[0015] Based on the pre-trained YOLO object detection model and the image, the first 2D detection box and the second 2D detection box are obtained.

[0016] The beneficial effect of adopting the above further scheme is that based on the pre-trained YOLO target detection model, the first 2D detection frame and the second 2D detection frame can be accurately and quickly determined.

[0017] Further, the determining of the target orientation and the target offset of the vehicle to be detected based on the first 2D detection frame and the second 2D detection frame includes:

[0018] If the position of the first 2D detection frame is to the left of the position of the second 2D detection frame, determining that the target orientation of the vehicle to be detected is positive;

[0019] If the position of the first 2D detection frame is to the right of the position of the second 2D detection frame, determining that the target orientation of the vehicle to be detected is negative;

[0020] The side of the first 2D detection frame parallel to the ground is determined as the x-axis offset in the target offset, and the side of the second 2D detection frame perpendicular to the ground is determined as the y-axis offset in the target offset.

[0021] The beneficial effect of adopting the above further scheme is that since the first 2D detection frame and the second 2D detection frame correspond to the front position and the rear position of the vehicle to be detected respectively, the front position and the rear position can correspond to the chassis position of the vehicle. Therefore, based on the first 2D detection frame and the second 2D detection frame, the target orientation and target offset of the vehicle to be detected can be accurately determined.

[0022] Further, the obtaining of the first 2D detection frame and the second 2D detection frame corresponding to the vehicle to be detected and the determination of the target orientation and the target offset of the vehicle to be detected based on the first 2D detection frame and the second 2D detection frame are determined based on a pre-trained detection model, and the detection model is trained based on the following method:

[0023] Acquire training data, wherein the training data includes sample data of each sample vehicle, each of the sample data includes a first identifier and a second identifier corresponding to the sample vehicle, each of the first identifiers represents a real vehicle orientation of the corresponding sample vehicle, and each of the second identifiers represents a real offset of the corresponding sample vehicle;

[0024] According to the training data, the initial model is trained to obtain a first prediction result and a second prediction result corresponding to each of the sample vehicles, each of the first prediction results is a predicted vehicle orientation of the corresponding sample vehicle, and each of the second prediction results is a predicted offset of the corresponding sample vehicle;

[0025] Determining a total loss value of the initial model according to the first prediction result, the second prediction result, the first identifier and the second identifier of each of the sample vehicles;

[0026] When the total loss value satisfies a preset training end condition, the initial model that satisfies the training end condition is determined as the detection model; when the total loss value does not meet the preset training end condition, the parameters of the initial model are adjusted to retrain the initial model based on the adjusted parameters until the total loss value meets the preset training end condition.

[0027] The beneficial effect of adopting the above further solution is that, based on the above model training process, the target orientation and target offset of the vehicle to be detected can be accurately determined.

[0028] Further, the total loss value of the initial model is determined according to the first prediction result, the second prediction result, the first identifier and the second identifier of each of the sample vehicles, including:

[0029] Determining a first loss value according to the first prediction result and the first identifier of each of the sample vehicles;

[0030] Determine a second loss value according to the second prediction result and the second identification of each of the sample vehicles;

[0031] A total loss value of the initial model is determined according to the first loss value and the second loss value.

[0032] The beneficial effect of adopting the above further scheme is that during the training process, not only the loss between the actual offset and the predicted offset is taken into account, but also the loss between the actual vehicle orientation and the predicted vehicle orientation is taken into account, thereby further improving the training accuracy of the model.

[0033] Further, for each of the sample vehicles, the sample data of the sample vehicle is obtained based on the following method:

[0034] Obtaining a 3D detection frame corresponding to the sample vehicle;

[0035] Determine a third 2D detection frame corresponding to the sample vehicle from the 3D detection frame, where the third 2D detection frame is a 2D detection frame corresponding to the entire vehicle of the sample vehicle;

[0036] Based on the 3D detection frame, determining a first 2D detection frame and a second 2D detection frame corresponding to the sample vehicle;

[0037] Determine a real vehicle orientation and a real offset of the sample vehicle based on the first 2D detection frame, the second 2D detection frame, and the third 2D detection frame corresponding to the sample vehicle;

[0038] Based on the actual vehicle orientation and the actual offset of the sample vehicle, sample data of the sample vehicle is determined.

[0039] The beneficial effect of adopting the above further solution is that, for the sample data of each sample vehicle, the real vehicle orientation and real offset of the sample vehicle can be more accurately described based on the 3D detection frame, making the sample data more accurate.

[0040] Further, for each of the sample vehicles, the method further comprises:

[0041] Acquire a sample image corresponding to the sample vehicle;

[0042] Based on the pre-trained YOLO object detection model and the sample image, at least two fourth 2D detection frames corresponding to the sample vehicle are obtained, each of the fourth 2D detection frames being a 2D detection frame corresponding to the entire vehicle of the sample vehicle;

[0043] Matching the third 2D detection frame with the at least two fourth 2D detection frames to obtain a target 2D detection frame that matches the third 2D detection frame;

[0044] The determining the real vehicle orientation and the real offset of the sample vehicle based on the first 2D detection frame, the second 2D detection frame, and the third 2D detection frame corresponding to the sample vehicle includes:

[0045] Based on the first 2D detection frame, the second 2D detection frame and the target 2D detection frame corresponding to the sample vehicle, a real vehicle orientation and a real offset of the sample vehicle are determined.

[0046] The beneficial effect of adopting the above further scheme is that there is an error in directly determining the third 2D detection frame corresponding to the sample vehicle from the 3D detection frame, that is, it cannot accurately express the whole vehicle position of the vehicle. The at least two fourth 2D detection frames detected based on the pre-trained yolo target detection model can more accurately express the whole vehicle position of the vehicle. Therefore, the target 2D detection frame obtained by matching the third 2D detection frame and the at least two fourth 2D detection frames can more accurately express the whole vehicle position of the vehicle, so that the actual vehicle orientation and actual offset of the sample vehicle subsequently determined based on the target 2D detection frame are more accurate.

[0047] In a second aspect, in order to solve the above technical problems, the present invention also provides a vehicle chassis position positioning device, the device comprising:

[0048] An acquisition module is used to acquire a first 2D detection frame and a second 2D detection frame corresponding to the vehicle to be detected, wherein the first 2D detection frame is a 2D detection frame corresponding to the front of the vehicle to be detected perpendicular to the ground, and the second 2D detection frame is a 2D detection frame corresponding to the rear of the vehicle to be detected perpendicular to the ground;

[0049] A processing module, configured to determine a target orientation and a target offset of the vehicle to be detected based on the first 2D detection frame and the second 2D detection frame, wherein the target offset includes an x-axis offset and a y-axis offset, the target offset is a side length of the 2D detection frame, and the target orientation represents a relative position between the first 2D detection frame and the second 2D detection frame of the vehicle to be detected;

[0050] A positioning module is used to determine a bottom quadrilateral detection frame corresponding to the vehicle to be detected based on the target orientation and the target offset, so as to determine the chassis position of the vehicle to be detected according to the bottom quadrilateral detection frame.

[0051] In a third aspect, in order to solve the above-mentioned technical problem, the present invention also provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the computer program, the vehicle chassis position positioning method of the present application is implemented.

[0052] In a fourth aspect, in order to solve the above-mentioned technical problem, the present invention further provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the vehicle chassis position positioning method of the present application is implemented.

[0053] Additional aspects and advantages of the present application will be partially given in the following description, which will become apparent from the following description, or will be understood through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for describing the embodiments of the present invention are briefly introduced below.

[0055] Figure 1 A schematic flow chart of a vehicle chassis position positioning method provided by one embodiment of the present invention;

[0056] Figure 2 A schematic diagram of a 2D detection frame and an offset provided by an embodiment of the present invention;

[0057] Figure 3 A schematic diagram of a process flow of a bottom quadrilateral detection frame provided by an embodiment of the present invention;

[0058] Figure 4 A schematic diagram of a training data determination process provided by an embodiment of the present invention;

[0059] Figure 5 A schematic diagram of a model training process provided by an embodiment of the present invention;

[0060] Figure 6 A schematic structural diagram of a vehicle chassis position positioning device provided by one embodiment of the present invention;

[0061] Figure 7 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0062] The principles and features of the present invention are described below. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0063] The technical solution of the present invention and how the technical solution of the present invention solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present invention will be described below in conjunction with the accompanying drawings.

[0064] The solution provided by the embodiment of the present invention can be applied to any application scenario where the chassis position of a vehicle needs to be determined. The solution provided by the embodiment of the present invention can be executed by any electronic device, for example, it can be a user's terminal device, including at least one of the following: a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, a smart TV, and a smart vehicle-mounted device.

[0065] The embodiment of the present invention provides a possible implementation method, such as Figure 1 As shown, a flow chart of a vehicle chassis location positioning method is provided. The solution can be executed by any electronic device, for example, a terminal device, or a terminal device and a server. For the convenience of description, the method provided by the embodiment of the present invention will be described below by taking the terminal device as the execution subject as an example. Figure 1 The method may include the following steps:

[0066] Step S10, obtaining a first 2D detection frame and a second 2D detection frame corresponding to the vehicle to be detected, wherein the first 2D detection frame is a 2D detection frame corresponding to the front of the vehicle to be detected perpendicular to the ground, and the second 2D detection frame is a 2D detection frame corresponding to the rear of the vehicle to be detected perpendicular to the ground;

[0067] Step S20, based on the first 2D detection frame and the second 2D detection frame, determining a target orientation and a target offset of the vehicle to be detected, wherein the target offset includes an x-axis offset and a y-axis offset, the target offset is the side length of the 2D detection frame, and the target orientation represents a relative position between the first 2D detection frame and the second 2D detection frame of the vehicle to be detected;

[0068] Step S30, based on the target orientation and the target offset, determining a bottom quadrilateral detection frame corresponding to the vehicle to be detected, so as to determine the chassis position of the vehicle to be detected according to the bottom quadrilateral detection frame.

[0069] Through the method of the present invention, on the basis of the first 2D detection frame corresponding to the vehicle to be detected and the second 2D detection frame corresponding to the vehicle to be detected, the characteristics that can characterize the chassis position of the vehicle to be detected, that is, the target orientation and the target offset, are determined, so that the chassis position can be accurately located based on the target orientation and the target offset, and compared with the 3D detection frame solution, the data processing speed can be accelerated, which is convenient for edge deployment.

[0070] The solution of the present invention is further described below in conjunction with the following specific embodiments. In this embodiment, a vehicle chassis position positioning method provided may include the following steps:

[0071] Step S10, obtaining a first 2D detection frame and a second 2D detection frame corresponding to the vehicle to be detected, wherein the first 2D detection frame is a 2D detection frame corresponding to the front of the vehicle to be detected perpendicular to the ground, and the second 2D detection frame is a 2D detection frame corresponding to the rear of the vehicle to be detected perpendicular to the ground;

[0072] The vehicle to be detected refers to a vehicle whose chassis position needs to be determined, and the vehicle to be detected can be a moving vehicle or a vehicle parked at a certain place. In addition, considering that in the prior art, the chassis position of a vehicle with a long length cannot be accurately determined, the vehicle to be detected can be a vehicle with a long length, for example, a vehicle larger than a set gear, such as a trailer.

[0073] The above-mentioned obtaining of the first 2D detection frame and the second 2D detection frame corresponding to the vehicle to be detected includes:

[0074] Acquire an image corresponding to the vehicle to be detected;

[0075] Based on the pre-trained YOLO object detection model and the image, the first 2D detection box and the second 2D detection box are obtained.

[0076] It should be noted that the above-mentioned first 2D detection frame and the second 2D detection frame can be implemented based on other methods in the prior art, for example, determined from the 3D detection frame corresponding to the vehicle to be detected, that is, the first 2D detection frame and the second 2D detection frame are respectively two 2D detection frames in the 3D detection frame. This solution does not limit how the first 2D detection frame and the second 2D detection frame are obtained.

[0077] As an example, see Figure 2 , Figure 2 The quadrilateral A in is the first 2D detection box, Figure 2 The quadrilateral B in is the second 2D detection box.

[0078] Step S20, based on the first 2D detection frame and the second 2D detection frame, determining a target orientation and a target offset of the vehicle to be detected, wherein the target offset includes an x-axis offset and a y-axis offset, the target offset is the side length of the 2D detection frame, and the target orientation represents a relative position between the first 2D detection frame and the second 2D detection frame of the vehicle to be detected;

[0079] Different relative positions can be represented based on the orientation mark. For example, if the first 2D detection frame is on the left side of the second 2D detection frame, the orientation mark is positive. Conversely, if the first 2D detection frame is on the right side of the second 2D detection frame, the orientation mark is negative.

[0080] The target direction can also be understood as: characterizing the offset direction of the body of the vehicle to be detected compared to a reference direction, where the reference direction is a standard direction corresponding to the driving direction of the vehicle to be detected, and the standard direction is a north-south direction or an east-west direction.

[0081] Among them, since the first 2D detection frame and the second 2D detection frame are determined based on the image corresponding to the vehicle to be detected, due to the inclination of the imaging angle of the image, the body of the vehicle to be detected appears to be tilted. Therefore, the target offset can be understood as the offset of the vehicle to be detected in the image compared to the position corresponding to the due south and due north directions, that is, the target offset is used to describe the offset of the body of the vehicle to be detected in the image compared to the position of the body in the real scene.

[0082] As an example, see Figure 2 As shown, Figure 2 The vehicle in the image is shown with its body tilted. Figure 2 The x-axis offset and y-axis offset are shown in the figure. In the present application, the xoy coordinate system can be constructed with the lower left corner of the circumscribed rectangle of the vehicle in the image as the origin of the xoy coordinate system. The x-axis offset refers to the offset of the front position of the vehicle to be detected compared to the origin on the x-axis, and the y-axis offset refers to the offset of the rear position of the vehicle to be detected compared to the origin on the y-axis.

[0083] In the present application, the x-axis offset and the y-axis offset can be determined based on the side length of the first 2D detection frame and the side length of the second 2D detection frame, respectively. Then, one possible implementation of the above S20 is:

[0084] S201, if the position of the first 2D detection frame is to the left of the position of the second 2D detection frame, it is determined that the target orientation of the vehicle to be detected is positive, which can be recorded as 1; or, if the first 2D detection frame is to the left of the reference direction and the second 2D detection frame is to the right of the reference direction, it is determined that the target orientation of the vehicle to be detected is positive;

[0085] S202, if the position of the first 2D detection frame is to the right of the position of the second 2D detection frame, it is determined that the target orientation of the vehicle to be detected is negative, which can be recorded as -1; or, if the first 2D detection frame is biased to the right relative to the reference direction and the second 2D detection frame is biased to the left relative to the reference direction, it is determined that the target orientation of the vehicle to be detected is negative;

[0086] S203, determining a side of the first 2D detection frame parallel to the ground as an x-axis offset in the target offset, which may be recorded as offset_x, and determining a side of the second 2D detection frame perpendicular to the ground as a y-axis offset in the target offset, which may be recorded as offset_y.

[0087] In the present application, if the first 2D detection frame and the second 2D detection frame are determined based on the 3D detection frame corresponding to the vehicle to be detected, in the above step S20, another implementation method of determining the target offset based on the first 2D detection frame and the second 2D detection frame is:

[0088] S1, determining a circumscribed rectangle corresponding to the vehicle to be detected from the 3D detection frame corresponding to the vehicle to be detected, where the circumscribed rectangle is a 2D detection frame corresponding to the body of the vehicle to be detected;

[0089] S2, taking the bottom side of the first 2D detection frame corresponding to the vehicle to be detected and parallel to the ground on the circumscribed rectangle as the x-axis offset in the target offset;

[0090] S3, taking the side perpendicular to the ground corresponding to the second 2D detection frame corresponding to the vehicle to be detected on the circumscribed rectangle as the y-axis offset in the target offset.

[0091] As an example, the above-mentioned bounding rectangle can be found in Figure 2 The quadrilateral C shown in .

[0092] Since there is an error in the circumscribed rectangle determined from the 3D detection frame corresponding to the vehicle to be detected, the present application also includes:

[0093] Acquire an image corresponding to the vehicle to be detected;

[0094] Based on the pre-trained YOLO object detection model and the image, at least two fifth 2D detection frames corresponding to the sample vehicle are obtained, each of the fifth 2D detection frames being a 2D detection frame corresponding to the entire vehicle of the sample vehicle;

[0095] The circumscribed rectangle is matched with the at least two fifth 2D detection frames to obtain a target detection frame matching the circumscribed rectangle; after the target detection frame is determined, the circumscribed rectangle in the above steps S2 and S3 can be replaced with the target detection frame.

[0096] Step S30, based on the target orientation and the target offset, determining a bottom quadrilateral detection frame corresponding to the vehicle to be detected, so as to determine the chassis position of the vehicle to be detected according to the bottom quadrilateral detection frame.

[0097] Among them, one implementable method of the above S30 is: based on the target offset, the size of the bottom edge quadrilateral detection frame can be determined, and based on the target orientation, the offset direction of the bottom edge quadrilateral detection frame compared to the reference direction can be determined, and thus based on the offset direction and the size of the bottom edge quadrilateral detection frame, the chassis position is determined, wherein the reference direction is the standard direction corresponding to the driving direction of the vehicle to be detected, and the standard direction is the north-south direction or the east-west direction.

[0098] As an example, see Figure 3 Schematic diagram of the bottom quadrilateral detection box shown.

[0099] Optional, see Figure 4 and Figure 5 , the above S10 and S20 may be determined based on a pre-trained detection model, and the detection model is trained based on the following method:

[0100] S11, obtain training data ( Figure 4 ), wherein the training data includes sample data of each sample vehicle, each of the sample data includes a first identifier and a second identifier of the corresponding sample vehicle, each of the first identifiers represents a real vehicle orientation of the corresponding sample vehicle, and each of the second identifiers represents a real offset of the corresponding sample vehicle;

[0101] S12, training the initial model according to the training data to obtain a first prediction result and a second prediction result corresponding to each of the sample vehicles, wherein each of the first prediction results is a predicted vehicle orientation of the corresponding sample vehicle, and each of the second prediction results is a predicted offset of the corresponding sample vehicle;

[0102] The training data can be sample images. If the training data is used as Figure 5 The input data in the training data is input into the initial model, and the image features are obtained through feature extraction, and then the first prediction result corresponding to each sample vehicle is output based on the image features (corresponding to Figure 5 The vehicle direction classification shown in toward) and the second prediction result (corresponding to Figure 5 The x and y axis offsets shown in , offset_x, offset_y).

[0103] S13, determining a total loss value of the initial model according to the first prediction result, the second prediction result, the first identifier and the second identifier of each of the sample vehicles;

[0104] S14, when the total loss value satisfies a preset training end condition, the initial model that satisfies the training end condition is determined as the detection model, and when the total loss value does not satisfy the preset training end condition, the parameters of the initial model are adjusted to retrain the initial model based on the adjusted parameters until the total loss value satisfies the preset training end condition (corresponding to Figure 5 The loss summation and gradient backpropagation shown in ).

[0105] Optionally, in the above S13, determining the total loss value of the initial model according to the first prediction result, the second prediction result, the first identifier and the second identifier of each of the sample vehicles includes:

[0106] S131, determining a first loss value (corresponding to Figure 5 Vehicle orientation classification loss shown in );

[0107] The implementation process of S131 can be implemented based on formula (1):

[0108]

[0109] Where n represents the number of images corresponding to the sample vehicle, which can also be called the number of first prediction results, and w n Represents the loss calculation weight, y n Indicates the nth first identifier, x n Indicates the nth first prediction result, L towards Represents the vehicle orientation classification loss, that is, the first loss value.

[0110] In the present application scheme, a binary cross entropy loss function may be used to calculate the first loss value.

[0111] S132, determining a second loss value (corresponding to Figure 5 offset regression loss as shown in );

[0112] Wherein, each second prediction result includes a prediction result corresponding to the x-axis offset and a prediction result corresponding to the y-axis offset, and the second loss value includes a loss value corresponding to the x-axis offset and a loss value corresponding to the y-axis offset;

[0113] The implementation process of S132 can be implemented based on formula (2) and formula (3):

[0114]

[0115] Where n represents the number of second prediction results, g n Indicates the nth second identifier, xn Indicates the nth second prediction result of the x-axis offset, y n Indicates the nth second prediction result of the y-axis offset, L x Indicates the loss value corresponding to the x-axis offset, L y Indicates the loss value corresponding to the y-axis offset.

[0116] S133: Determine a total loss value of the initial model according to the first loss value and the second loss value.

[0117] Optionally, for the sample data of each sample vehicle, the sample data also includes a third identifier, a fourth identifier and a fifth identifier, wherein the third identifier represents a real 2D detection frame of the sample vehicle in the corresponding image, the fourth identifier represents a classification result of the sample vehicle, that is, what type of vehicle it is; and the fifth identifier represents a real probability that the sample vehicle exists in the region of interest;

[0118] After inputting the training data into the initial model, the predicted target detection result, predicted classification result and predicted probability of the image corresponding to each sample vehicle can also be obtained. The predicted target detection result refers to the predicted 2D detection box of the sample image in the corresponding image, the predicted classification result refers to the category of the sample vehicle, and the predicted probability refers to the probability of the sample vehicle in the area of ​​interest.

[0119] The method may further include:

[0120] Based on each of the third identifiers and each predicted target detection result, a third loss value (corresponding to Figure 5 The bounding box regression loss shown in );

[0121] Based on each fourth identification and each predicted classification result, a fourth loss value (corresponding to Figure 5 Class loss shown in );

[0122] Based on each fifth identification and each predicted probability, determine the fifth loss value ( Figure 5 objectness loss as shown in );

[0123] Then the above S133 may further specifically include:

[0124] A total loss value of the initial model is determined according to the first loss value, the second loss value, the third loss value, the fourth loss value and the fifth loss value.

[0125] Optionally, for each of the sample vehicles, the sample data of the sample vehicle is obtained based on the following method:

[0126] Get the 3D detection frame corresponding to the sample vehicle ( Figure 43D object detection data shown in );

[0127] Determine a third 2D detection frame corresponding to the sample vehicle from the 3D detection frame ( Figure 4 ), the third 2D detection frame is the 2D detection frame corresponding to the whole vehicle of the sample vehicle;

[0128] Based on the 3D detection frame, determine the first 2D detection frame and the second 2D detection frame corresponding to the sample vehicle; based on the first 2D detection frame, the second 2D detection frame and the third 2D detection frame corresponding to the sample vehicle, determine the real vehicle orientation and the real offset (corresponding to Figure 4 Obtaining vehicle orientation and offset as shown in );

[0129] Based on the actual vehicle orientation and the actual offset of the sample vehicle, sample data of the sample vehicle is determined.

[0130] Optionally, for each of the sample vehicles, the method further includes:

[0131] Acquire a sample image corresponding to the sample vehicle;

[0132] Based on the pre-trained YOLO object detection model and the sample image, at least two fourth 2D detection frames corresponding to the sample vehicle (each fourth 2D detection frame may also be referred to as a bounding box) are obtained, and each of the fourth 2D detection frames is a 2D detection frame corresponding to the entire vehicle of the sample vehicle;

[0133] The third 2D detection frame is matched with the at least two fourth 2D detection frames to obtain a target 2D detection frame (corresponding to Figure 4 The shown and boundingbox match);

[0134] The above-mentioned determining the real vehicle orientation and the real offset of the sample vehicle based on the first 2D detection frame, the second 2D detection frame and the third 2D detection frame corresponding to the sample vehicle includes:

[0135] Based on the first 2D detection frame, the second 2D detection frame and the target 2D detection frame corresponding to the sample vehicle, a real vehicle orientation and a real offset of the sample vehicle are determined.

[0136] Optionally, for each of the sample vehicles, determining the real vehicle orientation and the real offset of the sample vehicle based on the first 2D detection frame, the second 2D detection frame, and the third 2D detection frame corresponding to the sample vehicle includes:

[0137] Determining a real vehicle orientation corresponding to the sample vehicle based on a first 2D detection frame and a second 2D detection frame corresponding to the sample vehicle;

[0138] Based on the third 2D detection frame, the first 2D detection frame, and the second 2D detection frame, a real offset of the sample vehicle is determined.

[0139] Optionally, the determining the real offset of the sample vehicle based on the third 2D detection frame, the first 2D detection frame, and the second 2D detection frame includes:

[0140] The bottom edge of the first 2D detection frame corresponding to the sample vehicle and parallel to the ground on the third 2D detection frame is used as the x-axis offset in the real offset;

[0141] The side perpendicular to the ground corresponding to the second 2D detection frame corresponding to the sample vehicle on the third 2D detection frame is used as the y-axis offset in the real offset.

[0142] Through the solution of the present invention, the present invention outputs two additional key points (target vehicle orientation and target offset) to describe the vehicle position based on the YOLO target detection model. Compared with the solution of directly using the bounding box with post-processing logic, the vehicle position can be accurately located, and end-to-end training can be achieved, simplifying the event detection process; compared with the 3D target detection solution, the inference speed can be kept consistent with the pre-trained base target detection model while maintaining the chassis positioning accuracy.

[0143] Based on Figure 1 Based on the same principle as the method shown in , the embodiment of the present invention also provides a vehicle chassis position positioning device 20, such as Figure 6 As shown in , the vehicle chassis position positioning device 20 may include an acquisition module 210, a processing module 220 and a positioning module 230, wherein:

[0144] An acquisition module 210 is used to acquire a first 2D detection frame and a second 2D detection frame corresponding to the vehicle to be detected, wherein the first 2D detection frame is a 2D detection frame corresponding to the front of the vehicle to be detected perpendicular to the ground, and the second 2D detection frame is a 2D detection frame corresponding to the rear of the vehicle to be detected perpendicular to the ground;

[0145] A processing module 220 is used to determine a target orientation and a target offset of the vehicle to be detected based on the first 2D detection frame and the second 2D detection frame, wherein the target offset includes an x-axis offset and a y-axis offset, the target offset is a side length of the 2D detection frame, and the target orientation represents a relative position between the first 2D detection frame and the second 2D detection frame of the vehicle to be detected;

[0146] The positioning module 230 is used to determine the bottom quadrilateral detection frame corresponding to the vehicle to be detected based on the target orientation and the target offset, so as to determine the chassis position of the vehicle to be detected according to the bottom quadrilateral detection frame.

[0147] Optionally, when acquiring the first 2D detection frame and the second 2D detection frame corresponding to the vehicle to be detected, the acquisition module 210 is specifically used to:

[0148] Acquire an image corresponding to the vehicle to be detected;

[0149] Based on the pre-trained YOLO object detection model and the image, the first 2D detection box and the second 2D detection box are obtained.

[0150] Optionally, when determining the target orientation and target offset of the vehicle to be detected based on the first 2D detection frame and the second 2D detection frame, the processing module 220 is specifically configured to:

[0151] If the position of the first 2D detection frame is to the left of the position of the second 2D detection frame, determining that the target orientation of the vehicle to be detected is positive;

[0152] If the position of the first 2D detection frame is to the right of the position of the second 2D detection frame, determining that the target orientation of the vehicle to be detected is negative;

[0153] The side of the first 2D detection frame parallel to the ground is determined as the x-axis offset in the target offset, and the side of the second 2D detection frame perpendicular to the ground is determined as the y-axis offset in the target offset.

[0154] Optionally, the processing of the acquisition module 210 and the processing module 220 is determined based on a pre-trained detection model, and the detection model is trained based on the following model training module:

[0155] Acquire training data, wherein the training data includes sample data of each sample vehicle, each of the sample data includes a first identifier and a second identifier corresponding to the sample vehicle, each of the first identifiers represents a real vehicle orientation of the corresponding sample vehicle, and each of the second identifiers represents a real offset of the corresponding sample vehicle;

[0156] According to the training data, the initial model is trained to obtain a first prediction result and a second prediction result corresponding to each of the sample vehicles, each of the first prediction results is a predicted vehicle orientation of the corresponding sample vehicle, and each of the second prediction results is a predicted offset of the corresponding sample vehicle;

[0157] Determining a total loss value of the initial model according to the first prediction result, the second prediction result, the first identifier and the second identifier of each of the sample vehicles;

[0158] When the total loss value satisfies a preset training end condition, the initial model that satisfies the training end condition is determined as the detection model; when the total loss value does not meet the preset training end condition, the parameters of the initial model are adjusted to retrain the initial model based on the adjusted parameters until the total loss value meets the preset training end condition.

[0159] Optionally, when determining the total loss value of the initial model according to the first prediction result, the second prediction result, the first identifier and the second identifier of each of the sample vehicles, the model training module is specifically used to:

[0160] Determining a first loss value according to the first prediction result and the first identifier of each of the sample vehicles;

[0161] Determine a second loss value according to the second prediction result and the second identification of each of the sample vehicles;

[0162] A total loss value of the initial model is determined according to the first loss value and the second loss value.

[0163] Optionally, for each of the sample vehicles, the sample data of the sample vehicle is obtained based on the following method:

[0164] Obtaining a 3D detection frame corresponding to the sample vehicle;

[0165] Determine a third 2D detection frame corresponding to the sample vehicle from the 3D detection frame, where the third 2D detection frame is a 2D detection frame corresponding to the entire vehicle of the sample vehicle;

[0166] Based on the 3D detection frame, determining a first 2D detection frame and a second 2D detection frame corresponding to the sample vehicle;

[0167] Determine a real vehicle orientation and a real offset of the sample vehicle based on the first 2D detection frame, the second 2D detection frame, and the third 2D detection frame corresponding to the sample vehicle;

[0168] Based on the actual vehicle orientation and the actual offset of the sample vehicle, sample data of the sample vehicle is determined.

[0169] Optionally, for each of the sample vehicles, the device further includes:

[0170] A matching module is used to obtain a sample image corresponding to the sample vehicle; based on the pre-trained YOLO target detection model and the sample image, obtain at least two fourth 2D detection frames corresponding to the sample vehicle, each of which is a 2D detection frame corresponding to the entire vehicle of the sample vehicle; match the third 2D detection frame with the at least two fourth 2D detection frames to obtain a target 2D detection frame that matches the third 2D detection frame;

[0171] The above-mentioned model training module is specifically used to determine the real vehicle orientation and the real offset of the sample vehicle based on the first 2D detection frame, the second 2D detection frame and the third 2D detection frame corresponding to the sample vehicle:

[0172] Based on the first 2D detection frame, the second 2D detection frame and the target 2D detection frame corresponding to the sample vehicle, a real vehicle orientation and a real offset of the sample vehicle are determined.

[0173] The vehicle chassis position positioning device of the embodiment of the present invention can execute the vehicle chassis position positioning method provided by the embodiment of the present invention, and the implementation principle is similar. The actions performed by each module and unit in the vehicle chassis position positioning device in each embodiment of the present invention correspond to the steps in the vehicle chassis position positioning method in each embodiment of the present invention. For the detailed functional description of each module of the vehicle chassis position positioning device, please refer to the description in the corresponding vehicle chassis position positioning method shown in the previous text, which will not be repeated here.

[0174] Among them, the above-mentioned vehicle chassis position positioning device can be a computer program (including program code) running in a computer device, for example, the vehicle chassis position positioning device is an application software; the device can be used to execute the corresponding steps in the method provided in the embodiment of the present invention.

[0175] In some embodiments, the vehicle chassis position positioning device provided in the embodiment of the present invention can be implemented by a combination of software and hardware. As an example, the vehicle chassis position positioning device provided in the embodiment of the present invention can be a processor in the form of a hardware decoding processor, which is programmed to execute the vehicle chassis position positioning method provided in the embodiment of the present invention. For example, the processor in the form of a hardware decoding processor can adopt one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field programmable gate arrays (FPGAs) or other electronic components.

[0176] In other embodiments, the vehicle chassis position positioning device provided by the embodiments of the present invention can be implemented in a software manner. Figure 6 A vehicle chassis position positioning device stored in a memory is shown, which can be software in the form of a program and a plug-in, and includes a series of modules, including an acquisition module 210, a processing module 220 and a positioning module 230, for implementing the vehicle chassis position positioning method provided in an embodiment of the present invention.

[0177] The modules involved in the embodiments of the present invention may be implemented by software or hardware, wherein the name of a module does not limit the module itself in some cases.

[0178] Based on the same principle as the method shown in the embodiments of the present invention, an electronic device is also provided in the embodiments of the present invention, which may include but is not limited to: a processor and a memory; the memory is used to store a computer program; the processor is used to execute the method shown in any embodiment of the present invention by calling the computer program.

[0179] In an alternative embodiment, an electronic device is provided, such as Figure 7 As shown, Figure 7The electronic device 4000 shown includes: a processor 4001 and a memory 4003. The processor 4001 and the memory 4003 are connected, such as through a bus 4002. Optionally, the electronic device 4000 may also include a transceiver 4004, which may be used for data interaction between the electronic device and other electronic devices, such as data transmission and / or data reception. It should be noted that in actual applications, the transceiver 4004 is not limited to one, and the structure of the electronic device 4000 does not constitute a limitation on the embodiments of the present invention.

[0180] Processor 4001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It may implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of the present invention. Processor 4001 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0181] The bus 4002 may include a path to transmit information between the above components. The bus 4002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus 4002 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 7 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0182] The memory 4003 can be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compressed optical disk, laser disk, optical disk, digital versatile disk, Blu-ray disk, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited to this.

[0183] The memory 4003 is used to store the application code (computer program) for executing the solution of the present invention, and the execution is controlled by the processor 4001. The processor 4001 is used to execute the application code stored in the memory 4003 to implement the content shown in the above method embodiment.

[0184] The electronic device may also be a terminal device. Figure 7 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0185] An embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer-readable storage medium is run on a computer, the computer can execute the corresponding contents of the aforementioned method embodiment.

[0186] According to another aspect of the present invention, a computer program product or a computer program is also 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 executes the methods provided in the implementation modes of the above-mentioned various embodiments.

[0187] Computer program code for performing the operations of the present invention may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0188] It should be understood that the flow charts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the module, the program segment, or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a sequence different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0189] The computer-readable storage medium provided by the embodiment of the present invention may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to, an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, device, or device.

[0190] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device executes the method shown in the above embodiment.

[0191] The above description is only a preferred embodiment of the present invention and an explanation of the technical principles used. Those skilled in the art should understand that the disclosure scope involved in the present invention is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosed concept. For example, the above features are replaced with the technical features with similar functions disclosed in the present invention (but not limited to) by each other.

Claims

1. A method for locating a vehicle chassis, characterized in that: The following steps are involved: Obtain a first 2D detection frame and a second 2D detection frame corresponding to the vehicle to be detected, wherein the first 2D detection frame is a 2D detection frame corresponding to the front of the vehicle to be detected perpendicular to the ground, and the second 2D detection frame is a 2D detection frame corresponding to the rear of the vehicle to be detected perpendicular to the ground; Based on the first 2D detection frame and the second 2D detection frame, determine the target orientation and target offset of the vehicle to be detected, the target offset includes an x-axis offset and a y-axis offset, the target offset is the side length of the 2D detection frame, and the target orientation represents the relative position of the first 2D detection frame and the second 2D detection frame of the vehicle to be detected; Based on the target orientation and the target offset, a bottom quadrilateral detection frame corresponding to the vehicle to be detected is determined, so as to determine the chassis position of the vehicle to be detected according to the bottom quadrilateral detection frame.

2. The method according to claim 1, characterized in that: The step of obtaining a first 2D detection frame and a second 2D detection frame corresponding to the vehicle to be detected includes: Acquire an image corresponding to the vehicle to be detected; Based on the pre-trained YOLO object detection model and the image, the first 2D detection box and the second 2D detection box are obtained.

3. The method according to claim 1, characterized in that: The determining, based on the first 2D detection frame and the second 2D detection frame, a target orientation and a target offset of the vehicle to be detected includes: If the position of the first 2D detection frame is to the left of the position of the second 2D detection frame, determining that the target orientation of the vehicle to be detected is positive; If the position of the first 2D detection frame is to the right of the position of the second 2D detection frame, determining that the target orientation of the vehicle to be detected is negative; The side of the first 2D detection frame parallel to the ground is determined as the x-axis offset in the target offset, and the side of the second 2D detection frame perpendicular to the ground is determined as the y-axis offset in the target offset.

4. The method according to any one of claims 1 to 3, characterized in that The obtaining of the first 2D detection frame and the second 2D detection frame corresponding to the vehicle to be detected and determining the target orientation and the target offset of the vehicle to be detected based on the first 2D detection frame and the second 2D detection frame are determined based on a pre-trained detection model, and the detection model is trained based on the following method: Acquire training data, wherein the training data includes sample data of each sample vehicle, each of the sample data includes a first identifier and a second identifier corresponding to the sample vehicle, each of the first identifiers represents a real vehicle orientation of the corresponding sample vehicle, and each of the second identifiers represents a real offset of the corresponding sample vehicle; According to the training data, the initial model is trained to obtain a first prediction result and a second prediction result corresponding to each of the sample vehicles, each of the first prediction results is a predicted vehicle orientation of the corresponding sample vehicle, and each of the second prediction results is a predicted offset of the corresponding sample vehicle; Determining a total loss value of the initial model according to the first prediction result, the second prediction result, the first identifier and the second identifier of each of the sample vehicles; When the total loss value satisfies a preset training end condition, the initial model that satisfies the training end condition is determined as the detection model; when the total loss value does not meet the preset training end condition, the parameters of the initial model are adjusted to retrain the initial model based on the adjusted parameters until the total loss value meets the preset training end condition.

5. The method according to claim 4, characterized in that Determining the total loss value of the initial model according to the first prediction result, the second prediction result, the first identifier and the second identifier of each of the sample vehicles includes: Determining a first loss value according to the first prediction result and the first identifier of each of the sample vehicles; Determine a second loss value according to the second prediction result and the second identification of each of the sample vehicles; A total loss value of the initial model is determined according to the first loss value and the second loss value.

6. The method according to claim 4, characterized in that For each of the sample vehicles, the sample data of the sample vehicle is obtained based on the following method: Obtaining a 3D detection frame corresponding to the sample vehicle; Determine a third 2D detection frame corresponding to the sample vehicle from the 3D detection frame, where the third 2D detection frame is a 2D detection frame corresponding to the entire vehicle of the sample vehicle; Based on the 3D detection frame, determining a first 2D detection frame and a second 2D detection frame corresponding to the sample vehicle; Determine a real vehicle orientation and a real offset of the sample vehicle based on the first 2D detection frame, the second 2D detection frame, and the third 2D detection frame corresponding to the sample vehicle; Based on the actual vehicle orientation and the actual offset of the sample vehicle, sample data of the sample vehicle is determined.

7. The method according to claim 6, characterized in that For each of the sample vehicles, the method further includes: Acquire a sample image corresponding to the sample vehicle; Based on the pre-trained YOLO object detection model and the sample image, at least two fourth 2D detection frames corresponding to the sample vehicle are obtained, each of the fourth 2D detection frames being a 2D detection frame corresponding to the entire vehicle of the sample vehicle; Matching the third 2D detection frame with the at least two fourth 2D detection frames to obtain a target 2D detection frame that matches the third 2D detection frame; The determining the real vehicle orientation and the real offset of the sample vehicle based on the first 2D detection frame, the second 2D detection frame, and the third 2D detection frame corresponding to the sample vehicle includes: Based on the first 2D detection frame, the second 2D detection frame and the target 2D detection frame corresponding to the sample vehicle, a real vehicle orientation and a real offset of the sample vehicle are determined.

8. A vehicle chassis position positioning device, characterized in that: include: An acquisition module is used to acquire a first 2D detection frame and a second 2D detection frame corresponding to the vehicle to be detected, wherein the first 2D detection frame is a 2D detection frame corresponding to the front of the vehicle to be detected perpendicular to the ground, and the second 2D detection frame is a 2D detection frame corresponding to the rear of the vehicle to be detected perpendicular to the ground; A processing module, configured to determine a target orientation and a target offset of the vehicle to be detected based on the first 2D detection frame and the second 2D detection frame, wherein the target offset includes an x-axis offset and a y-axis offset, the target offset is a side length of the 2D detection frame, and the target orientation represents a relative position between the first 2D detection frame and the second 2D detection frame of the vehicle to be detected; A positioning module is used to determine a bottom quadrilateral detection frame corresponding to the vehicle to be detected based on the target orientation and the target offset, so as to determine the chassis position of the vehicle to be detected according to the bottom quadrilateral detection frame.

9. An electronic device, characterized in that: The method comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images