Coordinate determination method and device of non-rigid body target object, equipment and storage medium

Abstract

The invention provides a coordinate determination method and device of a non-rigid body target object, equipment and a storage medium. The method for determining the coordinates of the non-rigid target object comprises the following steps: acquiring an image of the non-rigid target object; inputting the image of the non-rigid target object into a pre-trained non-rigid recognition model to obtain first non-rigid information output by the non-rigid recognition model; performing 3D calculation according to the first non-rigid body information to obtain third coordinate information of a 3D target frame of the first target body in the real coordinate system and fourth coordinate information of a 3D target frame of the second target body in the real coordinate system; and performing coordinate correction on the third coordinate information and the fourth coordinate information to obtain target coordinate information of the non-rigid target object. According to the method, the coordinates of the non-rigid target object can be obtained by performing identification, 3D calculation and coordinate correction processing on the image, the coordinates of the non-rigid target object can be determined without arranging a sensor, and the cost of identifying the coordinates of the non-rigid target object is reduced.

Description

Technical Field

[0001] The present application relates to the field of target detection technology, and in particular to a method, device, equipment and storage medium for determining the coordinates of a non-rigid target. Background Art

[0002] A non-rigid body refers to an object whose phase and position relationships between its internal parts change during physical motion. For example, a non-rigid vehicle, such as a trailer truck, produces various special postures during turning, which causes the relative positions of the various parts of the vehicle to change.

[0003] At present, it is necessary to install sensors on non-rigid targets to obtain the positions of various internal parts. This method relies on sensors such as lidar to obtain depth information, which increases hardware costs, is expensive, and cannot obtain the boundary positions of non-rigid targets. Summary of the invention

[0004] In view of this, the purpose of the present application is to provide a method, device, equipment and storage medium for determining the coordinates of a non-rigid target object, which can accurately determine the coordinates of the non-rigid target object and reduce the coordinate recognition cost.

[0005] To achieve the above objectives, the present application discloses the following technical solutions:

[0006] In a first aspect, the present application provides a method for determining the coordinates of a non-rigid target object, comprising:

[0007] Acquire an image of a non-rigid target object, wherein the non-rigid target object includes a first target object and at least one second target object;

[0008] Inputting the image of the non-rigid target object into a pre-trained non-rigid recognition model to obtain first non-rigid information output by the non-rigid recognition model; the first non-rigid information includes: first coordinate information of a 2D target frame of the first target object, second coordinate information of a 2D target frame of the second target object, and a first posture category of the non-rigid target object;

[0009] Performing 3D solution according to the first non-rigid body information to obtain third coordinate information of the 3D target frame of the first target in the real coordinate system and fourth coordinate information of the 3D target frame of the second target in the real coordinate system;

[0010] Coordinate correction is performed on the third coordinate information and the fourth coordinate information to obtain target coordinate information of the non-rigid target object.

[0011] Optionally, the non-rigid body recognition model is trained in the following manner:

[0012] Acquire training data, wherein the training data includes a training target image and a label corresponding to the training target image;

[0013] Inputting the training non-rigid image into the model to be trained, obtaining second non-rigid information output by the model to be trained, wherein the second non-rigid information includes fifth coordinate information of a 2D target frame of a training target body in the training non-rigid image and a second posture category;

[0014] The model to be trained is trained using the second non-rigid body information and the label until a loss value is less than a threshold, thereby obtaining a non-rigid body recognition model, wherein the loss value is used to measure the gap between the second non-rigid body information and the label.

[0015] Optionally, if the non-rigid target is in a front-view posture and the non-rigid target is a preset target, the method further includes: acquiring a first preset length of a first target and a second preset length of a second target in the non-rigid target;

[0016] The 3D solution process includes:

[0017] The first coordinate information and the second coordinate information are converted into a real coordinate system according to the camera model principle to obtain a first width coordinate of the first target body in the real coordinate system and a second width coordinate of the second target body in the real coordinate system;

[0018] Determine a first ratio of width to height in the first coordinate information according to the first coordinate information; determine a second ratio of width to height in the second coordinate information according to the second coordinate information;

[0019] Determine a first height coordinate of the first target body in a real coordinate system according to the first ratio and the first width coordinate; determine a second height coordinate of the second target body in a real coordinate system according to the second ratio and the second width coordinate;

[0020] The third coordinate information is determined according to the first width coordinate, the first height coordinate, and the first preset length; and the fourth coordinate information is determined according to the second width coordinate, the second height coordinate, and the second preset length.

[0021] Optionally, if the non-rigid target object is in a squint posture and the non-rigid target object is a preset target object, the non-rigid target information further includes: at least two first key point coordinates in a 2D target frame of the first target object and at least two second key point coordinates in a 2D target frame of the second target object;

[0022] The method further includes: obtaining a first preset width of a first target body and a second preset width of a second target body in the non-rigid target object;

[0023] The solution process includes:

[0024] Determine the first straight line coordinates where the first key point is located according to at least two coordinates of the first key point; determine the second straight line coordinates where the second key point is located according to at least two coordinates of the second key point;

[0025] Determine a first length coordinate of the first target body in a real coordinate system according to the first straight line coordinate and the first coordinate information; determine a second length coordinate of the second target body in a real coordinate system according to the second straight line and the second coordinate information;

[0026] Determine a third width coordinate of the first target body in the real coordinate system according to the first length coordinate and the first preset width; determine a fourth width coordinate of the second target body in the real coordinate system according to the second length coordinate and the second preset width;

[0027] Determine a first ratio of width to height in the first coordinate information according to the first coordinate information; determine a second ratio of width to height in the second coordinate information according to the second coordinate information;

[0028] Determine a third height coordinate of the first target body in the real coordinate system according to the first ratio and the third width coordinate; determine a fourth height coordinate of the second target body in the real coordinate system according to the second ratio and the fourth width coordinate;

[0029] The third coordinate information is determined according to the first length coordinate, the third width coordinate, and the third height coordinate; and the fourth coordinate information is determined according to the second length coordinate, the fourth width coordinate, and the fourth height coordinate.

[0030] Optionally, the coordinate correction includes height correction, and the coordinate correction of the third coordinate information and the fourth coordinate information includes:

[0031] Determine that the height information of the third coordinate information and / or the height information in the fourth coordinate information is greater than a preset height threshold;

[0032] The height information of the third coordinate information and / or the height information in the fourth coordinate information is adjusted to the preset height threshold to complete the height correction.

[0033] Optionally, the coordinate correction includes position correction, and the coordinate correction of the third coordinate information and the fourth coordinate information includes:

[0034] Determine the target traction position according to the third coordinate information or the fourth coordinate information;

[0035] Determine the intersection of the center line of the first target body and the center line of the second target body according to the third coordinate information and the fourth coordinate information;

[0036] Determine the origin position in the real coordinate system;

[0037] It is determined that the distance between the intersection point and the target traction position is greater than a preset distance threshold, and coordinate correction is performed on the third coordinate information or the fourth coordinate information based on the origin position and the intersection point.

[0038] Optionally, performing coordinate correction on the third coordinate information or the fourth coordinate information based on the origin position and the intersection point includes:

[0039] Determine a first distance according to the origin position and the third coordinate information; determine a second distance according to the origin position and the fourth coordinate information;

[0040] If the first distance is greater than or equal to the second distance, position correction is performed on the third coordinate information so that the target traction position is corrected;

[0041] If the first distance is smaller than the second distance, the fourth coordinate information is position-corrected so that the target traction position after position correction is the same as the intersection point.

[0042] In a second aspect, the present application provides a device for determining the coordinates of a non-rigid target, the device comprising:

[0043] An acquisition unit, configured to acquire an image of a non-rigid target object, wherein the non-rigid target object includes a first target object and a second target object;

[0044] A model output unit is used to input the image of the non-rigid target object into a pre-trained non-rigid recognition model to obtain first non-rigid information output by the non-rigid recognition model; the first non-rigid information includes: first coordinate information of a 2D target frame of the first target object, second coordinate information of the 2D target frame of the second target object, and a posture category of the non-rigid target object;

[0045] A 3D solving unit, used for performing 3D solving according to the first non-rigid body information, to obtain third coordinate information of the 3D target frame of the first target body in the real coordinate system and fourth coordinate information of the 3D target frame of the second target body in the real coordinate system;

[0046] The coordinate correction unit is used to perform coordinate correction on the third coordinate information and the fourth coordinate information to obtain the target coordinate information of the non-rigid target object.

[0047] Optionally, the non-rigid body recognition model is trained in the following manner:

[0048] Acquire training data, wherein the training data includes a training target image and a label corresponding to the training target image;

[0049] Inputting the training non-rigid image into the model to be trained, obtaining second non-rigid information output by the model to be trained, wherein the second non-rigid information includes fifth coordinate information of a 2D target frame of a training target body in the training non-rigid image and a second posture category;

[0050] The model to be trained is trained using the second non-rigid body information and the label until a loss value is less than a threshold, thereby obtaining a non-rigid body recognition model, wherein the loss value is used to measure the gap between the second non-rigid body information and the label.

[0051] Optionally, if the non-rigid target is in a front-view posture and the non-rigid target is a preset target, the acquisition unit is further used to: acquire a first preset length of a first target and a second preset length of a second target in the non-rigid target;

[0052] Then the 3D solving unit includes:

[0053] A conversion subunit, used to convert the first coordinate information and the second coordinate information into a real coordinate system according to a camera model principle, to obtain a first width coordinate of the first target body in the non-rigid target object in the real coordinate system, and a second width coordinate of the second target body in the real coordinate system;

[0054] a determination subunit, configured to determine a first ratio of width to height in the first coordinate information according to the first coordinate information; and to determine a second ratio of width to height in the second coordinate information according to the second coordinate information;

[0055] The determination subunit is further used to determine a first height coordinate of the first target body in the real coordinate system according to the first ratio and the first width coordinate; and to determine a second height coordinate of the second target body in the real coordinate system according to the second ratio and the second width coordinate;

[0056] The determination subunit is further used to determine the third coordinate information according to the first width coordinate, the first height coordinate and the first preset length; and to determine the fourth coordinate information according to the second width coordinate, the second height coordinate and the second preset length.

[0057] Optionally, if the non-rigid target object is in a squint posture and the non-rigid target object is a preset target object, the non-rigid target information further includes: at least two first key point coordinates in a 2D target frame of the first target object and at least two second key point coordinates in a 2D target frame of the second target object;

[0058] The acquisition unit is further used to acquire a first preset width of a first target body and a second preset width of a second target body in the non-rigid target object;

[0059] Then the 3D solving unit includes:

[0060] a determination subunit, configured to determine the first straight line coordinates where the first key point is located according to at least two coordinates of the first key point; and to determine the second straight line coordinates where the second key point is located according to at least two coordinates of the second key point;

[0061] The determination subunit is further used to determine a first length coordinate of the first target body in the real coordinate system according to the first straight line coordinate and the first coordinate information; and to determine a second length coordinate of the second target body in the real coordinate system according to the second straight line and the second coordinate information;

[0062] The determination subunit is further used to determine a third width coordinate of the first target body in the real coordinate system according to the first length coordinate and the first preset width; and to determine a fourth width coordinate of the second target body in the real coordinate system according to the second length coordinate and the second preset width;

[0063] The determination subunit is further used to determine a first ratio of width to height in the first coordinate information according to the first coordinate information; and to determine a second ratio of width to height in the second coordinate information according to the second coordinate information;

[0064] The determination subunit is further used to determine a third height coordinate of the first target body in the real coordinate system according to the first ratio and the third width coordinate; and to determine a fourth height coordinate of the second target body in the real coordinate system according to the second ratio and the fourth width coordinate;

[0065] The determination subunit is further used to determine the third coordinate information according to the first length coordinate, the third width coordinate and the third height coordinate; and to determine the fourth coordinate information according to the second length coordinate, the fourth width coordinate and the fourth height coordinate.

[0066] Optionally, the coordinate correction includes height correction, and the coordinate correction unit includes:

[0067] A determination subunit, configured to determine that the height information of the third coordinate information and / or the height information in the fourth coordinate information is greater than a preset height threshold;

[0068] The adjustment subunit is used to adjust the height information of the third coordinate information and / or the height information in the fourth coordinate information to the preset height threshold to complete the height correction.

[0069] Optionally, the coordinate correction includes position correction, and the coordinate correction unit includes:

[0070] a determination subunit, configured to determine a target traction position according to the third coordinate information or the fourth coordinate information;

[0071] The determination subunit is further used to determine the intersection of the center line of the first target body and the center line of the second target body according to the third coordinate information and the fourth coordinate information;

[0072] Determine the subunit, and also determine the origin position in the real coordinate system;

[0073] The correction subunit is used to determine that the distance between the intersection point and the target traction position is greater than a preset distance threshold, and then perform coordinate correction on the third coordinate information or the fourth coordinate information based on the origin position and the intersection point.

[0074] Optionally, the correction subunit is specifically used to:

[0075] Determine a first distance according to the origin position and the third coordinate information; determine a second distance according to the origin position and the fourth coordinate information;

[0076] If the first distance is greater than or equal to the second distance, position correction is performed on the third coordinate information so that the target traction position is corrected;

[0077] If the first distance is smaller than the second distance, the fourth coordinate information is position-corrected so that the target traction position after position correction is the same as the intersection point.

[0078] In a third aspect, the present application provides a coordinate determination device for a non-rigid target object, comprising:

[0079] Memory for storing computer programs;

[0080] A processor is used to execute the computer program stored in the memory to implement the steps of the coordinate determination method of the non-rigid target object as provided in the first aspect.

[0081] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the steps of the method for determining the coordinates of the non-rigid target object as provided in the first aspect.

[0082] The present application provides a method, device, equipment and storage medium for determining the coordinates of a non-rigid target. In a method for determining the coordinates of a non-rigid target, an image of the non-rigid target is obtained; the image of the non-rigid target is input into a pre-trained non-rigid recognition model to obtain the first non-rigid information output by the non-rigid recognition model; 3D solution is performed based on the first non-rigid information to obtain the third coordinate information of the 3D target frame of the first target in the real coordinate system and the fourth coordinate information of the 3D target frame of the second target in the real coordinate system; the third coordinate information and the fourth coordinate information are coordinate corrected to obtain the target coordinate information of the non-rigid target. Through the above method, the coordinates of the non-rigid target can be obtained by image recognition, 3D solution and coordinate correction processing, and the coordinates of the non-rigid target can be determined without setting up a sensor, thereby reducing the cost of coordinate recognition of the non-rigid target. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application, and a person skilled in the art can also obtain other drawings based on these drawings.

[0084] Figure 1 It is a flow chart of a method for determining the coordinates of a non-rigid target object provided in an embodiment of the present application;

[0085] Figure 2 is a schematic diagram of a 2D target frame in a front-view posture provided in an embodiment of the present application;

[0086] Figure 3 is a schematic diagram of a 2D target frame in a squint posture provided in an embodiment of the present application;

[0087] Figure 4 is a schematic diagram of a 3D target frame provided in an embodiment of the present application;

[0088] Figure 5is a schematic diagram of a 2D target frame of a trailer of a vehicle provided in an embodiment of the present application;

[0089] Figure 6 is a schematic diagram of a position correction process provided by an embodiment of the present application;

[0090] Figure 7 It is a structural schematic diagram of a coordinate determination device for a non-rigid target object provided in an embodiment of the present application. DETAILED DESCRIPTION

[0091] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0092] To facilitate understanding of the technical solution provided by the present application, a method for determining the coordinates of a non-rigid target object provided by the present application will be described below in conjunction with the accompanying drawings. Figure 1 , which is a flow chart of a method for determining the coordinates of a non-rigid target provided in an embodiment of the present application, such as Figure 1 As shown, the method includes S101-S104.

[0093] S101: Acquire an image of a non-rigid target object, where the non-rigid target object includes a first target object and at least one second target object.

[0094] In the embodiment of the present application, an image of a non-rigid target object is first obtained. The present application does not limit the specific method of obtaining an image of a non-rigid target object. As an example, an image of a non-rigid target object can be obtained by a surveillance camera. In the present application, the first target object and the second target object are non-rigidly connected, and multiple second target objects can also be non-rigidly connected.

[0095] The present application does not limit the specific content of the non-rigid target object. As an example, the non-rigid target object may be a vehicle, and the first target body and the second target body may be two vehicle bodies of the vehicle, such as a tractor head and a trailer in a trailer truck. The trailer may be one or more trailers. It can be understood that the tractor head and the trailer are non-rigidly connected, and multiple trailers may also be non-rigidly connected.

[0096] S102: Inputting the image of the non-rigid target object into a pre-trained non-rigid recognition model to obtain first non-rigid information output by the non-rigid recognition model.

[0097] In the embodiment of the present application, after obtaining an image of a non-rigid target object, the image is input into a non-rigid recognition model to obtain first non-rigid information output by the non-rigid recognition model. The first non-rigid information includes: first coordinate information of a 2D target frame of a first target object, second coordinate information of a 2D target frame of a second target object, and a first posture category of the non-rigid target object.

[0098] It can be understood that the first posture category includes: front view posture and side view posture. The 2D target frame can display the area where the target object is located. In the front view posture, the 2D target frame can only display the width and height of the target object. In the side view posture, the 2D target frame can only display the height and length of the target object. The coordinate information of the 2D target frame is the coordinate information of the non-rigid target object in the image coordinate system.

[0099] Taking a non-rigid target as a trailer truck as an example, the shape of the 2D target box can be as follows Figure 2 and Figure 3 As shown, Figure 2 A schematic diagram of a 2D target frame in a front-view posture provided in an embodiment of the present application, wherein the solid-line frame is the above-mentioned 2D target frame. Figure 3 is a schematic diagram of a 2D target frame in a squint posture provided in an embodiment of the present application, wherein the solid line frame is the above-mentioned 2D target frame.

[0100] S103: Perform 3D calculation according to the first non-rigid body information to obtain third coordinate information of the 3D target frame of the first target in the real coordinate system and fourth coordinate information of the 3D target frame of the second target in the real coordinate system.

[0101] In the embodiment of the present application, after obtaining the first non-rigid body information, 3D solution can be performed to obtain the third coordinate information of the 3D target frame of the first target body in the real coordinate system and the fourth coordinate information of the 3D target frame of the second target in the real coordinate system. It can be understood that the real coordinate system can be a world coordinate system or a vehicle coordinate system defined with one wheel of the vehicle itself as the origin. The third coordinate information includes the length, width and height of the first target body in the real coordinate system and the specific positions of each vertex of the boundary, and the fourth coordinate information includes the length, width and height of the second target body in the real coordinate system and the specific positions of each vertex of the boundary.

[0102] Taking a non-rigid target as a trailer truck as an example, the shape of the 3D target box is as follows Figure 4 As shown, Figure 4 A schematic diagram of a 3D target frame provided in an embodiment of the present application, wherein the solid line frame is the above-mentioned 3D target frame

[0103] The present application does not limit the specific steps of performing 3D solution based on the first non-rigid body information, which will be described in detail below.

[0104] S104: performing coordinate correction on the third coordinate information and the fourth coordinate information to obtain target coordinate information of the non-rigid target object.

[0105] In the embodiment of the present application, after obtaining the third coordinate information and the fourth coordinate information, the third coordinate information and the fourth coordinate information are coordinate-corrected to reduce the error of the third coordinate information and the fourth coordinate information. The present application does not limit the specific manner of coordinate-correcting the third coordinate information and the fourth coordinate information, which is described below.

[0106] Through the above method, the present application can obtain the image of the non-rigid target, and then perform a series of recognition, 3D solution, and coordinate correction processing based on the image to obtain the coordinates of the non-rigid target. Compared with the prior art that requires the use of sensors to obtain the coordinates of the non-rigid target, the present application does not need to set up additional sensors, thereby reducing the cost of coordinate recognition of the non-rigid target. In addition, the boundary information of the non-rigid target can be obtained.

[0107] As a possible implementation manner, after obtaining the target coordinate information of the non-rigid target object, the method of the present application further includes: converting the 3D target frame containing the target coordinate information into the image according to the camera model principle.

[0108] It is understandable that the present application will also convert the 3D target frame of the target coordinate information into the image coordinate system for visual display, which is helpful for the user to directly determine the coordinates of the non-rigid target object from the image.

[0109] As a possible implementation method, the non-rigid body recognition model is trained in the following way. The training steps include:

[0110] A1: Acquire training data, where the training data includes training target images and labels corresponding to the training target images.

[0111] In the embodiments of the present application, the method for acquiring the training target object image is not limited. The present application can collect the training target object image by crawler technology. The training target object image can also be collected by a data acquisition system. As an example, when the training target object image to be collected is an image of a trailer truck, the present application can use the forward-looking camera installed on the experimental vehicle to collect a large number of pictures of the trailer truck.

[0112] In the embodiment of the present application, the training target image is first obtained, and then the label corresponding to the training target image is obtained by annotating it through the labelme image data annotation tool. It can be understood that the label is used to identify the 2D target frame and posture category in the training target image. The label can also include at least 2 annotated key point coordinates in the 2D target frame. When the training target image is a trailer truck image, the annotated key point coordinates in the label can be the coordinates of the two farthest same-side wheel contact points.

[0113] This application does not limit the specific method of marking, and the marking method requires that the first target body and the second target body in the non-rigid target object be marked separately. For example, for a trailer truck, it will be marked separately according to the tractor and the trailer. If a trailer truck has 1 tractor head and 1 trailer, the trailer truck is regarded as two vehicles and marked separately in the way that the head is 1 vehicle and the trailer is 1 vehicle. If a truck has 1 head and 2 trailers, the truck is regarded as three vehicles and marked separately in the way that the head is 1 vehicle and the trailer is divided into 2 vehicles. For the part of the vehicle that is blocked in the picture, the occlusion completion operation can also be performed.

[0114] A2: Input the training non-rigid image into the model to be trained to obtain the second non-rigid information output by the model to be trained, where the second non-rigid information includes the fifth coordinate information of the 2D target box of the training target body in the training non-rigid image and the second posture category.

[0115] In the embodiment of the present application, the training non-rigid image is input into the model to be trained, and the second non-rigid information output by the model to be trained can be obtained. The second non-rigid information includes the fifth coordinate information of the 2D target box of the training target body in the training non-rigid image and the second posture category. It can be understood that the second posture category includes a normal view posture and a squinting posture.

[0116] As an example, the second non-rigid body information also includes key point coordinates in a 2D target box of the training target body.

[0117] A3: The model to be trained is trained using the second non-rigid body information and the label until the loss value is less than a threshold value, thereby obtaining a non-rigid body recognition model, wherein the loss value is used to measure the gap between the second non-rigid body information and the label.

[0118] After obtaining the second non-rigid information, the training non-rigid image and the label corresponding to the training non-rigid image are used to train the model to be trained until the loss value is less than the threshold. It is understandable that the network structure and parameters of the model will be continuously adjusted during the model training process. When the difference between the second non-rigid information output by the model and the label is less than the threshold, the training ends and the network structure and parameters of the model are fixed. According to the fixed network structure and parameters of the model, the trained non-rigid recognition model can be obtained.

[0119] As a possible implementation manner, if the non-rigid target is in a front view posture and the non-rigid target is a preset target, the method further includes: obtaining a first preset length of a first target and a second preset length of a second target in the non-rigid target.

[0120] It is understandable that in the embodiment of the present application, when the non-rigid target is a preset target, the first preset length of the first target and the second preset length of the second target can be directly obtained. For example, when the method of the present application is applied to trailer trucks in a port scenario, the trailer trucks are all standard, and the lengths of their tractor heads and trailers are known.

[0121] The 3D solution process includes the following steps:

[0122] B1: According to the camera model principle, the first coordinate information and the second coordinate information are converted into the real coordinate system to obtain the first width coordinate of the first target body in the non-rigid target object in the real coordinate system and the second width coordinate of the second target body in the real coordinate system.

[0123] In the embodiment of the present application, according to the principle of the camera model, only the width coordinate is converted to the real coordinate system to obtain the first width coordinate of the first target body in the non-rigid target object in the real coordinate system and the second width coordinate of the second target body in the real coordinate system.

[0124] Specifically, the coordinates of the two endpoints of the bottom line segment of the 2D target box can be converted from the image coordinate system to the real coordinate system based on the camera model principle, and the two coordinate points of the width line segment of the non-rigid target object in the real world can be obtained, and then the real width of the non-rigid target object can be calculated.

[0125] B2: Determine a first ratio of width to height in the first coordinate information according to the first coordinate information; and determine a second ratio of width to height in the second coordinate information according to the second coordinate information.

[0126] In the embodiment of the present application, a first ratio of width to height in the first coordinate information is determined in the image coordinate system, and a second ratio of width to height in the second coordinate information is determined.

[0127] B3: Determine a first height coordinate of the first target body in the real coordinate system according to the first ratio and the first width coordinate; determine a second height coordinate of the second target body in the real coordinate system according to the second ratio and the second width coordinate.

[0128] B4: Determine the third coordinate information according to the first width coordinate, the first height coordinate and the first preset length; determine the fourth coordinate information according to the second width coordinate, the second height coordinate and the second preset length.

[0129] In the embodiment of the present application, after determining the ratio of width to height and a width coordinate, the height coordinate can be determined, and then the third coordinate information of the first target body and the fourth coordinate information of the second target body can be obtained by combining the width coordinate, the height coordinate and the preset length. It can be understood that the third coordinate information can represent the length, width and height of the first target body in the real coordinate system.

[0130] The fourth coordinate information may represent the length, width, and height of the second target body in the real coordinate system.

[0131] Through the above method, the real length of the preset target object can be combined to perform 3D solution, so as to more accurately determine the coordinates of the non-rigid target object.

[0132] As a possible implementation method, the present application can also first determine the height coordinate of the target body through the camera model principle, and then determine the width coordinate based on the ratio of width to height, and continue to obtain the third coordinate information of the first target body and the fourth coordinate information of the second target body according to steps B3-B4.

[0133] As a possible implementation method, if the non-rigid target object is in a squinting posture and the non-rigid target object is a preset target object, the non-rigid target information also includes: at least 2 first key point coordinates in the 2D target frame of the first target body and at least 2 second key point coordinates in the 2D target frame of the second target body.

[0134] The present application provides a method for determining the coordinates of a non-rigid target object, further comprising: obtaining a first preset width of a first target object and a second preset width of a second target object in the non-rigid target object.

[0135] It is understandable that in the embodiment of the present application, when the non-rigid target is a preset target, the first preset width of the first target and the second preset width of the second target can be directly obtained. For example, when the method of the present application is applied to trailer trucks in a port scene, the trailer trucks are all standard, and the widths of their tractor head and trailer are both known real widths.

[0136] The solution process includes:

[0137] C1: Determine the first straight line coordinates where the first key point is located based on at least two first key point coordinates; determine the second straight line coordinates where the second key point is located based on at least two second key point coordinates.

[0138] C2: Determine the first length coordinate of the first target body in the real coordinate system based on the first straight line coordinate and the first coordinate information; determine the second length coordinate of the second target body in the real coordinate system based on the second straight line and the second coordinate information.

[0139] In the embodiment of the present application, the coordinates of the key points and the coordinate information of the 2D target frame are combined to determine the length of the target body in the real coordinate system.

[0140] This application takes the 3D solution of a trailer truck image as an example, where the key points are the two front and rear wheel contact points. Figure 5 As shown, Figure 5 A schematic diagram of a 2D target frame of a trailer of a vehicle provided in an embodiment of the present application. The present application will determine the dotted line L1 based on the two front and rear wheel contact points, and determine the length based on the coordinate information of the 2D target frame and the coordinate information of the dotted line L1. Specifically, the straight line L2 on the right side of the 2D target frame can be intersected downward with the dotted line L1 at point A, and the straight line L3 on the left side of the 2D target frame can be intersected with the dotted line L1 at point B, thereby determining the line segment AB, which is the line segment of the vehicle length. By determining the coordinates of point A and point B in the image coordinate system, and converting them to the real coordinate system according to the camera model principle, the length coordinates of the trailer can be obtained.

[0141] C3: Determine the third width coordinate of the first target body in the real coordinate system according to the first length coordinate and the first preset width; determine the fourth width coordinate of the second target body in the real coordinate system according to the second length coordinate and the second preset width.

[0142] In the embodiment of the present application, after the first length coordinate and the second length coordinate are obtained, the third width coordinate and the fourth width coordinate can be determined according to the first preset width and the second preset width respectively.

[0143] The present application does not limit the specific process of the third width coordinate and the fourth width coordinate. The following also takes the 3D solution of the trailer truck image as an example, where the key points are the two front and rear wheel contact points.

[0144] After obtaining the coordinates of point A and point B in the image coordinate system as mentioned above, convert point A and point B to the real coordinate system, and the coordinates of point A1 and point B1 can be obtained respectively. Draw a perpendicular line on the ground plane in the real coordinate system that intersects A1B1 and point B1, and take a line segment with a size of preset width. Then there will be 2 points on the perpendicular line whose distance from B1 is the preset width. Let these two points be W1 and W2 respectively. In this application, it is assumed that the origin of the real coordinate system is 0, then the distances |OW1| and |OW2| from W1 and W2 to the origin O are calculated respectively. If |OW1| is greater than |OW2|, the other endpoint of the width line segment is W1, otherwise it is W2. According to the rules of rectangles, after obtaining the long side and short side of the rectangle, the four sides of the rectangle can be obtained. The rectangle corresponds to the four bottom sides of the 3D target box of the target body. That is, the third width coordinate and the fourth width coordinate are obtained.

[0145] C4: determining a first ratio of width to height in the first coordinate information according to the first coordinate information; and determining a second ratio of width to height in the second coordinate information according to the second coordinate information.

[0146] C5: Determine the third height coordinate of the first target body in the real coordinate system according to the first ratio and the third width coordinate; determine the fourth height coordinate of the second target body in the real coordinate system according to the second ratio and the fourth width coordinate.

[0147] C6: determining the third coordinate information according to the first length coordinate, the third width coordinate, and the third height coordinate; determining the fourth coordinate information according to the second length coordinate, the fourth width coordinate, and the fourth height coordinate.

[0148] In an embodiment of the present application, after obtaining the third width coordinate and the fourth width coordinate, the height coordinate can be determined by combining the width-to-height ratio and a width coordinate, and then by combining the width coordinate, the height coordinate and the preset length, the third coordinate information of the first target body and the fourth coordinate information of the second target body can be obtained.

[0149] Through the above method, the real width of the preset target object can be combined to perform 3D solution, so as to more accurately determine the coordinates of the non-rigid target object.

[0150] As a possible implementation manner, the coordinate correction includes height correction, and S104 performs coordinate correction on the third coordinate information and the fourth coordinate information, including:

[0151] Determine that the height information of the third coordinate information and / or the height information in the fourth coordinate information is greater than a preset height threshold; adjust the height information of the third coordinate information and / or the height information in the fourth coordinate information to the preset height threshold to complete the height correction.

[0152] It is understandable that the present application can set a preset height threshold to determine whether the height information of the third coordinate information and / or the height information in the fourth coordinate information is greater than the preset height threshold, and if the height information of the third coordinate information and / or the height information in the fourth coordinate information is greater than the preset height threshold, the height information of the third coordinate information and / or the height information in the fourth coordinate information is adjusted to the preset height threshold.

[0153] The present application does not limit the specific value of the preset height threshold. As an example, the preset height threshold can be the standard height of the preset target plus 0.5 meters. By the above method, the height information in the third coordinate information and the height in the fourth coordinate information are corrected, further improving the accuracy of the coordinate information of the non-rigid target.

[0154] As a possible implementation manner, the coordinate correction includes position correction, and S104 performs coordinate correction on the third coordinate information and the fourth coordinate information, including the following steps:

[0155] D1: Determine the target traction position according to the third coordinate information or the fourth coordinate information.

[0156] D2: Determine the intersection point of the center line of the first target body and the center line of the second target body according to the third coordinate information and the fourth coordinate information.

[0157] It is understandable that since the first target body and the second target body are predicted as independent target bodies, position errors will occur, resulting in a large separation between the first target body and the second target body after the 3D solution is completed. Therefore, the present application requires position correction of the positions of the first target body and the second target body.

[0158] The first target body and the second target body are connected based on the target traction position, the first target body and the second target body are overlapped, and the center lines of the first target body and the second target body need to intersect at the target traction position. The present application can determine the target traction position based on the third coordinate information or the fourth coordinate information. It is understandable that the relative position of the target traction position in the first target body or the second target body is fixed.

[0159] D3: Determine the origin position in the real coordinate system.

[0160] D4: Determine that the distance between the intersection point and the target traction position is greater than a preset distance threshold, and perform coordinate correction on the third coordinate information or the fourth coordinate information based on the origin position and the intersection point.

[0161] It is understandable that the present application will calculate the distance between the intersection of the center lines of the first target body and the second target body and the target traction position. If it is determined that the distance is greater than the preset distance threshold, it means that the two do not overlap, which means that there is a deviation in the relative position between the first target body and the second target body, and coordinate correction is required. In addition, the present application performs coordinate correction on the third coordinate information or the fourth coordinate information based on the origin position and the intersection. The specific method of performing coordinate correction on the third coordinate information or the fourth coordinate information based on the origin position and the intersection includes the following steps:

[0162] E1: Determine a first distance according to the origin position and the third coordinate information; determine a second distance according to the origin position and the fourth coordinate information.

[0163] E2: If the first distance is greater than or equal to the second distance, position correction is performed on the third coordinate information so that the target traction position after position correction is the same as the intersection point.

[0164] E3: If the first distance is smaller than the second distance, position correction is performed on the fourth coordinate information so that the target traction position after position correction is the same as the intersection point.

[0165] It is understandable that the present application can use the camera that captures the image of the non-rigid target object as the origin position in the real coordinate system. The closer the target object is to the camera, the more accurate the result obtained by the recognition and solution.

[0166] The present application determines the first distance based on the origin position and the third coordinate information, and determines the second distance based on the origin position and the fourth coordinate information, that is, the present application calculates the first distance between the origin position and the first target body, and the second distance between the origin position and the second target body.

[0167] If the first distance is greater than or equal to the second distance, it means that the first target is farther from the camera or the distance is the same, then the position of the second target is fixed and the position of the first target is adjusted. If the first distance is less than the second distance, it means that the first target is closer to the camera, then the position of the first target is fixed and the position of the second target is adjusted. And after the adjustment, the intersection of the center lines of the first target and the second target is the same as the target traction position.

[0168] The following uses the trailer truck image as an example to illustrate the position correction process. Figure 6 , Figure 6 A schematic diagram of a position correction process provided by an embodiment of the present application. Figure 6As shown, the tractor head is the first target body mentioned above, and the truck trailer is the second target body mentioned above. Point P is the traction pin, which is the target traction position recorded above, and point T is the intersection of the center line of the tractor head and the center line of the trailer after solution, which is the intersection of the center line of the first target body and the center line of the second target body mentioned above. The embodiment of the present application assumes that the position of the truck trailer is corrected so that the intersection of the center line of the tractor head and the center line of the trailer after correction coincides with the position of the traction pin.

[0169] By using the above method, the position information in the third coordinate information or the position in the fourth coordinate information is corrected, thereby further improving the accuracy of the coordinate information of the non-rigid target object.

[0170] See also Figure 7 , Figure 7 A schematic diagram of the structure of a coordinate determination device for a non-rigid target object provided in an embodiment of the present application, the device includes: an acquisition unit 701, a model output unit 702, a 3D solution unit 703 and a coordinate correction unit 704.

[0171] An acquisition unit 701 is used to acquire an image of a non-rigid target object, wherein the non-rigid target object includes a first target object and a second target object;

[0172] The model output unit 702 is used to input the image of the non-rigid target object into a pre-trained non-rigid recognition model to obtain first non-rigid information output by the non-rigid recognition model; the first non-rigid information includes: first coordinate information of a 2D target frame of the first target object, second coordinate information of a 2D target frame of the second target object, and a posture category of the non-rigid target object;

[0173] A 3D solving unit 703 is used to perform 3D solving according to the first non-rigid body information to obtain third coordinate information of the 3D target frame of the first target in the real coordinate system and fourth coordinate information of the 3D target frame of the second target in the real coordinate system;

[0174] The coordinate correction unit 704 is used to perform coordinate correction on the third coordinate information and the fourth coordinate information to obtain target coordinate information of the non-rigid target object.

[0175] Acquire an image of a non-rigid target object; input the image of the non-rigid target object into a pre-trained non-rigid recognition model to obtain the first non-rigid information output by the non-rigid recognition model; perform 3D solution based on the first non-rigid information to obtain the third coordinate information of the 3D target frame of the first target in the real coordinate system and the fourth coordinate information of the 3D target frame of the second target in the real coordinate system; perform coordinate correction on the third coordinate information and the fourth coordinate information to obtain the target coordinate information of the non-rigid target object. Through the above device, the coordinates of the non-rigid target object can be obtained by image recognition, 3D solution, and coordinate correction processing. The coordinates of the non-rigid target object can be determined without setting up a sensor, thereby reducing the cost of coordinate recognition of the non-rigid target object.

[0176] As a possible implementation method, the non-rigid body recognition model is trained in the following way:

[0177] Acquire training data, wherein the training data includes a training target image and a label corresponding to the training target image;

[0178] Inputting the training non-rigid image into the model to be trained, obtaining second non-rigid information output by the model to be trained, wherein the second non-rigid information includes fifth coordinate information of a 2D target frame of a training target body in the training non-rigid image and a second posture category;

[0179] The model to be trained is trained using the second non-rigid body information and the label until a loss value is less than a threshold, thereby obtaining a non-rigid body recognition model, wherein the loss value is used to measure the gap between the second non-rigid body information and the label.

[0180] As a possible implementation, if the non-rigid target is in a front-view posture and the non-rigid target is a preset target, the acquisition unit is further used to: acquire a first preset length of a first target and a second preset length of a second target in the non-rigid target;

[0181] Then the 3D solving unit includes:

[0182] A conversion subunit, used to convert the first coordinate information and the second coordinate information into a real coordinate system according to a camera model principle, to obtain a first width coordinate of the first target body in the non-rigid target object in the real coordinate system, and a second width coordinate of the second target body in the real coordinate system;

[0183] a determination subunit, configured to determine a first ratio of width to height in the first coordinate information according to the first coordinate information; and to determine a second ratio of width to height in the second coordinate information according to the second coordinate information;

[0184] The determination subunit is further used to determine a first height coordinate of the first target body in the real coordinate system according to the first ratio and the first width coordinate; and to determine a second height coordinate of the second target body in the real coordinate system according to the second ratio and the second width coordinate;

[0185] The determination subunit is further used to determine the third coordinate information according to the first width coordinate, the first height coordinate and the first preset length; and to determine the fourth coordinate information according to the second width coordinate, the second height coordinate and the second preset length.

[0186] As a possible implementation manner, if the non-rigid target object is in a squint posture and the non-rigid target object is a preset target object, the non-rigid target information further includes: at least two first key point coordinates in a 2D target frame of the first target object and at least two second key point coordinates in a 2D target frame of the second target object;

[0187] The acquisition unit is further used to acquire a first preset width of a first target body and a second preset width of a second target body in the non-rigid target object;

[0188] Then the 3D solving unit includes:

[0189] a determination subunit, configured to determine the first straight line coordinates where the first key point is located according to at least two coordinates of the first key point; and to determine the second straight line coordinates where the second key point is located according to at least two coordinates of the second key point;

[0190] The determination subunit is further used to determine a first length coordinate of the first target body in the real coordinate system according to the first straight line coordinate and the first coordinate information; and to determine a second length coordinate of the second target body in the real coordinate system according to the second straight line and the second coordinate information;

[0191] The determination subunit is further used to determine a third width coordinate of the first target body in the real coordinate system according to the first length coordinate and the first preset width; and to determine a fourth width coordinate of the second target body in the real coordinate system according to the second length coordinate and the second preset width;

[0192] The determination subunit is further used to determine a first ratio of width to height in the first coordinate information according to the first coordinate information; and to determine a second ratio of width to height in the second coordinate information according to the second coordinate information;

[0193] The determination subunit is further used to determine a third height coordinate of the first target body in the real coordinate system according to the first ratio and the third width coordinate; and to determine a fourth height coordinate of the second target body in the real coordinate system according to the second ratio and the fourth width coordinate;

[0194] The determination subunit is further used to determine the third coordinate information according to the first length coordinate, the third width coordinate and the third height coordinate; and to determine the fourth coordinate information according to the second length coordinate, the fourth width coordinate and the fourth height coordinate.

[0195] As a possible implementation manner, the coordinate correction includes height correction, and the coordinate correction unit includes:

[0196] A determination subunit, configured to determine that the height information of the third coordinate information and / or the height information in the fourth coordinate information is greater than a preset height threshold;

[0197] The adjustment subunit is used to adjust the height information of the third coordinate information and / or the height information in the fourth coordinate information to the preset height threshold to complete the height correction.

[0198] As a possible implementation manner, the coordinate correction includes position correction, and the coordinate correction unit includes:

[0199] a determination subunit, configured to determine a target traction position according to the third coordinate information or the fourth coordinate information;

[0200] The determination subunit is further used to determine the intersection of the center line of the first target body and the center line of the second target body according to the third coordinate information and the fourth coordinate information;

[0201] Determine the subunit, and also determine the origin position in the real coordinate system;

[0202] The correction subunit is used to determine that the distance between the intersection point and the target traction position is greater than a preset distance threshold, and then perform coordinate correction on the third coordinate information or the fourth coordinate information based on the origin position and the intersection point.

[0203] As a possible implementation manner, the correction subunit is specifically used for:

[0204] Determine a first distance according to the origin position and the third coordinate information; determine a second distance according to the origin position and the fourth coordinate information;

[0205] If the first distance is greater than or equal to the second distance, position correction is performed on the third coordinate information so that the target traction position is corrected;

[0206] If the first distance is smaller than the second distance, the fourth coordinate information is position-corrected so that the target traction position after position correction is the same as the intersection point.

[0207] The present application also provides a coordinate determination device for a non-rigid target object, which may include a memory and a processor, wherein a computer program is stored in the memory, and when the processor calls the computer program in the memory, the steps provided in the above embodiment may be implemented. Of course, the terminal may also include various network interfaces, power supplies and other components.

[0208] It should be noted that the coordinate determination device for a non-rigid target object provided in an embodiment of the present application has the technical effects of any of the above embodiments, and the embodiments of the present application will not be described in detail here.

[0209] The present application also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed, the steps provided in the above embodiment can be implemented. The storage medium may include: a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes.

[0210] It should be noted that a computer-readable storage medium provided in an embodiment of the present application has the technical effects of any of the above embodiments, and the embodiments of the present application are not described in detail here.

[0211] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0212] Those skilled in the art can understand that the flowchart shown in the figure is only an example in which the embodiments of the present application can be implemented, and the scope of application of the embodiments of the present application is not limited by any aspect of the flowchart.

[0213] In the several embodiments provided in the present application, it should be understood that the disclosed methods, devices and equipment can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of the device or unit can be electrical, mechanical or other forms.

[0214] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, each functional unit in each embodiment of the present application may be integrated into a processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0215] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in the various embodiments of the present application.

[0216] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for determining the coordinates of a non-rigid target object, characterized in that: The method comprises: Acquire an image of a non-rigid target object, wherein the non-rigid target object includes a first target object and at least one second target object; Inputting the image of the non-rigid target object into a pre-trained non-rigid recognition model to obtain first non-rigid information output by the non-rigid recognition model; the first non-rigid information includes: first coordinate information of a 2D target frame of the first target object, second coordinate information of a 2D target frame of the second target object, and a first posture category of the non-rigid target object; Performing 3D calculation according to the first non-rigid body information to obtain third coordinate information of the 3D target frame of the first target in the real coordinate system and fourth coordinate information of the 3D target frame of the second target in the real coordinate system; Coordinate correction is performed on the third coordinate information and the fourth coordinate information to obtain target coordinate information of the non-rigid target object.

2. The method according to claim 1, characterized in that: The non-rigid body recognition model is trained in the following way: Acquire training data, wherein the training data includes a training target image and a label corresponding to the training target image; Inputting the training non-rigid image into the model to be trained, obtaining second non-rigid information output by the model to be trained, wherein the second non-rigid information includes fifth coordinate information of a 2D target frame of a training target body in the training non-rigid image and a second posture category; The model to be trained is trained using the second non-rigid body information and the label until a loss value is less than a threshold, thereby obtaining a non-rigid body recognition model, wherein the loss value is used to measure the gap between the second non-rigid body information and the label.

3. The method according to claim 1, characterized in that If the non-rigid target is in a front-view posture and the non-rigid target is a preset target, the method further includes: obtaining a first preset length of a first target and a second preset length of a second target in the non-rigid target; The 3D solution process includes: The first coordinate information and the second coordinate information are converted into a real coordinate system according to the camera model principle to obtain a first width coordinate of the first target body in the real coordinate system and a second width coordinate of the second target body in the real coordinate system; Determine a first ratio of width to height in the first coordinate information according to the first coordinate information; determine a second ratio of width to height in the second coordinate information according to the second coordinate information; Determine a first height coordinate of the first target body in a real coordinate system according to the first ratio and the first width coordinate; determine a second height coordinate of the second target body in a real coordinate system according to the second ratio and the second width coordinate; The third coordinate information is determined according to the first width coordinate, the first height coordinate, and the first preset length; and the fourth coordinate information is determined according to the second width coordinate, the second height coordinate, and the second preset length.

4. The method according to claim 1, characterized in that If the non-rigid target is in a squint posture and the non-rigid target is a preset target, the non-rigid target information further includes: at least two first key point coordinates in a 2D target frame of the first target and at least two second key point coordinates in a 2D target frame of the second target; The method further includes: obtaining a first preset width of a first target body and a second preset width of a second target body in the non-rigid target object; The 3D solution process includes: Determine the first straight line coordinates where the first key point is located according to at least two coordinates of the first key point; determine the second straight line coordinates where the second key point is located according to at least two coordinates of the second key point; Determine a first length coordinate of the first target body in a real coordinate system according to the first straight line coordinate and the first coordinate information; determine a second length coordinate of the second target body in a real coordinate system according to the second straight line and the second coordinate information; Determine a third width coordinate of the first target body in the real coordinate system according to the first length coordinate and the first preset width; determine a fourth width coordinate of the second target body in the real coordinate system according to the second length coordinate and the second preset width; Determine a first ratio of width to height in the first coordinate information according to the first coordinate information; determine a second ratio of width to height in the second coordinate information according to the second coordinate information; Determine a third height coordinate of the first target body in the real coordinate system according to the first ratio and the third width coordinate; determine a fourth height coordinate of the second target body in the real coordinate system according to the second ratio and the fourth width coordinate; The third coordinate information is determined according to the first length coordinate, the third width coordinate, and the third height coordinate; and the fourth coordinate information is determined according to the second length coordinate, the fourth width coordinate, and the fourth height coordinate.

5. The method according to claim 1, characterized in that The coordinate correction includes height correction, and the coordinate correction of the third coordinate information and the fourth coordinate information includes: Determining that the height information in the third coordinate information and / or the height information in the fourth coordinate information is greater than a preset height threshold; The height information in the third coordinate information and / or the height information in the fourth coordinate information is adjusted to the preset height threshold to complete the height correction.

6. The method according to claim 1, characterized in that The coordinate correction includes position correction, and the coordinate correction of the third coordinate information and the fourth coordinate information includes: Determine a target traction position according to the third coordinate information or the fourth coordinate information; Determine the intersection of the center line of the first target body and the center line of the second target body according to the third coordinate information and the fourth coordinate information; Determine the origin position in the real coordinate system; If it is determined that the distance between the intersection point and the target traction position is greater than a preset distance threshold, coordinate correction is performed on the third coordinate information or the fourth coordinate information based on the origin position and the intersection point.

7. The method according to claim 6, characterized in that The method of performing coordinate correction on the third coordinate information or the fourth coordinate information based on the origin position and the intersection point includes: Determine a first distance according to the origin position and the third coordinate information; determine a second distance according to the origin position and the fourth coordinate information; If the first distance is greater than or equal to the second distance, position correction is performed on the third coordinate information so that the target traction position is corrected; If the first distance is smaller than the second distance, the fourth coordinate information is position-corrected so that the target traction position after position correction is the same as the intersection point.

8. A coordinate determination device for a non-rigid target, characterized in that: The device comprises: An acquisition unit, used for acquiring an image of a non-rigid target object, wherein the non-rigid target object includes a first target object and a second target object; A model output unit is used to input the image of the non-rigid target object into a pre-trained non-rigid recognition model to obtain first non-rigid information output by the non-rigid recognition model; the first non-rigid information includes: first coordinate information of a 2D target frame of the first target object, second coordinate information of the 2D target frame of the second target object, and a posture category of the non-rigid target object; A 3D solving unit, used for performing 3D solving according to the first non-rigid body information, to obtain third coordinate information of the 3D target frame of the first target body in the real coordinate system and fourth coordinate information of the 3D target frame of the second target body in the real coordinate system; The coordinate correction unit is used to perform coordinate correction on the third coordinate information and the fourth coordinate information to obtain the target coordinate information of the non-rigid target object.

9. A coordinate determination device for a non-rigid target object, characterized in that: include: Memory for storing computer programs; A processor, configured to execute the computer program stored in the memory to implement the steps of the method for determining the coordinates of a non-rigid target object as claimed in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: The computer program is executed by a processor to implement the steps of the method for determining the coordinates of a non-rigid target object according to any one of claims 1 to 7.

Images