A monocular vision ranging fusion method and system based on fuzzy selection

Abstract

The present invention provides a monocular vision ranging fusion method and system based on fuzzy selection, which belongs to the field of target tracking technology. The method includes: obtaining image data; obtaining a projection ranging result based on the image data; obtaining a target frame width ranging result based on the image data; performing weighted fusion on the projection ranging result and the target frame width ranging result to obtain a fused ranging result; performing matched filtering tracking on the target vehicle to construct a target trajectory, which records the projection ranging result, the target frame width ranging result, and the fused ranging result in each frame image; calculating the external parameter change value of the current frame through transient correction; and determining the optimal ranging result among the projection ranging result, the target frame width ranging result, and the fused ranging result through fuzzy selection based on the external parameter change value as the observation quantity for subsequent tracking filtering. The present invention can improve ranging accuracy, improve target tracking effect, and improve the performance of intelligent driving systems.

Description

Technical Field

[0001] The present invention relates to the technical field of target tracking, and in particular to a monocular vision ranging fusion method and system based on fuzzy selection. Background Art

[0002] In intelligent driving systems, the accuracy of monocular vision in perceiving target distance and speed is a crucial performance metric in the perception process. On the one hand, due to the constraints of actual hardware computing power in engineering, monocular vision cannot directly obtain speed measurement results. It is necessary to derive the speed component from distance information, thus translating the speed requirement into a higher precision requirement for distance measurement results. On the other hand, target distance and speed measurement information serves as key input for subsequent functional decision-making, planning, and control, directly impacting the performance of intelligent driving systems.

[0003] Currently, in the target tracking process, the distance measurement result is often determined by converting the pixel width of the target in the image. However, in practical applications, the image may be blurred, and the target detection and pixel width calculation may not be accurate enough, resulting in unstable distance and speed measurement results, which in turn affects the target tracking effect and the performance of the intelligent driving system. Summary of the Invention

[0004] The main purpose of the present invention is to provide a monocular vision ranging fusion method and system based on fuzzy selection to solve the problem in the prior art that images may be blurred, target detection and pixel width calculation may be inaccurate, resulting in unstable ranging and speed measurement results, which in turn affects the target tracking effect and the performance of the intelligent driving system.

[0005] To achieve the above object, according to one aspect of the present invention, a monocular vision ranging fusion method based on fuzzy selection is provided, comprising:

[0006] S1: Acquire image data;

[0007] S2: Obtaining a projection ranging result according to the image data;

[0008] S3: Obtaining a target frame width ranging result according to the image data;

[0009] S4: performing weighted fusion on the projection ranging result and the target frame width ranging result to obtain a fused ranging result;

[0010] S5: Perform matched filtering tracking on the target vehicle to construct a target trajectory, wherein the target trajectory records the projection ranging result, the target frame width ranging result, and the fusion ranging result in each frame image;

[0011] S6: Calculate the external parameter change value of the current frame through transient correction;

[0012] S7: According to the extrinsic parameter change value, the optimal ranging result among the projection ranging result, the target frame width ranging result and the fusion ranging result is determined through fuzzy selection as the observation value for subsequent tracking filtering.

[0013] Furthermore, the S2 specifically includes:

[0014] S201: Using a vehicle detection deep learning model, obtain the coordinates of the rear frame of the target vehicle in the current frame image;

[0015] S202: Obtain camera intrinsic and extrinsic parameters;

[0016] S203: Based on the camera intrinsic parameters and extrinsic parameters, the midpoint of the bottom edge of the rear frame is projected into the carrier coordinate system through inverse perspective transformation to obtain a projection ranging result.

[0017] Furthermore, the S3 specifically includes:

[0018] S301: Using a vehicle detection deep learning model, obtain the pixel width of the rear frame of the target vehicle in the current frame image;

[0019] S302: Calculate the target frame width ranging result based on the pixel width of the vehicle tail frame and the preset prior actual physical width:

[0020]

[0021] Among them, x w Indicates the target frame width ranging result, f x Indicates the camera's intrinsic focal length, W pre represents the actual physical width, and w represents the pixel width of the rear frame.

[0022] Furthermore, the S4 is specifically:

[0023] The projection ranging result and the target frame width ranging result are weightedly fused according to the following formula:

[0024] x f =βx p +(1-β)x w

[0025] Among them, x f Indicates the fusion ranging result, x p Represents the projection distance measurement result, x w represents the target frame width ranging result, and β represents the weight coefficient of the projection ranging result.

[0026] Furthermore, determining the optimal ranging result among the projection ranging result, the target frame width ranging result, and the fusion ranging result through fuzzy selection in S7 specifically includes:

[0027] S701: Construct an evaluation factor set U = {degree of closeness between the detected value and the predicted value, whether the current value jumps based on historical detection results, changes in external parameters, and distance to the target};

[0028] S702: Construct an evaluation set V = {available, cannot be determined, unavailable};

[0029] S703: Construct a weight vector A = {a1, a2, a3, a4} for each evaluation factor, where a1 + a2 + a3 + a4 = 1, where a1 represents the weight of the closeness between the detected value and the predicted value, a2 represents the weight of whether the current value has changed based on historical detection results, a3 represents the weight of the external parameter change, and a4 represents the weight of the distance to the target;

[0030] S704: The projected distance measurement result x p , the target frame width ranging result x w And the fused ranging result x f Evaluation is performed based on each evaluation factor to obtain the evaluation matrix R = {R1, R2, R3, R4}, where R1 represents the evaluation result of the closeness between the detection value and the predicted value, R2 represents the evaluation result of whether the current value jumps based on the historical detection results, R3 represents the evaluation result of the external parameter change, and R4 represents the evaluation result of the distance to the target;

[0031] S705: Calculate the comprehensive fuzzy evaluation vector based on the weight vectors of each evaluation factor and the evaluation matrix:

[0032] B=AR

[0033] Among them, B represents the comprehensive fuzzy evaluation vector, A represents the weight vector of each evaluation factor, and R represents the evaluation matrix;

[0034] S706: Determine the distance measurement result corresponding to the value with the highest probability in the comprehensive fuzzy evaluation vector as the optimal distance measurement result.

[0035] Furthermore, the evaluation result R1 of the degree of closeness between the detected value and the predicted value is specifically:

[0036] Predict the target distance value of the current frame based on the target uniform motion model, and determine the unusable probability of each distance measurement result under the factor of the closeness between the detection value and the predicted value:

[0037]

[0038] Among them, r eiIndicates the unavailable probability of the i-th ranging result, min indicates the minimum value, abs indicates the absolute value, x i Indicates the i-th distance measurement result, i = 1, 2, 3, respectively represents the projection distance measurement result x p , target frame width ranging result x w And the fusion ranging result x f , x est Indicates the estimated distance value of the target through prediction, and rt indicates the maximum error ratio;

[0039] According to the unavailable probability of each ranging result, the evaluation result R1 of the closeness between the detection value and the predicted value is calculated:

[0040] R1=[1-r ei ,0,r ei ].

[0041] Furthermore, the evaluation result R2 based on whether the current value of the historical detection result jumps is specifically:

[0042] Based on the variance and average value of the ranging results, the unusable probability of each ranging result is determined based on whether the current value of the historical detection results has jumped:

[0043]

[0044] Among them, r ei Indicates the unavailable probability of the i-th ranging result, min indicates the minimum value, abs indicates the absolute value, x i Indicates the i-th distance measurement result, i = 1, 2, 3, respectively represents the projection distance measurement result x p , target frame width ranging result x w And the fusion ranging result x f , x mi represents the average value of the i-th distance measurement result, max represents the maximum value, δ i represents the variance of the i-th ranging result;

[0045] According to the unavailability probability of each ranging result, the evaluation result R2 based on whether the current value of the historical detection result jumps is calculated:

[0046] R2=[1-r ei ,0,r ei ].

[0047] Furthermore, the evaluation result R3 of the external parameter change is specifically:

[0048] According to the maximum threshold of the external parameter change value, the unavailable probability of each ranging result is determined by the assignment method based on whether the current value of the historical detection result jumps:

[0049]

[0050] Among them, r e represents the unavailable probability of the ranging result, Δp represents the extrinsic parameter change value calculated based on the ranging result, and θ represents the maximum threshold of the extrinsic parameter change value;

[0051] According to the unavailability probability of each ranging result, the evaluation result R3 of the external parameter change is calculated:

[0052] R3=[1-r ei ,0,r ei ].

[0053] Furthermore, the evaluation result R4 of the target distance is specifically:

[0054] Set the target vertical distance range: range = [0, 5, 20, 60, 80, 100, 150, 200, 250] m, and the available probability at different distance ranges:

[0055] The projection distance measurement result x p Available probability at different distance ranges:

[0056] r xp =[0.3,0.4,0.8,0.8,0.9,0.75,0.6,0.2,0.1];

[0057] The target frame width ranging result x w Available probability at different distance ranges:

[0058] r xw =[1,1,0.9,0.8,0.9,0.85,0.8,0.6,0.6];

[0059] The fused ranging result x f Available probability at different distance ranges:

[0060] r xf =[0.8,0.9,1,1,1,0.95,0.9,0.5,0.5];

[0061] According to the target, the current frame distance value is estimated by prediction, and the availability probability r of each ranging result is calculated i ;

[0062] According to the available probability of each ranging result, the evaluation result R4 of the closeness between the detection value and the predicted value is calculated:

[0063] R4=[r i ,0,1-r i ];

[0064] Among them, r i Represents the availability probability of the i-th ranging result.

[0065] According to one aspect of the present invention, a monocular vision ranging fusion system based on fuzzy selection is provided, comprising:

[0066] processor;

[0067] A memory having computer-readable instructions stored thereon, wherein when the computer-readable instructions are executed by the processor, the above-mentioned monocular vision ranging fusion method based on fuzzy selection is implemented.

[0068] By applying the technical solution of the present invention, projection ranging results, target frame width ranging results, and fused ranging results can be obtained in multiple aspects. Through fuzzy selection, the optimal ranging result is determined as the observation quantity for subsequent tracking filtering, avoiding ranging based solely on pixel width, reducing the impact of image blur, improving ranging accuracy, enhancing target tracking effects, and improving the performance of intelligent driving systems.

[0069] In addition to the above-described objects, features and advantages, the present invention has other objects, features and advantages. The present invention will be further described in detail below with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] The accompanying drawings, which constitute part of the present invention, are intended to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are intended to explain the present invention and do not constitute an undue limitation of the present invention. In the accompanying drawings:

[0071] Figure 1 A schematic diagram of a flow chart of a monocular vision ranging fusion method based on fuzzy selection provided by an embodiment of the present invention is shown;

[0072] Figure 2 The figure shows a structural diagram of a monocular vision ranging fusion system based on fuzzy selection provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0073] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments of the present invention can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0074] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0075] It should be noted that the terms "first," "second," and the like in the specification and claims of the present invention and the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a particular order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate to facilitate the embodiments of the present invention described herein. In addition, the terms "including," "having," and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus comprising a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to such processes, methods, products, or apparatuses.

[0076] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof.

[0077] Reference Manual Figure 1 , which shows a flow chart of a monocular vision ranging fusion method based on fuzzy selection provided by an embodiment of the present invention.

[0078] The embodiment of the present invention provides a monocular vision ranging fusion method based on fuzzy selection, including:

[0079] S1: Acquire image data.

[0080] Specifically, image data in front of the car can be captured by a camera.

[0081] S2: Obtain the projection distance measurement result based on the image data.

[0082] In a possible implementation, S2 specifically includes:

[0083] S201: Use the vehicle detection deep learning model to obtain the rear frame coordinates of the target vehicle in the current frame image.

[0084] S202: Obtain camera intrinsic and extrinsic parameters.

[0085] Among them, camera intrinsic parameters are parameters related to the camera's optical properties and sensor geometry, which do not depend on the scene content, including focal length, optical center, distortion coefficient, etc.

[0086] The camera extrinsics describe the position and orientation of the camera relative to a world coordinate system. These parameters depend on the specific settings of the camera in the scene, including rotation, translation, etc.

[0087] S203: Based on the camera intrinsic parameters and extrinsic parameters, the midpoint of the bottom edge of the rear frame is projected into the carrier coordinate system through inverse perspective transformation to obtain a projection ranging result.

[0088] It should be noted that the camera intrinsic parameters (such as focal length and optical center) and extrinsic parameters (i.e., the position and orientation of the camera relative to the vehicle) provide the necessary geometric and positional information to make the transformation from the image plane to the vehicle coordinate system more accurate.

[0089] In this invention, by combining a deep learning model to identify and determine the vehicle's position, and then using the camera's internal and external parameters to perform precise ranging through inverse perspective transformation, the system can more comprehensively understand and predict the dynamic environment on the road, providing solid technical support for intelligent driving.

[0090] S3: Obtain the target frame width ranging result based on the image data.

[0091] In a possible implementation, S3 specifically includes:

[0092] S301: Using a vehicle detection deep learning model, obtain the pixel width of the rear frame of the target vehicle in the current frame image.

[0093] S302: Calculate the target frame width ranging result based on the pixel width of the rear frame and the preset prior actual physical width:

[0094]

[0095] Among them, x w Indicates the target frame width ranging result, f x Indicates the camera's intrinsic focal length, W pre represents the actual physical width, and w represents the pixel width of the rear frame.

[0096] The actual physical width can be determined in advance, which is a standard data that is usually easy to obtain.

[0097] In the present invention, the distance is calculated by the pixel width of the rear frame in the image, which is a cost-effective, simple to implement and highly adaptable method, and is very suitable for application in autonomous driving and driver assistance systems.

[0098] S4: Perform weighted fusion on the projection ranging result and the target frame width ranging result to obtain a fused ranging result.

[0099] In a possible implementation, S4 is specifically:

[0100] The projection ranging result and the target frame width ranging result are weightedly fused according to the following formula:

[0101] x f =βx p +(1-β)x w

[0102] Among them, x f Indicates the fusion ranging result, x p Represents the projection distance measurement result, x w represents the target frame width ranging result, and β represents the weight coefficient of the projection ranging result.

[0103] Among them, those skilled in the art can set the size of the weight coefficient β of the projection ranging result according to actual conditions, and the present invention does not limit it. Optionally, β can be set to 0.5.

[0104] It should be noted that projection ranging relies on the geometric transformation of 3D space and the camera position, while target box width ranging relies on the known size of the target and its representation in the image. These two methods are theoretically complementary, and fusing their results can leverage the advantages of each method and make up for the shortcomings of each method.

[0105] In this paper, by weightedly fusing the projection ranging results with the target frame width ranging results, we can achieve more accurate, stable, and reliable distance perception capabilities, which are crucial for the safe operation of autonomous vehicles. This method leverages the advantages of multiple ranging technologies and improves the overall performance and robustness of the system.

[0106] S5: Perform matched filtering tracking on the target vehicle to construct a target trajectory. The target trajectory records the projection ranging results, target frame width ranging results, and fusion ranging results in each frame image.

[0107] S6: Calculate the extrinsic parameter change value of the current frame through transient correction.

[0108] Transient correction is a technique used to adjust and correct for changes in camera extrinsics (position and orientation) caused by vehicle motion, vibration, or external environmental factors such as uneven roads. In autonomous driving and advanced driver assistance systems, accurately calculating and correcting changes in extrinsic parameters for the current frame is crucial to ensure the accuracy and reliability of the perception system.

[0109] Among them, the external parameter change value can be used to evaluate the superiority of the projection ranging results, the target frame width ranging results and the fusion ranging results.

[0110] Specifically, the image frames that can be captured continuously are used to analyze the camera's motion and external changes. The external parameters obtained during the initial calibration are used as a benchmark. The image changes and sensor data between consecutive frames are analyzed to estimate the preliminary changes in the camera's external parameters. Key feature points are extracted from consecutive image frames and matched between multiple frames. The feature point matching results are used to estimate the relative motion of the camera using visual odometry technology. The results of the visual odometry are combined with sensor data such as IMU and GPS, and the estimation of external parameter changes is optimized using Kalman filters, particle filters, or other sensor fusion technologies. Short-term dynamic changes are analyzed, such as sudden acceleration or deceleration, vehicle tilt or steering, and how these factors affect the camera's external parameters. The external parameters are optimized and corrected based on the fused data. The corrected external parameters are updated to the system for the next image processing and object detection.

[0111] S7: According to the change value of the external parameter, the optimal ranging result among the projection ranging result, the target frame width ranging result and the fusion ranging result is determined through fuzzy selection as the observation value for subsequent tracking filtering.

[0112] Among them, fuzzy selection is a decision-making process based on fuzzy logic, which is used to select the best option among multiple alternative options with uncertain or fuzzy attributes.

[0113] In a possible implementation, determining the optimal ranging result among the projection ranging result, the target frame width ranging result, and the fused ranging result through fuzzy selection in S7 specifically includes:

[0114] S701: Construct an evaluation factor set U = {degree of closeness between the detection value and the predicted value, whether the current value jumps based on historical detection results, changes in external parameters, and distance to the target}.

[0115] The core goal of a ranging system is to accurately measure distance. If the measured value (actual distance) is close to the predicted value (the distance predicted based on a model or previous data), this indicates that the ranging system is performing well under the current conditions and providing reliable data. The degree of closeness between the measured value and the predicted value can also indicate whether the system is correctly calibrated and whether further adjustments or optimization are needed.

[0116] In dynamic environments, continuity and stability are crucial. If new ranging results show significant jumps compared to historical data, this could indicate an abnormal sensor reading or an unexpected event in the environment. Determining whether the current value has changed based on historical detection results can increase confidence in the current measurement and reduce misjudgments caused by occasional noise or errors.

[0117] The extrinsic parameters (such as position and orientation) of an autonomous vehicle's cameras and other sensors can change due to vehicle motion, road conditions, or other external factors. These changes directly impact the interpretation of sensor data. These extrinsic parameter changes can help the system adjust its algorithms to adapt to new operating conditions, ensuring consistent and accurate data processing.

[0118] The distance of an object affects its representation in sensor data. For example, in a vision system, distant objects may appear smaller and less clear. This can affect detection and recognition accuracy. Target distance can help the system determine processing priorities (for example, close objects may require immediate response) and adopt appropriate strategies for processing data at different distances.

[0119] S702: Construct an evaluation set V = {available, cannot be determined, unavailable}.

[0120] S703: Construct the weight vector A = {a1, a2, a3, a4} of each evaluation factor, a1 + a2 + a3 + a4 = 1, a1 represents the weight of the closeness between the detection value and the predicted value, a2 represents the weight of whether the current value jumps based on the historical detection results, a3 represents the weight of the external parameter change, and a4 represents the weight of the distance to the target.

[0121] S704: Projection distance measurement result x p , target frame width ranging result x w And the fusion ranging result x f Evaluation is performed based on each evaluation factor to obtain the evaluation matrix R = {R1, R2, R3, R4}, where R1 represents the evaluation result of the closeness between the detection value and the predicted value, R2 represents the evaluation result of whether the current value jumps based on the historical detection results, R3 represents the evaluation result of the external parameter change, and R4 represents the evaluation result of the target distance.

[0122] Optionally, the evaluation result R1 of the degree of closeness between the detected value and the predicted value is specifically:

[0123] Predict the target distance value of the current frame based on the target uniform motion model, and determine the unusable probability of each distance measurement result under the factor of the closeness between the detection value and the predicted value:

[0124]

[0125] Among them, r ei Indicates the unavailable probability of the i-th ranging result, min indicates the minimum value, abs indicates the absolute value, x i Indicates the i-th distance measurement result, i = 1, 2, 3, respectively represents the projection distance measurement result x p , target frame width ranging result x w And the fusion ranging result x f , x est Indicates the estimated distance value of the target through prediction, and rt indicates the maximum error ratio.

[0126] According to the unavailable probability of each ranging result, the evaluation result R1 of the closeness between the detection value and the predicted value is calculated:

[0127] R1=[1-r ei ,0,r ei ].

[0128] Optionally, the evaluation result R2 based on whether the current value of the historical detection result jumps is specifically:

[0129] Based on the variance and average value of the ranging results, the unusable probability of each ranging result is determined based on whether the current value of the historical detection results has jumped:

[0130]

[0131] Among them, r ei Indicates the unavailable probability of the i-th ranging result, min indicates the minimum value, abs indicates the absolute value, x i Indicates the i-th distance measurement result, i = 1, 2, 3, respectively represents the projection distance measurement result x p , target frame width ranging result x w And the fusion ranging result x f , x mi represents the average value of the i-th distance measurement result, max represents the maximum value, δ i Represents the variance of the i-th ranging result.

[0132] According to the unavailability probability of each ranging result, the evaluation result R2 based on whether the current value of the historical detection result jumps is calculated:

[0133] R2=[1-r ei ,0,r ei ].

[0134] Optionally, the evaluation result R3 of the external parameter change is specifically:

[0135] According to the maximum threshold of the external parameter change value, the unavailable probability of each ranging result is determined by the assignment method based on whether the current value of the historical detection result jumps:

[0136]

[0137] Among them, r e represents the unavailable probability of the ranging result, Δp represents the extrinsic parameter change value calculated based on the ranging result, and θ represents the maximum threshold of the extrinsic parameter change value.

[0138] According to the unavailability probability of each ranging result, the evaluation result R3 of the external parameter change is calculated:

[0139] R3=[1-r ei ,0,r ei ].

[0140] Optionally, the evaluation result R4 of the target distance is specifically:

[0141] Set the target vertical distance range: range = [0, 5, 20, 60, 80, 100, 150, 200, 250] m, and the available probability at different distance ranges:

[0142] Projection distance measurement result x p Available probability at different distance ranges:

[0143] r xp =[0.3,0.4,0.8,0.8,0.9,0.75,0.6,0.2,0.1].

[0144] Target frame width ranging result x w Available probability at different distance ranges:

[0145] r xw =[1,1,0.9,0.8,0.9,0.85,0.8,0.6,0.6].

[0146] Fusion ranging result x f Available probability at different distance ranges:

[0147] r xf =[0.8,0.9,1,1,1,0.95,0.9,0.5,0.5].

[0148] The current frame distance value is estimated by prediction according to the target, and the availability probability of each ranging result is calculated.

[0149] According to the available probability of each ranging result, the evaluation result R4 of the closeness between the detection value and the predicted value is calculated:

[0150] R4=[r i ,0,1-r i ];

[0151] Among them, r i Represents the availability probability of the i-th ranging result.

[0152] S705: Calculate the comprehensive fuzzy evaluation vector based on the weight vectors of each evaluation factor and the evaluation matrix:

[0153] B=AR

[0154] Among them, B represents the comprehensive fuzzy evaluation vector, A represents the weight vector of each evaluation factor, and R represents the evaluation matrix.

[0155] S706: Determine the distance measurement result corresponding to the value with the highest probability in the comprehensive fuzzy evaluation vector as the optimal distance measurement result.

[0156] In the present invention, fuzzy selection allows for consideration of multiple factors (such as the proximity of the detected value to the predicted value, the stability of the ranging result, external parameter changes, and the distance of the target), all of which may affect the accuracy and reliability of the ranging. By combining these factors to select the ranging result, higher accuracy can be achieved in complex environments. In a dynamic and ever-changing environment, a single ranging method may not be reliable enough. Fuzzy selection can select the most appropriate ranging result in the current environment by evaluating the performance of different ranging results under various environmental factors and conditions, thereby improving the adaptability and robustness of the system.

[0157] From the above description, it can be seen that the monocular vision ranging fusion method based on fuzzy selection provided by the present invention can obtain projection ranging results, target frame width ranging results, and fused ranging results in multiple aspects. Through fuzzy selection, the optimal ranging result is determined as the observation quantity for subsequent tracking filtering, avoiding ranging based solely on pixel width, reducing the impact of image blur, improving ranging accuracy, improving target tracking effect, and improving the performance of the intelligent driving system.

[0158] Reference Manual Figure 2 , which shows a structural diagram of a monocular vision ranging fusion system based on fuzzy selection provided by an embodiment of the present invention.

[0159] An embodiment of the present invention provides a monocular vision ranging fusion system 20 based on fuzzy selection, comprising:

[0160] Processor 201;

[0161] The memory 202 stores computer-readable instructions, and when the computer-readable instructions are executed by the processor 201, the monocular vision ranging fusion method based on fuzzy selection is implemented.

[0162] From the above description, it can be seen that the monocular vision ranging fusion system based on fuzzy selection provided by the present invention can obtain projection ranging results, target frame width ranging results and fused ranging results in multiple aspects. Through fuzzy selection, the optimal ranging result is determined as the observation quantity for subsequent tracking filtering, avoiding ranging based solely on pixel width, reducing the impact of image blur, improving ranging accuracy, improving target tracking effect, and improving the performance of the intelligent driving system.

[0163] Unless otherwise specifically stated, the relative arrangement of the parts and steps, the numerical expressions and the numerical values ​​set forth in these embodiments do not limit the scope of the present invention. At the same time, it should be understood that, for ease of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship. The techniques, methods and equipment known to those of ordinary skill in the relevant art may not be discussed in detail, but where appropriate, the techniques, methods and equipment should be considered as part of the authorization specification. In all examples shown and discussed here, any specific values ​​should be interpreted as being merely exemplary and not as limiting. Therefore, other examples of the exemplary embodiments may have different values. It should be noted that similar numbers and letters represent similar items in the following figures, and therefore, once an item is defined in one figure, it does not need to be further discussed in subsequent figures.

[0164] For ease of description, spatially relative terms such as "above", "above", "on the upper surface of", "above", etc. may be used herein to describe the spatial positional relationship of a device or feature to other devices or features as shown in the figures. It should be understood that spatially relative terms are intended to include different orientations of the device in use or operation in addition to the orientation described in the figures. For example, if the device in the drawings is inverted, the device described as "above other devices or structures" or "above other devices or structures" will be positioned as "below other devices or structures" or "below other devices or structures". Thus, the exemplary term "above" can include both "above" and "below". The device can also be positioned in other different ways (rotated 90 degrees or in other orientations), and the spatially relative descriptions used here are interpreted accordingly.

[0165] In the description of the present invention, it should be understood that the directions or positional relationships indicated by directional words such as "front, back, up, down, left, right", "horizontal, vertical, perpendicular, horizontal" and "top, bottom" are usually based on the directions or positional relationships shown in the accompanying drawings. They are only for the convenience of describing the present invention and simplifying the description. Unless otherwise specified, these directional words do not indicate or imply that the device or element referred to must have a specific direction or be constructed and operated in a specific direction. Therefore, they cannot be understood as limiting the scope of protection of the present invention; the directional words "inside and outside" refer to the inside and outside relative to the outline of each component itself.

[0166] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A monocular vision ranging fusion method based on fuzzy selection, characterized in that: include: S1: Acquire image data; S2: Obtaining a projection ranging result according to the image data; S3: Obtaining a target frame width ranging result according to the image data; S4: performing weighted fusion on the projection ranging result and the target frame width ranging result to obtain a fused ranging result; S5: Perform matched filtering tracking on the target vehicle to construct a target trajectory, wherein the target trajectory records the projection ranging result, the target frame width ranging result, and the fusion ranging result in each frame image; S6: Calculate the external parameter change value of the current frame through transient correction; S7: According to the extrinsic parameter change value, the optimal ranging result among the projection ranging result, the target frame width ranging result and the fusion ranging result is determined through fuzzy selection as the observation value for subsequent tracking filtering; The determining of the optimal ranging result among the projection ranging result, the target frame width ranging result, and the fusion ranging result through fuzzy selection in S7 specifically includes: S701: Construct an evaluation factor set U = {degree of closeness between the detected value and the predicted value, whether the current value jumps based on historical detection results, changes in external parameters, and distance to the target}; S702: Construct an evaluation set V = {available, cannot be determined, unavailable}; S703: Construct a weight vector A = {a1, a2, a3, a4} for each evaluation factor, where a1 + a2 + a3 + a4 = 1, where a1 represents the weight of the closeness between the detected value and the predicted value, a2 represents the weight of whether the current value has changed based on historical detection results, a3 represents the weight of the external parameter change, and a4 represents the weight of the distance to the target; S704: The projected distance measurement result x p , the target frame width ranging result x w And the fused ranging result x f Evaluation is performed based on each evaluation factor to obtain the evaluation matrix R = {R1, R2, R3, R4}, where R1 represents the evaluation result of the closeness between the detection value and the predicted value, R2 represents the evaluation result of whether the current value jumps based on the historical detection results, R3 represents the evaluation result of the external parameter change, and R4 represents the evaluation result of the distance to the target; S705: Calculate the comprehensive fuzzy evaluation vector based on the weight vectors of each evaluation factor and the evaluation matrix: B=AR Among them, B represents the comprehensive fuzzy evaluation vector, A represents the weight vector of each evaluation factor, and R represents the evaluation matrix; S706: Determine the distance measurement result corresponding to the value with the highest probability in the comprehensive fuzzy evaluation vector as the optimal distance measurement result.

2. The monocular vision ranging fusion method based on fuzzy selection according to claim 1 is characterized in that: The S2 specifically includes: S201: Using a vehicle detection deep learning model, obtain the coordinates of the rear frame of the target vehicle in the current frame image; S202: Obtain camera intrinsic and extrinsic parameters; S203: Based on the camera intrinsic parameters and extrinsic parameters, the midpoint of the bottom edge of the rear frame is projected into the carrier coordinate system through inverse perspective transformation to obtain a projection ranging result.

3. The monocular vision ranging fusion method based on fuzzy selection according to claim 1 is characterized in that: The S3 specifically includes: S301: Using a vehicle detection deep learning model, obtain the pixel width of the rear frame of the target vehicle in the current frame image; S302: Calculate the target frame width ranging result based on the pixel width of the vehicle tail frame and the preset prior actual physical width: Among them, x w Indicates the target frame width ranging result, f x Indicates the camera's intrinsic focal length, W pre represents the actual physical width, and w represents the width of the rear frame in pixels.

4. The monocular vision ranging fusion method based on fuzzy selection according to claim 1 is characterized in that: The S4 is specifically: The projection ranging result and the target frame width ranging result are weightedly fused according to the following formula: x f =βx p +(1-β)x w Among them, x f Indicates the fusion ranging result, x p Represents the projection distance measurement result, x w represents the target frame width ranging result, and β represents the weight coefficient of the projection ranging result.

5. The monocular vision ranging fusion method based on fuzzy selection according to claim 1 is characterized in that: The evaluation result R1 of the degree of closeness between the detected value and the predicted value is specifically: Predict the target distance value of the current frame based on the target uniform motion model, and determine the unusable probability of each distance measurement result under the factor of the closeness between the detection value and the predicted value: Among them, r ei Indicates the unavailable probability of the i-th ranging result, min indicates the minimum value, abs indicates the absolute value, x i Indicates the i-th distance measurement result, i = 1, 2, 3, respectively represents the projection distance measurement result x p , target frame width ranging result x w And the fusion ranging result x f , x est Indicates the estimated distance value of the target through prediction, and rt indicates the maximum error ratio; According to the unavailable probability of each ranging result, the evaluation result R1 of the closeness between the detection value and the predicted value is calculated: R1＝[1-r ei ,0,r ei ]。 6. The monocular vision ranging fusion method based on fuzzy selection according to claim 1 is characterized in that: The evaluation result R2 based on whether the current value of the historical detection result jumps is specifically: Based on the variance and average value of the ranging results, the unusable probability of each ranging result is determined based on whether the current value of the historical detection results has jumped: Among them, r ei Indicates the unavailable probability of the i-th ranging result, min indicates the minimum value, abs indicates the absolute value, x i Indicates the i-th distance measurement result, i = 1, 2, 3, respectively represents the projection distance measurement result x p , target frame width ranging result x w And the fusion ranging result x f , x mi represents the average value of the i-th distance measurement result, max represents the maximum value, δ i represents the variance of the i-th ranging result; According to the unavailability probability of each ranging result, the evaluation result R2 based on whether the current value of the historical detection result jumps is calculated: R2＝[1-r ei ,0,r ei ]。 7. The monocular vision ranging fusion method based on fuzzy selection according to claim 1 is characterized in that: The evaluation result R3 of the external parameter change is specifically: According to the maximum threshold of the external parameter change value, the unusable probability of each ranging result under the factor of whether the current value of the historical detection result jumps is determined: Among them, r e represents the unavailable probability of the ranging result, Δp represents the extrinsic parameter change value calculated based on the ranging result, and θ represents the maximum threshold of the extrinsic parameter change value; According to the unavailability probability of each ranging result, the evaluation result R3 of the external parameter change is calculated: R3＝[1-r ei ,0,r ei ]。 8. The monocular vision ranging fusion method based on fuzzy selection according to claim 1 is characterized in that: The evaluation result R4 of the target distance is specifically: Set the target vertical distance range: range = [0, 5, 20, 60, 80, 100, 150, 200, 250] m, and the available probability at different distance ranges: The projection distance measurement result x p Available probability at different distance ranges: r xp ＝[0.3,0.4,0.8,0.8,0.9,0.75,0.6,0.2,0.1]； The target frame width ranging result x w Available probability at different distance ranges: r xw ＝[1,1,0.9,0.8,0.9,0.85,0.8,0.6,0.6]； The fused ranging result x f Available probability at different distance ranges: r xf ＝[0.8,0.9,1,1,1,0.95,0.9,0.5,0.5]； According to the target, the current frame distance value is estimated by prediction, and the availability probability r of each ranging result is calculated i ; According to the available probability of each ranging result, the evaluation result R4 of the closeness between the detection value and the predicted value is calculated: R4＝[r i ,0,1-r i ]； Among them, r i Represents the availability probability of the i-th ranging result.

9. A monocular vision ranging fusion system based on fuzzy selection, characterized in that: include: processor; A memory having computer-readable instructions stored thereon, wherein when the computer-readable instructions are executed by the processor, the monocular vision ranging fusion method based on fuzzy selection according to any one of claims 1 to 8 is implemented.

Images