Prediction bounding box dynamic correction method based on multi-dimensional kinematics characteristics

Abstract

The invention provides a prediction bounding box dynamic correction method based on multi-dimensional kinematics characteristics, and belongs to the technical field of prediction bounding box dynamic correction. The method comprises the following steps: initializing parameters and preparing data; track validity judgment; calculating bounding box offset based on historical displacement offset; calculating track acceleration; detecting a turning state; calculating by combining displacement offset, acceleration and bounding box offset in a turning state; carrying out adaptive threshold calculation based on recent motion trend credibility; adjusting bounding box offset in combination with a self-adaptive threshold value; and adjusting the original prediction bounding box. According to the method, parameters can be dynamically adjusted, extra training data is not needed, the method can serve as a lightweight module to be integrated in an existing tracking or prediction system, bounding box prediction errors can be effectively reduced, the target positioning stability in a complex dynamic scene can be improved, and hardware cost and deployment complexity are saved.

Description

TECHNICAL FIELD

[0001] The application relates to a multi-dimensional kinematic feature-based dynamic correction method for a prediction bounding box and belongs to the technical field of dynamic correction of the prediction bounding box. BACKGROUND

[0002] The prediction bounding box is a direct carrier of the target space position, and the accuracy and stability of the prediction bounding box directly affect the target detection, tracking and trajectory prediction effect. Dynamic optimization of the prediction bounding box can correct the position of the bounding box in real time to avoid accumulation of positioning errors when the target motion state changes and the scene is disturbed, thereby providing reliable protection for downstream tasks.

[0003] At present, improving the positioning accuracy of the corrected prediction bounding box is still a difficulty to be solved. In the existing prediction bounding box correction method, the method is mainly divided into two types of physical model driving and data driving. Compared with the data driving method, the physical model driving method has a significant lightweight advantage. However, the existing model driving method optimizes the prediction bounding box through a single model such as a uniform motion model and a linear extrapolation model. Such a method only relies on a single motion assumption and is difficult to adapt to motion mutations, turns and other nonlinear motions in real scenes, which is easy to cause error accumulation. SUMMARY

[0004] The purpose of the present application is to solve the problem that the existing model-driven prediction bounding box correction method cannot adapt to various motion patterns and causes cumulative errors, and to provide a multi-dimensional kinematic feature-based dynamic correction method for a prediction bounding box.

[0005] The present application fuses the multi-dimensional kinematic features of the displacement offset, acceleration and turning state of the historical trajectory, and the adaptive threshold adjustment and prediction uncertainty perception mechanism based on the credibility of the recent motion trend to dynamically correct the original prediction bounding box, realize real-time robust correction of the prediction bounding box to complex dynamic behaviors such as sudden speed changes and turns of the target, and solve the problems of prediction error accumulation and inability to adapt to various scenes in the prior art in a high dynamic scene. The positioning accuracy breaks through the limitations of a single motion model, can effectively suppress noise accumulation, has the engineering practicality advantages of lightweight and no need for additional training data, can effectively reduce the prediction error of the bounding box, and improves the target positioning stability in a complex dynamic scene.

[0006] The purpose of the present application is achieved by the following technical solutions:

[0007] A multi-dimensional kinematic feature-based dynamic correction method for a prediction bounding box comprises the following steps:

[0008] Step one, parameter initialization and data preparation: set the dynamic correction control parameter of the prediction bounding box, determine the target input data, the input data includes: historical trajectory sequence data T and original prediction bounding box ;

[0009] Step two, trajectory validity determination: when the trajectory length n of the historical trajectory sequence data T in step one is less than 2, it will be judged as invalid trajectory, at this time, effective trajectory analysis cannot be carried out, and step nine is directly executed; When the trajectory length n of the historical trajectory sequence data T is greater than or equal to 2, the historical trajectory sequence data T is valid and step three is executed;

[0010] Step three, calculation of bounding box offset based on historical displacement offset: extract the motion trend of the historical trajectory sequence data T and give higher weight to the recent motion in the historical trajectory sequence data T, multiply the offset representing the position change of the adjacent two frames of the bounding box in the historical trajectory sequence data T by the corresponding weight to obtain the weighted average offset reflecting the overall motion trend of the target;

[0011] Step four, trajectory acceleration calculation: the motion state of the target can be quantitatively described by displacement-velocity-acceleration three-level features: displacement is offset; Velocity is approximately represented by offset; Acceleration is calculated by the difference between adjacent velocities;

[0012] Step five, turning state detection: based on the position information in the historical trajectory sequence data, the direction change is quantified by calculating the cosine similarity of the continuous direction vector of the position information, and then it is judged whether the target is in the turning state;

[0013] Step six, bounding box offset calculation combining displacement offset, acceleration and turning state: combine the weighted average offset obtained in step three with the acceleration obtained in step four to generate a basic prediction offset, when the target is in the turning state detected by step five, the basic prediction offset is enhanced to adapt to the motion characteristics of the target turning, and the enhanced prediction offset is obtained;

[0014] Step seven, adaptive threshold calculation based on the credibility of recent motion trend: the adaptive threshold is dynamically constructed to quantify the credibility of the target recent motion trend;

[0015] Step eight, bounding box offset adjustment combined with adaptive threshold: according to the adaptive threshold calculated in step seven, the enhanced prediction offset calculated in step six is further adjusted to obtain the correction offset;

[0016] Step nine, adjustment of the original predicted bounding box: the consistency of the original predicted offset calculated according to the original predicted bounding box and the modified offset obtained in step eight is verified by setting a tolerance range, and the final modified offset is obtained, and when the trajectory length n of the historical trajectory sequence data T in step one is less than 2, the original predicted offset is output.

[0017] Compared with the prior art, the beneficial effects of the present application are:

[0018] The present application fully utilizes the complementarity of the kinematic characteristics of displacement offset, acceleration change and turning state in the historical trajectory and the adaptive threshold adjustment mechanism based on the credibility of the recent motion trend, adopts a rule-driven dynamic correction strategy, realizes the adaptation to complex motion states such as speed change and turning without additional training data, significantly improves the robustness and continuity of the bounding box prediction, effectively suppresses the error accumulation caused by dynamic noise and instantaneous disturbance, and still maintains high stability in challenging scenes such as occlusion.

[0019] The present application includes optional operations, which are a bounding box correction judgment method based on the uncertainty value of the predicted bounding box, which dynamically adapts the prediction reliability to accurately control the correction opportunity, can actively reduce the correction operation in low credibility scenes, and further improves the real-time performance.

[0020] The present application can dynamically adjust parameters without additional training data, and can be integrated as a lightweight module in existing tracking or prediction systems, which significantly saves the hardware cost and deployment complexity while improving the positioning accuracy. BRIEF DESCRIPTION OF DRAWINGS

[0021] Figure 1 It is a basic principle diagram of a prediction bounding box dynamic correction method based on multi-dimensional kinematic characteristics of the present application.

[0022] Figure 2 It is a flowchart for the application of the present application. DETAILED DESCRIPTION

[0023] The present application will be further described in detail below with reference to the accompanying drawings: the present embodiment is implemented on the premise of the technical solution of the present application, and a detailed implementation manner is given, but the protection scope of the present application is not limited to the following embodiments.

[0024] As shown in Figure 1 and Figure 2 , a prediction bounding box dynamic correction method based on multi-dimensional kinematic characteristics related by the present embodiment includes the following steps:

[0025] Step one: parameter initialization and data preparation

[0026] Set the dynamic correction control parameters for the predicted bounding box and determine the input data;

[0027] This method requires the inclusion of nine dynamic correction control parameters for the predicted bounding box, including: historical weight decay factor. Predicting steps Acceleration Influence Factor Turn detection threshold Adjusting the threshold for prediction uncertainty Turning Influence Factors Adaptive threshold scaling factor Adaptive minimum threshold and the original prediction offset adjustment threshold

[0028] The input data for this method consists of three types: first, historical trajectory sequence data T, which is composed of bounding boxes at each time step; second, the original predicted bounding boxes. Third, predicting uncertainty values. The confidence level of this uncertainty ranges from [0-1]. It is worth noting that the closer the value is to 1, the higher the reliability of the corresponding predicted bounding box result.

[0029] Step 2: Determining the validity of the trajectory

[0030] Considering that subsequent analysis of historical trajectory sequence data is required, when the trajectory length n of historical trajectory sequence data T is less than 2, it will be judged as an invalid trajectory, and effective trajectory analysis cannot be performed at this time. Step 10 will be executed directly. When the trajectory length n of historical trajectory sequence data T is greater than 2, historical trajectory sequence data T is valid and step 3 will be executed.

[0031] Step 3: Boundary box adjustment decision based on predicted boundary box uncertainty

[0032] This step is optional. If no prediction bounding box uncertainty value is entered, skip this step and proceed to step four. Here, the prediction bounding box uncertainty value is in the range of [0,1]. The closer it is to 1, the higher the prediction uncertainty, indicating a worse prediction effect. If a prediction bounding box uncertainty value has been entered, the threshold will be adjusted according to the preset prediction uncertainty. Make a judgment: when the uncertainty value of the predicted bounding box is... If the value is less than the prediction uncertainty adjustment threshold, the original prediction bounding box is used directly; otherwise, proceed to step four to adjust the original prediction bounding box.

[0033] Step 4: Calculation of bounding box offset based on historical displacement offset

[0034] This step makes the prediction more consistent with the current motion state of the target by extracting the motion trend of the historical trajectory and giving higher weight to recent motion, inputting the historical trajectory sequence wherein, represents the bounding box at the th time step, is the top-left corner coordinate of the bounding box, is the bottom-right corner coordinate of the bounding box, n is the length of the trajectory, and here , otherwise no correction is needed;

[0035] offset is the position change of the bounding box between two adjacent frames, and its calculation formula is:

[0036]

[0037] wherein, , , to highlight the influence of recent offsets, the weight decays exponentially over time, and the weight of the historical offset between two adjacent frames is calculated as follows:

[0038]

[0039] wherein, is a historical weight decay factor, represents the decay weight of the kth offset, the smaller k is, the older the offset is, and the smaller the weight is, the numerator is raised to the power of , and the denominator is the sum of the decay weights of all offsets, i.e. , to normalize the offset weight;

[0040] After multiplying the offset by the corresponding weight and summing them up, the weighted average offset reflecting the overall motion trend is obtained , and its calculation formula is:

[0041]

[0042] Step five: trajectory acceleration calculation

[0043] The motion state of the target can be quantitatively described by three-level features of displacement, velocity, and acceleration: among them, the displacement is represented by the position change between adjacent frames, i.e. the offset ; the velocity reflects the displacement change per unit time, and in the continuous frame scenario, since the frame interval is a fixed value, the offset ​Approximate representation; acceleration reflects the rate of change of speed per unit time, which is calculated by the difference between adjacent speeds. By quantifying acceleration, the dynamic change trend in the target motion process can be effectively captured, and the cumulative error caused by relying only on average speed for prediction can be corrected in multi-step prediction scenarios, improving prediction accuracy.

[0044] The calculation of acceleration requires at least 3 frames of valid historical trajectory, i.e. the historical trajectory length n≥3. If the historical trajectory length is less than 3, skip this step and go to step six. In the kth time interval in the continuous frame, the speed of the target can be approximated as the offset between adjacent frames, i.e. , The core of acceleration is the rate of change of speed, i.e. the difference between the two adjacent speeds:

[0045]

[0046] wherein, , represents the change from the kth speed to the th speed, to reduce single-frame noise interference, the average of all speed changes is taken as the final acceleration.

[0047]

[0048] wherein, is the average acceleration, with a dimension of 4, containing the acceleration components of the four coordinates, is the top-left corner coordinate of the bounding box, is the bottom-right corner coordinate of the bounding box.

[0049] Step six: turn state detection

[0050] The core of turn state detection is to analyze the degree of change in the target's recent motion direction to determine whether its motion trajectory has significantly turned. This step is based on the position information of the last three frames in the historical trajectory, and the cosine similarity of consecutive direction vectors is calculated to quantify the direction change, and then determine whether it is in a turning state. The essential difference between straight-line motion and turning motion lies in the continuity of the motion direction vector: in the straight-line motion scenario, the direction vectors in adjacent time intervals have high consistency, and the included angle tends to 0°, and the corresponding cosine similarity tends to 1; while in the turning motion scenario, the included angle between adjacent direction vectors will significantly increase, resulting in a significant decrease in cosine similarity and less than a preset threshold. Based on this, by calculating the cosine similarity of the two consecutive direction vectors formed by the last three frames in the target's historical trajectory, the magnitude of the direction change can be quantitatively represented, and the precise detection of the target's turning state can be achieved.

[0051] Here also requires the historical trajectory length n≥3, if the historical trajectory length is less than 3, then skip this step, enter step seven, from the historical trajectory sequence The last three frames of the boundary box coordinates are extracted: , Respectively represent the third frame, the second frame, the last frame of the boundary box, the format is , Is the upper left corner coordinates of the boundary box, Is the right lower corner coordinates of the boundary box, based on the last three frames of coordinates, two continuous motion direction vectors are calculated: , Wherein, Indicates the direction vector from To , indicates the direction vector from To , each component of the vector corresponds to the direction change of the boundary box coordinates ; If the modulus of the direction vector is too small, that is, close to 0, it means that the target is almost stationary in this time period, and there is no turning phenomenon, at this time, it is defaulted to be non-turning state, the modulus calculation formula is:

[0052]

[0053]

[0054] , wherein, The four components of the direction vector, if Or , Is the minimum value, it is judged as non-turning state;

[0055] After that, the effective direction vector is normalized, and the normalization formula is:

[0056]

[0057] After that, the cosine similarity is calculated, which can reflect the consistency of the direction;

[0058]

[0059] , wherein, The included angle of two direction vectors, : the value is closer to 1, the direction is more consistent; the value is closer to-1, the direction is more opposite. (I=1,2,3,4) are the four components of the normalized direction vector , corresponding to the components of , the calculation method is The corresponding components of ; similarly, ​For the corresponding component divided by .

[0060] After calculating the cosine similarity value of the direction vectors of the two directions, if the cosine similarity is less than the turning detection threshold , it is determined that the turning state,

[0061]

[0062] Step seven: combine the displacement offset, acceleration, and boundary box offset amount of the turning state

[0063] First, combine the calculated offset based on historical displacement offset and acceleration to generate a basic predicted offset, and the calculation formula is as follows:

[0064]

[0065] wherein, is the basic predicted offset, is the weighted average offset calculated in step four, is a preset acceleration influence factor, is the average acceleration calculated in step five, is the predicted step number, for example, when predicting the first frame in the future, =1;

[0066] When step six detects that the target is in a turning state, that is, , the basic predicted offset needs to be enhanced to adapt to the motion characteristics when the target turns, and the calculation formula of the enhanced predicted offset is as follows:

[0067]

[0068] wherein, is the enhanced predicted offset, is a preset turning influence factor;

[0069] Step eight: adaptive threshold calculation based on the credibility of recent motion trends

[0070] The core logic of dynamically constructing an adaptive threshold to quantify the reliability of recent motion trends is as follows: when the target is in a stable motion state, that is, when the change ratio of the two most recent offsets is small, the recent motion trend has high reliability and is therefore given priority; when the target undergoes a sudden change in motion, that is, when the change ratio of the two most recent offsets is large, the recent motion trend may be affected by noise or instantaneous disturbances. In this case, it is necessary to use the basic offset obtained by fusing historical information for correction. By calculating the adaptive threshold based on the ratio of the two most recent offsets, the reliability of the recent trend and the dependence on historical patterns can be dynamically balanced, thereby effectively improving the robustness of the predicted offset.

[0071] offset sequence of historical trajectory Extract the last offset and the second-to-last offset These represent the recent trend and the previous trend, respectively, and are represented by the symbols: For each component of the directional vector, calculate the ratio of the absolute value of the recent offset to the previous offset, and take the maximum value as the overall change ratio.

[0072]

[0073]

[0074] in, The four possible values ​​represent Offset components of the four coordinates, It is the coordinate of the top left corner of the bounding box. These are the coordinates of the bottom right corner of the bounding box. for Under the offset component, the ratio of the absolute value of the most recent offset to the previous offset. It is the minimum value. Characterizes the maximum percentage change between the two most recent offsets;

[0075] Based on the change ratio With predictive adaptive threshold scaling factor Generate adaptive threshold The formula is:

[0076]

[0077] in, The minimum threshold of the adaptive threshold. The function is an interval-limited function that ensures the adaptive threshold is not lower than [a certain value]. ;

[0078] Step 9: Adjust the bounding box offset based on the adaptive threshold

[0079] According to the adaptive threshold value calculated in step eight, the enhanced offset calculated in step seven is adjusted Further adjustment is made according to the following rules:

[0080]

[0081] wherein, is the modified offset, is the L2 norm of the vector, used to quantify the overall size of the offset; when the modulus of the enhanced offset is ≤ the adaptive threshold value and the modulus of the recent offset, it is determined that the recent motion trend is reliable, and the is adopted as the adjusted offset; when the modulus of the enhanced offset is > the adaptive threshold value and the modulus of the recent offset, the enhanced offset is directly adopted .

[0082] Step ten: adjustment of the original prediction bounding box

[0083] The consistency of the original prediction offset and the modified offset is verified by setting a tolerance range, so as to realize the complementary advantages of the two prediction results;

[0084] Firstly, the original prediction offset needs to be calculated according to the original prediction bounding box , and the calculation formula is as follows:

[0085]

[0086] wherein, is the original prediction bounding box, is the bounding box of the last frame in the historical trajectory, and both of the two bounding boxes adopt the representation method of the upper left corner coordinates plus the lower right corner coordinates of the bounding box;

[0087] Then, the difference between the original prediction offset and the modified offset obtained in step nine is calculated, and the formula is:

[0088]

[0089] wherein, is the Euclidean distance between the original prediction offset and the modified offset, reflecting the overall deviation degree of the original prediction and the modified result, and the four values of i represent the direction vectors of the four coordinates,

[0090] The tolerance range is calculated according to the set original prediction offset adjustment threshold value , which is in the form of an interval If the original predicted offset is within the tolerance range or the history track length n < 2, the original predicted offset is adopted, otherwise the modified offset is adopted, and the formula is:

[0091]

[0092] wherein, is the final modified offset.

[0093] The above merely provides the preferred embodiments of the present application, which are different implementations based on the overall concept of the present application, and the protection scope of the present application is not limited to this. Any changes or replacements easily conceived by those skilled in the art within the technical scope disclosed by the present application should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for dynamic correction of predicted bounding boxes based on multidimensional kinematic features, characterized in that, Includes the following steps: Step 1: Parameter Initialization and Data Preparation: Set the dynamic correction control parameters for the predicted bounding box and determine the target input data, which includes: historical trajectory sequence data T and the original predicted bounding box. ; Step 2, Trajectory Validity Determination: When the trajectory length n of the historical trajectory sequence data T in Step 1 is less than 2, it will be determined as an invalid trajectory. At this time, effective trajectory analysis cannot be performed, and Step 9 will be executed directly. When the trajectory length n of the historical trajectory sequence data T is greater than or equal to 2, the historical trajectory sequence data T is valid and Step 3 will be executed. Step 3: Calculation of bounding box offset based on historical displacement: Extract the motion trend of historical trajectory sequence data T and give higher weight to recent motion in historical trajectory sequence data T. Multiply the offset representing the change in bounding box position between two adjacent frames in historical trajectory sequence data T with the corresponding weight to obtain the weighted average offset reflecting the overall motion trend of the target. Step 4: Trajectory Acceleration Calculation: The target's motion state can be quantitatively described by three levels of features: displacement, velocity, and acceleration. Displacement is the offset; velocity is approximately represented by the offset; and acceleration is calculated by the difference between adjacent velocities. Step 5, Turning State Detection: Based on the position information in the historical trajectory sequence data, the cosine similarity of the continuous direction vector of the position information is used to quantify the direction change, thereby determining whether the target is in a turning state; Step 6: Calculate the bounding box offset by combining displacement, acceleration, and turning state: Combine the weighted average offset obtained in Step 3 with the acceleration obtained in Step 4 to generate the basic predicted offset. When the target is detected to be turning in Step 5, the basic predicted offset is enhanced to adapt to the motion characteristics of the target when turning, and the enhanced predicted offset is obtained. Step 7: Adaptive threshold calculation based on recent motion trend credibility: Quantify the credibility of the target's recent motion trend by dynamically constructing an adaptive threshold; Step 8: Adjust the bounding box offset based on the adaptive threshold: Based on the adaptive threshold calculated in Step 7, further adjust the enhanced prediction offset calculated in Step 6 to obtain the corrected offset. Step 9: Adjustment of the original predicted bounding box: By setting a tolerance range, the consistency between the original predicted offset calculated based on the original predicted bounding box and the corrected offset obtained in Step 8 is checked to obtain the final corrected offset. When the trajectory length n of the historical trajectory sequence data T in Step 1 is less than 2, the original predicted offset is output.

2. The method for dynamic correction of predicted bounding boxes based on multidimensional kinematic features according to claim 1, characterized in that, The dynamic correction control parameters for the predicted bounding box mentioned in step one include: historical weight decay factor. Predicting steps Acceleration Influence Factor Turn detection threshold Turning Influence Factors Adaptive threshold scaling factor Adaptive minimum threshold and the original prediction offset adjustment threshold .

3. The method for dynamic correction of predicted bounding boxes based on multidimensional kinematic features according to claim 2, characterized in that, The multidimensional kinematic feature parameters mentioned in step one also include a prediction uncertainty adjustment threshold. The input data also includes the predicted bounding box uncertainty value. Input the uncertainty value of the predicted bounding box. When predicting the bounding box uncertainty value If the value is less than the prediction uncertainty adjustment threshold, the original prediction bounding box is used directly; otherwise, proceed to step two to adjust the original prediction bounding box.

4. The method for dynamic correction of predicted bounding boxes based on multidimensional kinematic features according to claim 2, characterized in that, The specific process for obtaining the weighted average offset reflecting the overall movement trend of the target in step three is as follows: Input historical trajectory sequence ,in, Indicates the first The bounding box at each time step. The coordinates of the top-left corner of the bounding box. Here, is the coordinate of the bottom right corner of the bounding box, and n is the trajectory length. Otherwise, no correction is needed; offset It refers to the positional change of the bounding boxes between two adjacent frames, and its calculation formula is: in, , To highlight the impact of recent offsets, the weights decay exponentially over time, and the offsets between two adjacent historical frames are considered. weight The calculation is as follows: in, Historical weight decay factor This represents the decay weight of the k-th offset. The smaller k is, the older the offset, and the smaller the weight. The numerator is... of The power is the sum of the attenuation weights of all offsets, i.e. , used to normalize the offset weights; The formula is to multiply the offset by its corresponding weight and then sum them up: Obtain the weighted average offset that reflects the overall movement trend. .

5. The method for dynamic correction of predicted bounding boxes based on multidimensional kinematic features according to claim 4, characterized in that, The specific calculation method for acceleration mentioned in step four is as follows: In this step, the historical trajectory length n ≥ 3. If the historical trajectory length is less than 3, skip this step and proceed to step five; in consecutive frames, the target's velocity in the k-th time interval. Approximately the offset between adjacent frames ,Right now , The core of acceleration is the rate of change of velocity, that is, the difference between two adjacent velocities: in, , This indicates the speed from the k-th speed to the... To reduce noise interference in a single frame, the average value of all velocity changes is taken as the final acceleration. in, The average acceleration has 4 dimensions, including... Acceleration components of the four coordinates, It is the coordinate of the top left corner of the bounding box. These are the coordinates of the bottom right corner of the bounding box.

6. The method for dynamic correction of predicted bounding boxes based on multidimensional kinematic features according to claim 5, characterized in that, The specific method for determining whether the target is in a turning state in step five is as follows: In this step, the length of the historical trajectory n ≥ 3. If the length of the historical trajectory n < 3, skip this step and proceed to step six, starting from the historical trajectory sequence. Extract the bounding box coordinates of the last three frames: , These represent the bounding boxes of the third-to-last frame, the second-to-last frame, and the last frame, respectively, in the format of... , It is the coordinate of the top left corner of the bounding box. These are the coordinates of the bottom right corner of the bounding box. Based on the coordinates of the most recent three frames, two consecutive motion direction vectors are calculated: ,in, Indicates from arrive directional vector, Indicates from arrive The direction vector, where each component of the vector corresponds to the bounding box coordinates. The change in direction; If the magnitude of the direction vector is too small, i.e., close to 0, it indicates that the target is almost stationary during that time period and there is no turning phenomenon. In this case, it is judged by default to a non-turning state. The formula for calculating the magnitude is: in, Let these be the four components of the direction vector, if or , If the value is a minimum, it is determined to be a non-turning state; The effective direction vectors are then normalized using the following formula: Next, cosine similarity is calculated, which reflects the consistency of direction. in, Let be the angle between two direction vectors. The closer the value is to 1, the more consistent the direction; the closer the value is to -1, the more opposite the direction. i = 1, 2, 3, 4, which are the normalized direction vectors. The four components correspond to The directional component is calculated as follows: The corresponding component divided by Similarly, for The corresponding component divided by ; After calculating the cosine similarity values ​​of the direction vectors in two directions, if the cosine similarity is less than the turning detection threshold... If so, it is determined to be a turning state. This is a function for determining the turning state.

7. The method for dynamic correction of predicted bounding boxes based on multidimensional kinematic features according to claim 6, characterized in that, The specific method for obtaining the enhanced predicted offset in step six is ​​as follows: First, the calculated offset based on historical displacement is combined with the acceleration to generate the basic predicted offset, and the calculation formula is as follows: in, Based on the predicted offset, This refers to the weighted average offset calculated in step three. The preset acceleration influence factor, The average acceleration calculated in step four, To predict the number of steps, When step five detects that the target is in a turning state, that is... The base prediction offset needs to be enhanced to adapt to the motion characteristics of the target during turning. The enhanced prediction offset is calculated using the following formula: in, For the enhanced prediction offset, This is the preset turning influence factor.

8. The method for dynamic correction of predicted bounding boxes based on multidimensional kinematic features according to claim 7, characterized in that, The specific method for dynamically constructing the adaptive threshold in step seven is as follows: The offset sequence of historical trajectories is Extract the last offset and the second-to-last offset These represent the recent trend and the previous trend, respectively, and are represented by the symbols: For each component of the directional vector, calculate the ratio of the absolute value of the recent offset to the previous offset, and take the maximum value as the overall change ratio. in, The four possible values ​​represent Offset components of the four coordinates, It is the coordinate of the top left corner of the bounding box. These are the coordinates of the bottom right corner of the bounding box. for Under the offset component, the ratio of the absolute value of the most recent offset to the previous offset. It is the minimum value. Characterizes the maximum percentage change between the two most recent offsets; Based on the maximum change ratio With predictive adaptive threshold scaling factor Generate adaptive threshold The formula is: in, To achieve an adaptive minimum threshold, The function is an interval-limited function that ensures the adaptive threshold is not lower than [a certain value]. .

9. The method for dynamic correction of predicted bounding boxes based on multidimensional kinematic features according to claim 8, characterized in that, The specific method for obtaining the corrected offset in step eight is as follows: Based on the adaptive threshold calculated in step seven, the enhanced offset calculated in step six is ​​adjusted accordingly. Further adjustments will be made, and the adjustment rules are as follows: in, To correct the offset, The L2 norm of the vector is used to quantize the overall size of the offset; when the magnitude of the enhanced offset is less than or equal to the adaptive threshold... When multiplying by the recent offset magnitude, the reliability of the recent motion trend is determined. As the adjusted offset.

10. The method for dynamic correction of predicted bounding boxes based on multidimensional kinematic features according to claim 9, characterized in that, The specific method for obtaining the final corrected offset in step nine is as follows: By setting a tolerance range, the consistency between the original predicted offset and the corrected offset is verified, thus achieving the complementary advantages of the two prediction results. First, the original predicted offset needs to be calculated based on the original predicted bounding box. The calculation formula is as follows: in, The original predicted bounding box, The bounding box is the last frame in the historical trajectory. Both bounding boxes are represented by the coordinates of the top left corner plus the coordinates of the bottom right corner. Next, the difference between the original predicted offset and the corrected offset obtained in step eight is calculated using the following formula: in, The Euclidean distance between the original prediction offset and the corrected offset reflects the overall degree of deviation between the original prediction and the corrected result. The four possible values ​​represent The direction vectors of the four coordinates, The tolerance range adjusts the threshold based on the set original prediction offset. The calculation is performed, and it is represented as an interval. If the original predicted offset is within the tolerance range or the historical trajectory length n < 2, then the original predicted offset is used; otherwise, the corrected offset is used. The formula is as follows: in, This is the final corrected offset.

Images