Dynamic object detection method based on improved epipolar constraint

Abstract

The invention relates to a dynamic object detection method based on improved epipolar constraint. The algorithm comprises the following steps: S1, image acquisition and preprocessing; S2, estimation of a basic matrix; S3, calculation of the distance between a polar line and a feature point and the polar line; according to the algorithm, prior semantic information which is captured by a system and is not used by an original fixed threshold algorithm is fully utilized, the recognition rate of a dynamic object and a static object is improved, and therefore the positioning precision of the system is improved; meanwhile, due to the universality, the tedious work that the threshold value needs to be continuously and manually adjusted when different scenes are coped with is avoided.

Description

Technical Field

[0001] The present invention belongs to the technical field of dynamic SLAM, and in particular relates to a dynamic object detection method based on improved epipolar constraints. Background Art

[0002] In the fields of computer vision and image processing, dynamic feature point removal is a key step in many applications, especially in scenarios such as moving object detection, video surveillance, 3D reconstruction, and robotic visual navigation. Its core goal is to accurately identify and remove feature points corresponding to dynamic objects from image sequences, thereby retaining feature points of the static background and providing a reliable data foundation for subsequent vision tasks.

[0003] Traditional methods for removing dynamic feature points are mainly based on the principle of epipolar constraints. Epipolar constraints are a basic concept in stereo vision, which describes the geometric relationship between a point and its corresponding point under two perspectives. Specifically, when a feature point is matched in two images, it must satisfy the constraint of being on the epipolar line. In the scenario of dynamic object removal, it is usually assumed that the feature points of the static background strictly satisfy the epipolar constraints, while the feature points of the dynamic object will deviate from the epipolar line due to movement. Therefore, by calculating the distance from the feature point to the epipolar line and comparing it with a preset fixed threshold, it is determined whether the point is a dynamic point. If the distance from the feature point to the epipolar line exceeds the threshold, the point is considered to belong to the dynamic object; otherwise, it is considered to be part of the static background.

[0004] This fixed threshold method based on epipolar constraints can achieve certain results in some simple scenarios. For example, when the dynamic object moves slowly, the scene is relatively simple, and the background is relatively static, by setting the threshold appropriately, some dynamic feature points can be removed more accurately. However, with the increasing complexity of application scenarios and the increasing requirements for dynamic object detection accuracy, the limitations of this method have gradually become apparent. Specifically, it has the following limitations:

[0005] 1) Lack of adaptability

[0006] Fixed thresholds cannot adapt to dynamic changes in different scenes. For example, when a dynamic object moves quickly or the scene is complex, a fixed threshold may not accurately distinguish between dynamic and static feature points. Furthermore, when the direction of motion of a dynamic object is close to the epipolar direction, judging only by the perpendicular distance difference can lead to misjudgment.

[0007] 2) Lack of universality

[0008] Different dynamic scenarios require different threshold settings. For example, the motion pattern and speed of the same object in an indoor environment may differ significantly from that in an outdoor setting. A fixed threshold struggles to maintain consistent performance across multiple scenarios. This requires researchers to manually adjust the threshold for each specific scenario, increasing the complexity of the method and making it difficult to apply.

[0009] 3) Limitations of Epipolar Constraints

[0010] The epipolar constraint itself only considers the distance between the feature point and the epipolar line, ignoring other factors that may affect dynamic object detection. This single judgment criterion is prone to misjudgment in complex scenes, reducing the accuracy and robustness of dynamic object removal. Summary of the Invention

[0011] To address the various limitations of fixed thresholds in epipolar constraint algorithms, the present invention proposes a maximum a posteriori probability method. This method replaces the original fixed empirical threshold with a method that compares and calculates the maximum a posteriori probability of the distances of each pixel point in different semantic information categories (people, trees, cars, etc.). If the static probability is less than the dynamic probability, the point is considered a dynamic point; otherwise, it is retained as a valid static point. Compared to the fixed threshold algorithm, the present invention fully utilizes the prior semantic information captured by the system that is not used by the original fixed threshold algorithm, thereby improving the recognition rate of dynamic and static objects and thus improving the positioning accuracy of the system. At the same time, due to its universality, the present invention avoids the tedious task of constantly adjusting the threshold to cope with different scenarios.

[0012] In order to achieve the above object, the technical solution of the present invention is as follows: Based on an improved epipolar constraint dynamic object detection method, the method comprises the following steps:

[0013] S1: Image acquisition and preprocessing,

[0014] S2: estimated fundamental matrix,

[0015] S3: Calculate the distance between the epipolar line and the feature point to the epipolar line,

[0016] S4: Introducing the maximum a posteriori probability method,

[0017] S5: Calculate the posterior probability and judge the dynamics,

[0018] S6: Dynamic object culling.

[0019] Among them, S1: image acquisition and preprocessing, the details are as follows:

[0020] Image acquisition: Use consecutive frames of a binocular camera or a monocular camera to capture two images I1 and I2, ensuring that the time interval between the two images is short enough to capture the motion of dynamic objects;

[0021] Image preprocessing: Preprocess the collected images, including grayscale conversion, denoising, and normalization. Denoising can use Gaussian filtering or other filtering methods to reduce the impact of image noise on subsequent processing;

[0022] Feature point extraction: Extract feature points from two images using algorithms such as SIFT, SURF, or ORB, and match them. The matched feature point pairs are recorded as (p1, p2).

[0023] Specifically, the two parallelograms represent the two frames of images captured by the camera, I1 and I2 respectively. O1 and O2 are the optical centers of the cameras that captured the two frames of images respectively. Points p1 and p2 are the pixel points of the three-dimensional point P on different imaging planes. At the same time, these two points are also a pair of matched feature points. The three points O1, O2, and P determine a plane, called the epipolar plane. The intersection lines of the epipolar plane with the two imaging planes I1 and I2 are l1 and l2 respectively, which are called epipolar lines. Then the coordinates of points p1 and p2 and their secondary coordinates p′1 and p′2 are expressed as:

[0024] p1=[u1,v1],p2=[u2,v2]

[0025] p′1=[u1,v1,1], p′2=[u2,v2,1]

[0026] Among them, S2: estimated basic matrix, as follows,

[0027] Select feature points: Select a set of feature points (usually more than 8 pairs) from the matched feature point pairs to estimate the fundamental matrix F. The selected feature points should be evenly distributed in the image to improve the robustness of the fundamental matrix estimation.

[0028] Based on the knowledge of epipolar geometry constraints, we know that the epipolar line l1 represents the possible projection position of the 3D point P on the imaging plane I1, while the epipolar line l2 represents the possible projection position of the 3D point P on the imaging plane I2. The epipolar line l2 on the imaging plane I2 can be expressed by the equation au + bv + c = 0. The geometric meaning of the fundamental matrix F is the mapping relationship from p1' to the epipolar line l2, that is:

[0029]

[0030] Calculate the fundamental matrix: Use the eight-point method to estimate the fundamental matrix F, let F be:

[0031]

[0032] The fundamental matrix F satisfies:

[0033]

[0034] Optimize the fundamental matrix: The fundamental matrix F is further optimized through nonlinear optimization methods to reduce the estimation error. During the optimization process, the reprojection error can be used as the optimization target to further improve the accuracy of the fundamental matrix.

[0035] Among them, S3: calculates the distance between the epipolar line and the feature point to the epipolar line, as follows:

[0036] After obtaining the epipolar line l2, the point-to-line distance formula is used to calculate the distance D from the pixel point p2 to the epipolar line to satisfy:

[0037]

[0038] In classic dynamic object detection methods, a fixed threshold is set and D is compared with the fixed threshold. If D is greater than the threshold, the point is considered dynamic. If D is less than or equal to the threshold, it is retained. In this invention, however, this step only requires calculating D, rather than using a one-size-fits-all approach of comparing with a fixed threshold.

[0039] Among them, S4: introduces the maximum a posteriori probability method, as follows:

[0040] Set the prior probability P(ω i ): Set the dynamic prior probability P(ω1) and the static prior probability P(ω2) according to the semantic information;

[0041] Set the conditional probability P(X|ω i ): Assume that the distance D from the feature point to the epipolar line follows a normal distribution, and set the mean μ and standard deviation σ for dynamic and static cases; where P(ω1|X) represents the probability that this pixel is a dynamic pixel when D is observed to be X, and P(ω2|X) represents the probability that this pixel is a static pixel when D is observed to be X;

[0042] Calculate the conditional probability P(X|ω i ): Calculate the conditional probability based on the probability density function of the normal distribution:

[0043]

[0044] Among them, P(X|ω1) is the probability of observing D as X when the current pixel is a dynamic pixel, and P(X|ω2) is the probability of observing D as X when the current pixel is a static pixel.

[0045] This step is one of the key steps of the present invention. The present invention does not use a fixed threshold to determine whether a pixel point is a dynamic point. Instead, the following parameters are set in this step: dynamic prior probability P(ω1), mean μ and standard deviation σ in dynamic and static conditions. The present invention sets parameters for objects that may be identified in the image according to the following categories. One category is animals and objects that people often carry, such as cats, dogs, people, backpacks, etc. The dynamic probability of such objects is high, the mean is large, and the standard deviation is small; one category is objects that basically exist as static objects in daily life, such as buildings, chairs, etc. The dynamic probability of such objects is low, the mean is small, and the standard deviation is small; and another category is various means of transportation, such as cars, bicycles, etc. The dynamic probability and static probability of such objects are equal, the mean is large, and the standard deviation is large. Figuratively speaking, this step corresponds to the fixed threshold set in the classic dynamic object detection method, and the parameters set here have the following advantages over the fixed threshold: since the system can read the object properties in the camera, its dynamic nature can be judged more accurately through this information, thereby improving the success rate of dynamic object detection; although the set parameters are more than the original method, once the parameters are set, the system can guarantee the success rate of dynamic object detection in different scenes and environments without adjusting the parameters, while the fixed threshold needs to be adjusted according to the environment to ensure its effect.

[0046] Among them, S5: calculates the posterior probability and determines the dynamics, as follows:

[0047] Calculate the posterior probability λ: Calculate the dynamic and static posterior probabilities according to the Bayesian formula:

[0048]

[0049] Judging dynamicity: If λ>1, the feature point is considered to be a dynamic point; otherwise, the feature point is considered to be a static point.

[0050] Steps S5 and S4 complement each other and are key steps in this invention. Step S4 replaces the fixed threshold parameters of the original method with more stringent and precise parameters. Similar to the original method, which compares D with a fixed threshold to determine its dynamic nature, step S5 calculates the posterior probabilities of the pixel's dynamic and static states and divides them. If the posterior probabilities are > 1, the probability of dynamic is higher, while if they are lower, the probability of static is higher.

[0051] Among them, S6: dynamic object culling, the details are as follows:

[0052] Eliminate dynamic points: Remove the feature points that are judged to be dynamic from the subsequent processing, and only retain the static feature points for subsequent visual tasks. Eliminating dynamic points can reduce the interference of dynamic objects on subsequent processing and improve the robustness of the system.

[0053] Optimization results: The basic matrix F is re-estimated based on the feature points after removal, and steps S3 to S5 are repeated to further optimize the accuracy of dynamic object removal. Through iterative optimization, the effect of dynamic object removal is gradually improved.

[0054] Compared with the prior art, the advantages of the present invention are as follows:

[0055] The accuracy of dynamic object detection is improved, thereby improving the positioning accuracy of the system. By capturing the system's prior information about objects, the present invention uses the maximum a posteriori probability method to improve the system's accuracy in detecting dynamic objects. The present invention selects the sitting and walking sequences in the TUM dataset for dataset experiments. This experiment verifies whether the improved method of the present invention can have better dynamic positioning accuracy than the original method. The compared indicators mainly include the root mean square error (RMSE), mean (Mean), and median (Median) of the absolute trajectory error (APE) and the relative trajectory error (RPE). The improvement of the change amplitude of the improved method compared with the original method.

[0056]

[0057] It can be seen from the table that no matter whether it is a sitting sequence with a lower environmental dynamics or a walking sequence with a higher dynamics, the system positioning accuracy of the improved method used in the present invention is higher.

[0058] Another major advantage of the present invention is that it directly abandons the tedious steps of constantly modifying the fixed threshold value in response to different environments such as indoors and outdoors, and uses the maximum a posteriori probability method, a mathematical method. Although the number of parameters to be set increases, as long as the pre-setting is successful, in theory, no further changes are required during the subsequent system operation process, achieving a once-and-for-all effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 is the epipolar constraint graph. DETAILED DESCRIPTION

[0060] Embodiment: A method for detecting dynamic objects in an image based on an improved epipolar constraint algorithm includes the following steps:

[0061] S1: Image acquisition and preprocessing

[0062] Image acquisition: Use consecutive frames from a binocular or monocular camera to capture two images, I1 and I2. Ensure that the time interval between the two images is short enough to capture the motion of dynamic objects.

[0063] Image preprocessing: This includes grayscale conversion, denoising, and normalization. Denoising can use Gaussian filtering or other filtering methods to reduce the impact of image noise on subsequent processing.

[0064] Feature point extraction: Extract feature points from the two images using algorithms such as SIFT, SURF, or ORB, and then match them. The matched feature point pair is denoted as (p1, p2). When extracting feature points, select those with good discrimination and stability to improve the accuracy of subsequent calculations.

[0065] Specifically, if Figure 1 As shown in the figure, the two blue parallelograms represent the two frames of images captured by the camera, I1 and I2, respectively. O1 and O2 are the optical centers of the cameras that captured these two frames of images, respectively. Points p1 and p2 are the pixels of the three-dimensional point P on different imaging planes. These two points are also a pair of matched feature points. Points O1, O2, and P define a plane called the epipolar plane. The intersection lines of the epipolar plane with the two imaging planes I1 and I2, respectively, are called epipolar lines. The coordinates of points p1 and p2, and their secondary coordinates p1' and p'2, are expressed as:

[0066] p1=[u1,v1],p2=[u2,v2]

[0067] p′1=[u1,v1,1], p′2=[u2,v2,1]

[0068] S2: Estimated fundamental matrix

[0069] Select feature points: Select a set of feature points (usually more than 8 pairs) from the matched feature point pairs to estimate the fundamental matrix F. The selected feature points should be evenly distributed in the image to improve the robustness of the fundamental matrix estimation.

[0070] Based on the knowledge of epipolar geometry constraints, we know that the epipolar line l1 represents the possible projection position of the 3D point P on the imaging plane I1, while the epipolar line l2 represents the possible projection position of the 3D point P on the imaging plane I2. The epipolar line l2 on the imaging plane I2 can be expressed by the equation au+bv+c=0. The geometric meaning of the fundamental matrix F is the mapping relationship from p′1 to the epipolar line l2, that is:

[0071]

[0072] Calculate the fundamental matrix: Use the eight-point method to estimate the fundamental matrix F. Let F be:

[0073]

[0074] The fundamental matrix F satisfies:

[0075]

[0076] Optimize the fundamental matrix: The fundamental matrix F is further optimized through nonlinear optimization methods to reduce the estimation error. During the optimization process, the reprojection error can be used as the optimization target to further improve the accuracy of the fundamental matrix.

[0077] S3: Calculate the distance between the epipolar line and the feature point to the epipolar line

[0078] After obtaining the epipolar line l2, the point-to-line distance formula is used to calculate the distance D from the pixel point p2 to the epipolar line to satisfy:

[0079]

[0080] If the distance is greater than the threshold, the point will be considered a dynamic point. If the distance between pixels is less than or equal to the threshold, it will be retained.

[0081] S4: Introducing the maximum a posteriori probability method

[0082] Set the prior probability P(ω i ): Set the dynamic prior probability P(ω1) and the static prior probability P(ω2) based on semantic information (such as people, trees, and cars). For example, the dynamic probability of a person is 0.7 and the static probability is 0.3; the static probability of a tree is 0.9 and the dynamic probability is 0.1. These prior probabilities can be adjusted according to the actual application scenario.

[0083] Set the conditional probability P(X|ω i ): Assume that the distance D from the feature point to the epipolar line follows a normal distribution and set the mean μ and standard deviation σ for the dynamic and static cases. For example, the mean is larger and the standard deviation is smaller in the dynamic case, while the mean and standard deviation are smaller in the static case. Here, P(ω1|X) represents the probability that a pixel is a dynamic pixel when D is observed to be X, and P(ω2|X) represents the probability that a pixel is a static pixel when D is observed to be X.

[0084] Calculate the conditional probability P(X|ω i ): Calculate the conditional probability based on the probability density function of the normal distribution:

[0085]

[0086] Among them, P(X|ω1) is the probability of observing D as X when the current pixel is a dynamic pixel, and P(X|ω2) is the probability of observing D as X when the current pixel is a static pixel.

[0087] S5: Calculate the posterior probability and determine the dynamics

[0088] Calculate the posterior probability λ: Calculate the dynamic and static posterior probabilities according to the Bayesian formula:

[0089]

[0090] Judging dynamicity: If λ>1, the feature point is considered to be a dynamic point; otherwise, the feature point is considered to be a static point.

[0091] S6: Dynamic Object Culling

[0092] Eliminate dynamic points: Feature points that are judged to be dynamic are removed from subsequent processing, retaining only static feature points for subsequent visual tasks (such as 3D reconstruction and camera positioning). Eliminating dynamic points can reduce the interference of dynamic objects on subsequent processing and improve the robustness of the system.

[0093] Optimization results: Re-estimate the fundamental matrix F based on the removed feature points and repeat steps S3 to S5 to further optimize the accuracy of dynamic object removal. Through iterative optimization, the effect of dynamic object removal can be gradually improved.

[0094] By setting reasonable means and standard deviations for the normal distributions of D under different semantics and different states, and setting empirical probabilities for different states under different semantics, the present invention uses the maximum a posteriori probability method to effectively solve the problems caused by fixed thresholds, improve the recognition rate of dynamic objects and static objects, and thus improve the positioning accuracy of the system.

[0095] Attached is the derivation formula of the maximum a posteriori probability method

[0096] Bayesian formula:

[0097]

[0098] P(ω i |X) is the posterior probability, which means that when the data X is observed, the event ω i Probability of occurrence.

[0099] P(X|ω i ) is the conditional probability, indicating that in the event ω i The probability of observing data X when occurs.

[0100] P(ω i ) is the prior probability, indicating that before there is any observation data, the event ω i Probability of occurrence.

[0101] P(X) is the total probability of observing X, which is calculated using the total probability formula:

[0102]

[0103] Specifically in the invention, P(ω i ) is divided into dynamic probability and static probability, which are the dynamic probability and static probability in a certain semantic information (people, trees, etc.), respectively, denoted as P(ω1) and P(ω2). P(X) is the probability of observing the current pixel to the limit distance D of X. The total probability formula can be written as:

[0104] P(X)=P(X|ω1)P(ω1)+P(X|ω2)P(ω2) (1-8)

[0105] P(X|ω i ) is divided into the dynamic case of P(X|ω1) and the static case of P(X|ω2). P(X|ω1) is the probability of observing D as X when the current pixel is a dynamic pixel, and P(X|ω2) is the probability of observing D as X when the current pixel is a static pixel. i |X) is also divided into P(ω1|X) and P(ω2|X). P(ω1|X) represents the probability that this pixel is a dynamic pixel when D is observed to be X, and P(ω2|X) represents the probability that this pixel is a static pixel when D is observed to be X. We give a specific application of the Bayesian formula in this invention:

[0106]

[0107] It is obvious that if λ>1, this pixel should be eliminated as a dynamic point, otherwise it should be retained as a static point. Observe the right side of the equation, P(ω1) and P(ω2) are prior probabilities. Among the several classic semantic information obtained by YOLO, we have given the corresponding empirical dynamic probability and static probability in advance, such as the dynamic probability of a person is 0.7 and the static probability is 0.3; the static probability of a tree is 0.1 and the dynamic probability is 0.9. This prior information has been determined by the system before this pixel is judged. P(X|ω1) and P(X|ω2) are conditional probabilities. In the case of uniform illumination, we assume that D follows a normal distribution, and the conditional probability P(X|ω i ) can be calculated using the probability density function of the normal distribution:

[0108]

[0109] Among them, μ i is the mean of the normal distribution of D under certain specific semantics in dynamic (denoted as μ1) and static (denoted as μ2) conditions, and σ i is the corresponding standard deviation. These parameters are all empirical parameters. For example, compared with semantic information with weaker uncertainty such as people and trees, which are obviously dynamic or static, the standard deviation of the semantic information of car will be set larger.

[0110] By setting reasonable means and standard deviations for the normal distributions of D under different semantics and different states, and setting empirical probabilities for different states under different semantics, the present invention uses the maximum a posteriori probability method to effectively solve the problems caused by fixed thresholds, improve the recognition rate of dynamic objects and static objects, and thus improve the positioning accuracy of the system.

Claims

1. A method for detecting dynamic objects in images based on an improved epipolar constraint algorithm, characterized in that: The method comprises the following steps: S1: Image acquisition and preprocessing, S2: estimated fundamental matrix, S3: Calculate the distance between the epipolar line and the feature point to the epipolar line, S4: Introducing the maximum a posteriori probability method, S5: Calculate the posterior probability and judge the dynamics, S6: Dynamic object culling.

2. The method for dynamic object detection based on improved epipolar constraint according to claim 1, characterized in that: S1: Image acquisition and preprocessing, as follows: Image acquisition: Use consecutive frames of a binocular camera or a monocular camera to capture two images I1 and I2, ensuring that the time interval between the two images is short enough to capture the motion of dynamic objects; Image preprocessing: Preprocess the collected images, including grayscale conversion, denoising, and normalization. Denoising uses Gaussian filtering or other filtering methods to reduce the impact of image noise on subsequent processing. Feature point extraction: Extract feature points from two images using algorithms such as SIFT, SURF, or ORB, and match them. The matched feature point pairs are recorded as (p1, p2). The two parallelograms represent the two frames of images captured by the camera, I1 and I2 respectively. O1 and O2 are the optical centers of the cameras that captured the two frames of images respectively. Points p1 and p2 are the pixel points of the three-dimensional space point P on different imaging planes. At the same time, these two points are also a pair of matched feature points. The three points O1, O2, and P determine a plane, called the epipolar plane. The intersection lines of the epipolar plane with the two imaging planes I1 and I2 are l1 and l2 respectively, which are called epipolar lines. The coordinates of points p1 and p2 and their secondary coordinates p'1 and p'2 are expressed as: p1=[u1,v1],p2=[u2,v2] p'1=[u1,v1,1], p'2=[u2,v2,1].

3. The method for dynamic object detection based on improved epipolar constraint according to claim 1, characterized in that: S2: Estimate the basic matrix, as follows, Select feature points: Select a set of feature points from the matched feature point pairs to estimate the basic matrix F. The selected feature points should be evenly distributed in the image to improve the robustness of the basic matrix estimation. From the knowledge of epipolar geometry constraints, we know that the epipolar line l1 represents the possible projection position of the three-dimensional space point P on the imaging plane I1, and the epipolar line l2 represents the possible projection position of the three-dimensional space point P on the imaging plane I2. The epipolar line l2 on the imaging plane I2 can be expressed by the equation au+bv+c=0. The geometric meaning of the basic matrix F is the mapping relationship from p1' to the epipolar line l2, that is: Calculate the fundamental matrix: Use the eight-point method to estimate the fundamental matrix F, let F be: The fundamental matrix F satisfies: Optimize the fundamental matrix: The fundamental matrix F is further optimized through nonlinear optimization methods to reduce the estimation error. During the optimization process, the reprojection error is used as the optimization target to further improve the accuracy of the fundamental matrix.

4. The method for dynamic object detection based on improved epipolar constraint according to claim 1, characterized in that: S3: Calculate the distance between the epipolar line and the feature point to the epipolar line, as follows: After obtaining the epipolar line l2, the point-to-line distance formula is used to calculate the distance D from the pixel point p2 to the epipolar line to satisfy:

5. The method for dynamic object detection based on improved epipolar constraint according to claim 1, characterized in that: S4: Introduce the maximum a posteriori probability method, as follows: Set the prior probability P(ω i ): Set the dynamic prior probability P(ω1) and the static prior probability P(ω2) according to the semantic information; Set the conditional probability P(X|ω i ): Assume that the distance D from the feature point to the epipolar line follows a normal distribution, and set the mean μ and standard deviation σ for dynamic and static cases; where P(ω1|X) represents the probability that this pixel is a dynamic pixel when D is observed to be X, and P(ω2|X) represents the probability that this pixel is a static pixel when D is observed to be X; Calculate the conditional probability P(X|ω i ): Calculate the conditional probability based on the probability density function of the normal distribution: Among them, P(X|ω1) is the probability of observing D as X when the current pixel is a dynamic pixel, and P(X|ω2) is the probability of observing D as X when the current pixel is a static pixel.

6. The method for dynamic object detection based on improved epipolar constraint according to claim 1, characterized in that: S5: Calculate the posterior probability and judge the dynamics, as follows: Calculate the posterior probability λ: Calculate the dynamic and static posterior probabilities according to the Bayesian formula: Judging dynamicity: If λ>1, the feature point is considered to be a dynamic point; otherwise, the feature point is considered to be a static point.

7. The method for dynamic object detection based on improved epipolar constraint according to claim 1, characterized in that: S6: Dynamic object culling, as follows: Eliminate dynamic points: Remove the feature points that are judged to be dynamic from the subsequent processing, and only retain the static feature points for subsequent visual tasks. Eliminating dynamic points can reduce the interference of dynamic objects on subsequent processing and improve the robustness of the system. Optimization results: The basic matrix F is re-estimated based on the feature points after removal, and steps S3 to S5 are repeated to further optimize the accuracy of dynamic object removal. Through iterative optimization, the effect of dynamic object removal is gradually improved.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method for dynamically generating and converting multiple views of system design based on publish-subscribe relationships as described in any one of claims 1 to 7 above is implemented.

9. A computer-readable storage medium having computer instructions stored thereon, characterized in that: When the computer instructions are executed by a processor, the method for dynamically generating and converting multiple views of system design based on publish-subscribe relationships according to any one of claims 1 to 7 is implemented.

Images