Blanking point determination method and device, computer device, and storage medium

Abstract

The application relates to a method and device for determining a blanking point, computer equipment, a storage medium and a computer program product. The method comprises: acquiring a target frame image, extracting a linear road element from the target frame image; determining a segmentation point of the linear road element according to the curvature of the linear road element; segment fitting the linear road element by using the segmentation point to obtain a fitting result; screening a target fitting result from the fitting result according to the coordinate difference between the intersection of the fitting result and a preset pixel point; and in the case that the target fitting result comprises a plurality of linear fitting results, obtaining a dynamic blanking point of the target frame image by using the intersection of the plurality of linear fitting results. The method can improve the determination accuracy of the blanking point and the robustness of the blanking point determination method.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of crowd-sourced map vehicle-side mapping, and in particular to a method and device for determining a vanishing point, a computer device, a storage medium, and a computer program product. BACKGROUND

[0002] A crowd-sourced map is a map obtained by processing map data collected by non-professional collection vehicles while driving on a road. In the field of intelligent driving, a crowd-sourced map has the advantages of low cost, rich data sources, and good real-time performance, in addition to providing lane-level road conditions, compared to a high-definition map made by professional collection vehicles, and can provide perception, positioning, and control information for autonomous vehicles.

[0003] Current vehicle-side crowd-sourced map construction is mostly implemented using a visual three-dimensional reconstruction algorithm: target elements in image data collected by a camera are obtained by performing deep learning processing on the image data. The target elements are processed using a three-dimensional reconstruction algorithm, thereby completing three-dimensional coordinate reconstruction of road elements. For three-dimensional reconstruction of lane lines, road signs, and other target elements, the inverse projection change method is generally used. The processing accuracy of the inverse projection change method is related to the angle between the camera optical axis and the road surface.

[0004] According to the principles of projective geometry, parallel straight lines in the real world will intersect at infinity in the presence of perspective distortion. The projection of this intersection point on the imaging plane is called a vanishing point. The position of the vanishing point is related to the angle between the camera optical axis and the road surface, so the angle can be estimated in real time using the vanishing point, thereby improving the accuracy of vehicle-side crowd-sourced mapping.

[0005] In traditional techniques, multiple spatial vanishing points in an image can be determined based on the principle of line intersection by detecting line segments in the image of a road to be detected. However, in some cases where no lines can be detected or too many lines are detected to directly use the line information, it is difficult to obtain the required vanishing point using the vanishing point determination method in traditional techniques, resulting in low robustness. SUMMARY

[0006] Therefore, it is necessary to provide a method and device for determining a vanishing point with high robustness, a computer device, a computer readable storage medium, and a computer program product to solve the above technical problems.

[0007] In a first aspect, the present application provides a method for determining a vanishing point. The method comprises:

[0008] obtaining a target frame image, and extracting road linear elements from the target frame image;

[0009] determining a segmentation point of the linear element of the road surface according to a curvature of the linear element of the road surface;

[0010] segment fitting the linear element of the road surface by using the segmentation point to obtain a fitting result;

[0011] screening a target fitting result from the fitting result according to a coordinate difference between an intersection of the fitting result and a preset pixel point;

[0012] in a case where the target fitting result includes a plurality of linear fitting results, obtaining a dynamic blanking point of the target frame image by using intersections of the plurality of linear fitting results.

[0013] In one of the embodiments, the determination method of the preset pixel point includes:

[0014] establishing a dynamic blanking point set by using dynamic blanking points of a plurality of continuous frame images collected before the target frame image;

[0015] determining an average blanking point corresponding to the target frame image by using mean coordinates of the dynamic blanking point set;

[0016] processing the dynamic blanking point set by using a time series model to obtain an extrapolated blanking point corresponding to the target frame image;

[0017] determining the preset pixel point according to the average blanking point or the extrapolated blanking point.

[0018] In one of the embodiments, in a case where the target fitting result does not include a plurality of linear fitting results, the method further includes:

[0019] determining an interframe distance corresponding to the target frame image according to image frames corresponding to the dynamic blanking point set;

[0020] in response to the interframe distance being greater than a first preset threshold, using the average blanking point as the dynamic blanking point of the target frame image;

[0021] or, in response to the interframe distance being less than the first preset threshold, using the extrapolated blanking point as the dynamic blanking point of the target frame image.

[0022] In one of the embodiments, the determination method of the preset pixel point includes:

[0023] obtaining an image acquisition parameter of the target frame image;

[0024] determining a static blanking point of the target frame image by using the image acquisition parameter;

[0025] determining the preset pixel point according to the static blanking point.

[0026] In one of the embodiments, the target fitting result is filtered from the fitting results according to a coordinate difference between the intersection point of the fitting result and a preset pixel point.

[0027] The coordinate difference between the intersection point and the preset pixel point in the longitudinal direction is obtained.

[0028] In response to the coordinate difference being less than a second preset threshold, the intersection point is taken as a target intersection point.

[0029] The target fitting result corresponding to the target intersection point is filtered from the fitting results.

[0030] In one of the embodiments, in the case where the target fitting result includes a plurality of linear fitting results, the dynamic blanking point of the target frame image is obtained by intersecting the plurality of linear fitting results.

[0031] The linear identifier and the linear length of each linear fitting result are obtained.

[0032] A random model is established by using the linear identifier and the linear length, and a linear weight of each linear fitting result is generated by using the random model.

[0033] Each linear fitting result is weighted by using the linear weight, and the dynamic blanking point is obtained by intersecting the weighted linear fitting results.

[0034] In a second aspect, the application further provides a blanking point determination device. The device includes:

[0035] An element extraction module is configured to obtain a target frame image and extract a road linear element from the target frame image.

[0036] A segmentation point determination module is configured to determine a segmentation point of the road linear element according to the curvature of the road linear element.

[0037] A segmentation fitting module is configured to segment and fit the road linear element by using the segmentation point to obtain a fitting result.

[0038] A result filtering module is configured to filter a target fitting result from the fitting results according to a coordinate difference between an intersection point of the fitting result and a preset pixel point.

[0039] A blanking point determination module is configured to obtain a dynamic blanking point of the target frame image by intersecting a plurality of linear fitting results in the case where the target fitting result includes the plurality of linear fitting results.

[0040] In a third aspect, the present application provides a computer device. The computer device comprises a memory and a processor, the memory stores a computer program, and the processor implements the method for determining the blanking point according to any one of the embodiments of the first aspect when executing the computer program.

[0041] In a fourth aspect, the present application provides a computer readable storage medium. The computer readable storage medium stores a computer program, and the computer program, when executed by a processor, implements the method for determining the blanking point according to any one of the embodiments of the first aspect.

[0042] In a fifth aspect, the present application provides a computer program product. The computer program product comprises a computer program, and the computer program, when executed by a processor, implements the method for determining the blanking point according to any one of the embodiments of the first aspect.

[0043] The method, device, computer device, storage medium and computer program product for determining the blanking point can obtain a target frame image, extract a road linear element from the target frame image, determine a segmentation point of the road linear element according to a curvature of the road linear element, perform segmented fitting on the road linear element by using the segmentation point, obtain a fitting result, filter a target fitting result from the fitting result according to a coordinate difference between an intersection of the fitting result and a preset pixel point, in the case that the target fitting result comprises a plurality of linear fitting results, obtain a dynamic blanking point of the target frame image by using intersections of the plurality of linear fitting results, can perform segmented fitting on the road linear element by using the segmentation point determined according to the curvature, thereby improving the generation accuracy of the fitting result, filter the target fitting result by using the coordinate difference between the intersection of the preset pixel point and the fitting result, to avoid errors in determining the blanking point caused by non-parallel fitting results, further improve the determination accuracy of the target fitting result, and obtain the dynamic blanking point by using the intersections of the plurality of linear fitting results in the target fitting result, thereby improving the determination accuracy of the blanking point and the robustness of the method for determining the blanking point. BRIEF DESCRIPTION OF DRAWINGS

[0044] Figure 1 An application environment diagram of the method for determining the blanking point in an embodiment;

[0045] Figure 2 A flowchart of the method for determining the blanking point in an embodiment;

[0046] Figure 3 A schematic diagram of a road linear element in an embodiment;

[0047] Figure 4 A flowchart of the method for determining the preset pixel point in an embodiment;

[0048] Figure 5Flowchart of a preset pixel point determination method in another embodiment;

[0049] Figure 6 Flowchart of a preset pixel point determination method in another embodiment;

[0050] Figure 7 Flowchart of a preset pixel point determination method in another embodiment;

[0051] Figure 8 Flowchart of a preset pixel point determination method in another embodiment;

[0052] Figure 9 Flowchart of a preset pixel point determination method in another embodiment;

[0053] Figure 10 Flowchart of a preset pixel point determination method in another embodiment;

[0054] Figure 11 Flowchart of a preset pixel point determination method in another embodiment;

[0055] Figure 12 Flowchart of a preset pixel point determination method in another embodiment; DETAILED DESCRIPTION

[0056] There are currently four methods for estimating the blanking point in China:

[0057] 1. First, the image of the road to be detected is detected. According to the detected line segment, a plurality of spatial blanking points in the image are determined. Among the plurality of spatial blanking points, the line segment corresponding to the target spatial blanking point with the highest confidence is fitted, and the direction indicated by the plurality of target line segments obtained by the straight line fitting is determined. The lane direction of the road to be detected. Obtain the spatial blanking point corresponding to the lane direction.

[0058] 2. The target road picture is input into the deep learning model by using the deep learning method, and the target road picture corresponding to the target segmentation probability graph is output. According to the target segmentation probability graph, the target road picture is determined.

[0059] 3. Randomly extract N line segments from the edge line segments to form a sample set of sample line segments. Create a group of M voting areas for each sample line segment, set M voting boxes, and execute one or more blanking points detected by multiple rounds of blanking point detection and voting.

[0060] 4. The horizontal line segment input unit calculates the intersection of two horizontal line segments. Through the angle point set fan beam projection unit, the fan beam selection unit and the point density calculation unit, the center of each cluster is taken as the candidate blanking point.

[0061] However, the above method is prone to the following problems when executed:

[0062] 1. Since the intersection of the extracted straight lines in the image is used to estimate the blanking point, the case where multiple straight lines exist in a curved scene and the correct blanking point cannot be obtained is not considered.

[0063] 2. The output blanking point cannot be solved for scenes with missing straight line information.

[0064] 3. The efficiency of using voting, clustering and other methods to remove intersection point gross errors is low, and voting clustering essentially uses the probability distribution of multiple candidate blanking points to obtain the probability density center of the blanking point, which cannot directly use the weight of the straight line information in the road information to directly obtain the optimal blanking point.

[0065] 4. Directly extracting the picture straight line does not consider the problem of large picture edge distortion. When the straight line distortion becomes a curve, the distorted straight line cannot be used to solve the blanking point. Moreover, the solved blanking point also has distortion and cannot be directly used for mapping operation of crowdsourcing map.

[0066] 5. When using a deep learning method to extract road linear elements, the generalization ability for different scenes is poor.

[0067] Therefore, in order to solve the problems existing in the prior art, the present application provides a blanking point determination method, device, computer equipment, storage medium and computer program product.

[0068] In order to make the purpose, technical scheme and advantages of the present application clearer, the present application will be further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not to limit the present application.

[0069] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties.

[0070] The blanking point determination method provided by the embodiments of the present application can be applied to, for example Figure 1The application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data required by the server 104 to process. The data storage system can be integrated on the server 104, or placed on the cloud or other network servers. Specifically, the server 104 can obtain the target frame image from the terminal 102, and extract the road linear element from the target frame image. According to the curvature of the road linear element, the segmentation point of the road linear element is determined. The road linear element is segmented and fitted using the segmentation point to obtain the fitting result. According to the coordinate difference between the intersection of the fitting result and the preset pixel point, the target fitting result is selected from the fitting result. In the case where the target fitting result includes multiple linear fitting results, the dynamic blanking point of the target frame image is obtained by intersecting the multiple linear fitting results.

[0071] Among them, the terminal 102 can be but not limited to various personal computers, notebook computers, smart phones, tablet computers, Internet of Things devices and portable wearable devices, and the Internet of Things devices can be smart televisions, smart vehicle-mounted devices, etc. The portable wearable device can be a smart watch, a smart bracelet, a head-mounted device, etc. The server 104 can be implemented by an independent server or a server cluster composed of multiple servers.

[0072] In one embodiment, as Figure 2 shown, a method for determining a blanking point is provided, which is applied to the server 104 in Figure 1 for example, including the following steps:

[0073] Step S202, obtaining a target frame image, and extracting a road linear element from the target frame image.

[0074] Among them, the target frame image can represent one or more frames of road images that need to determine the blanking point. In one example, the road image can be obtained by driving within a preset range by a non-professional image acquisition vehicle.

[0075] The road linear element can include but is not limited to any one or more of the following elements: virtual and real lane lines, road edges, curbs, guardrails, etc. The road linear element can be represented by a point string composed of several (which can be one or more) pixel point coordinates.

[0076] Specifically, the server can obtain the target frame image uploaded by the terminal in real time, or the target frame image collected at a preset time or in a preset time period pushed by the terminal every preset period. The target detection model based on the pre-trained deep learning network is used to extract the road linear element from the target frame image.

[0077] Step S204, according to the curvature of the road linear element, determine the segmentation point of the road linear element.

[0078] Among them, curvature can represent the degree to which the linear elements of the road surface deviate from the straight line.

[0079] Segment points can be used to represent the endpoints of different definition domains in pavement linear elements.

[0080] Specifically, the server can obtain the linear length of the road surface linear element. Perform circle fitting processing on the road surface linear element to obtain the radius of the road surface linear element. The ratio of the radius of the road surface linear element to the linear length is used as the curvature of the road surface linear element. Since the road surface linear elements in the curved scene are mostly in the form of curves, the curvature is generally large. The road surface linear elements in the non-curved scene are approximately straight lines, and the curvature is generally small. The server can use the comparison result of the curvature with the preset curvature threshold to determine the scene in which the road surface linear element is located, and then adopt different segmentation point setting methods for different scenes.

[0081] When the curvature is less than the preset curvature threshold, the deviation of the road surface linear elements from the straight line is small, e.g. Figure 3 As shown in the figure, the road surface linear elements can be directly approximated as straight lines, and the vanishing points can be obtained by intersecting the approximate straight lines of the road surface linear elements. In other words, the segmentation points of the road surface linear elements can be close to the starting point or end point of the road surface linear elements to improve the efficiency of subsequent segment fitting.

[0082] If the curvature is greater than the preset curvature threshold, the road surface linear element deviates significantly from a straight line and may be a curved road. It is necessary to set several (or one or more) segmentation points in the middle of the road surface linear element to improve the accuracy of subsequent segmented fitting.

[0083] Optionally, in some embodiments, the server may also pre-divide the pavement linear element into three segments of equal linear length or equal number of pixels based on the linear length of the pavement linear element or the number of pixels in the pavement linear element: a first segment, a middle segment, and a last segment. The segmentation points {pt1, pt2, pt3, pt4} are determined based on the domain range of the pavement linear element, where pt1 and pt4 are the starting point and end point of the linear element, respectively, and pt2 and pt3 are unknown. When the curvature is less than a preset curvature threshold, pt2 and pt3 can be selected from the last segment of the pavement linear element. When the curvature is greater than the preset curvature threshold, pt2 and pt3 can be selected from the middle segment and the last segment of the pavement linear element, respectively.

[0084] Step S206: Perform segmented fitting on the road surface linear elements using the segmented points to obtain fitting results.

[0085] The fitting results may include linear fitting results (such as a straight line or a line segment) and nonlinear fitting results (such as a curve).

[0086] Specifically, the server can acquire all pixel points between two adjacent segment points of the road linear element, and form a point string corresponding to the current segment by using the pixel points. The server can perform fitting processing on the point string, thereby obtaining a fitting result of each segment of the road linear element, or a fitting result of any segment.

[0087] Optionally, in some embodiments, the server can perform linear fitting on the point string according to the following formula:

[0088] Ax = b

[0089]

[0090] wherein the fitting residual e = Ax - b, x is a value that makes SSR = e T minimum.

[0091] {(x1, y1), (x2, y2), …, (xn, yn)} can represent all pixel points between two adjacent segment points of the road linear element. A is a horizontal coordinate matrix of the pixel points. b is a vertical coordinate matrix of the pixel points. x can represent a fitting coefficient. a can represent a slope of a straight line obtained after fitting. b can represent an intercept of the straight line with the vertical axis obtained after fitting.

[0092] Optionally, in some embodiments, the server can extract a linear fitting result obtained by fitting the first segment of the road linear element (i.e., the part between the starting point of the road linear element and the segment point adjacent to the starting point) according to the following formula:

[0093] Ax + By + C = 0

[0094] wherein A, B, and C are linear coefficients.

[0095] Optionally, in some embodiments, the server can also adjust the positions of the segment points according to the fitting result of the road linear element according to the following formula, thereby facilitating re-fitting by using the adjusted segment points to obtain a fitting result with better effect.

[0096] Minimize SSR(b), b = [pt2, pt3]

[0097] wherein pt2, pt3 are segment points in the road linear element except the starting point and the ending point. pt2 and pt3 are always within the domain of the road linear element. The definition of b can be determined by referring to the definition of the same parameter provided in the above embodiments.

[0098] In step S208, the target fitting result is selected from the fitting results according to the coordinate difference between the intersection of the fitting result and the preset pixel point.

[0099] The preset pixel point can be used to eliminate the non-parallel linear fitting result.

[0100] Specifically, since multiple non-parallel straight lines in the real world do not intersect at a point, and the blanking point is the intersection point of the parallel straight lines in the real world projected in the image. Therefore, the server needs to screen the target fitting result corresponding to the parallel straight line in the real world from the fitting result. The server can obtain a preset pixel point of the target frame image, determine a coordinate difference value of the intersection point of the preset pixel point and the fitting result. According to the coordinate difference value, the target fitting result is screened from the fitting result.

[0101] For example, the server can sort the coordinate difference value corresponding to each intersection point from small to large, and then use the sorting result to screen the fitting result corresponding to the intersection point in the top thirty positions. The screened fitting result is used as the target fitting result. Alternatively, the coordinate difference value can also be directly compared with a preset threshold, and the comparison result is used to screen the target fitting result.

[0102] Alternatively, in some embodiments, the preset pixel point can be a clustering centroid point calculated according to the intersection point set of the fitting result. Alternatively, the preset pixel point can also be a pixel point determined by using the extrinsic parameter of the image acquisition device corresponding to the target frame image. Alternatively, the preset pixel point can also be a pixel point determined by using the continuous multiple frames of images acquired before the target frame image.

[0103] Step S210, in the case where the target fitting result includes multiple linear fitting results, the dynamic blanking point of the target frame image is obtained by intersecting the multiple linear fitting results.

[0104] The dynamic blanking point can be used to represent the blanking point of the target frame image currently calculated.

[0105] Specifically, in the case where the target fitting result includes multiple linear fitting results, the server can obtain the intersection point between the multiple linear fitting results. In response to the case where three or more linear fitting results intersect at the same point, the point is taken as the dynamic blanking point of the target frame image.

[0106] In the blanking point determination method, the target frame image is obtained, the road linear element is extracted from the target frame image, the segmentation points of the road linear element are determined according to the curvature of the road linear element, the road linear element is segmented and fitted by using the segmentation points, and the fitting result is obtained. The segmentation points determined according to the curvature can improve the generation accuracy of the fitting result, so as to solve the problem that the fitting effect of the road linear element is poor in the curved road scene, and even the fitting is difficult. The target fitting result is obtained by screening the fitting result according to the coordinate difference between the intersection of the fitting result and the preset pixel point, so as to eliminate the fitting result which is not parallel by using the coordinate difference, thereby reducing the error of the target fitting result. In the case that the target fitting result includes multiple linear fitting results, the dynamic blanking point of the target frame image is obtained by using the intersection of the multiple linear fitting results, so as to improve the determination accuracy of the blanking point and the robustness of the blanking point determination method, and solve the problem that the determination of the blanking point is affected by distortion when the edge of the target frame image is distorted.

[0107] In one embodiment, as shown in FIG. 1, a preset pixel point determination method is provided, including: Figure 4

[0108] In step S402, the dynamic blanking points of the continuous multiple frames of images collected before the target frame image are used to establish a dynamic blanking point set.

[0109] Specifically, the server can acquire the continuous multiple frames of images collected before the target frame image according to the collection time of the target frame image, and the dynamic blanking point corresponding to each frame of image in the continuous multiple frames of images. The dynamic blanking point set corresponding to the continuous multiple frames of images is established by using the dynamic blanking point corresponding to each frame of image.

[0110] In step S404, the average blanking point corresponding to the target frame image is determined by using the mean coordinate of the dynamic blanking point set.

[0111] Specifically, the server can perform operation processing on the coordinates of each dynamic blanking point in the dynamic blanking point set, and determine the horizontal coordinate mean value and the vertical coordinate mean value corresponding to the dynamic blanking point set. The average blanking point corresponding to the target frame image is determined by using the horizontal coordinate mean value and the vertical coordinate mean value.

[0112] Optionally, in some embodiments, the dynamic blanking points of some image frames in the continuous multiple frames of images may have large deviations. The server can determine the median of the blanking point coordinates by using the dynamic blanking point of each frame of image. The absolute value of the difference between the dynamic blanking point coordinates of each frame of image and the median is obtained. The dynamic blanking points in the dynamic blanking point set whose difference absolute value exceeds a preset threshold value are eliminated. The average blanking point of the target frame image is calculated by using the eliminated dynamic blanking point set.

[0113] ​At step S406, the time series model is used to process the dynamic blanking point set to obtain an extrapolated blanking point corresponding to the target frame image.

[0114] The time series model can include, but is not limited to, an autoregressive moving average model (ARMA model) and an autoregressive integrated moving average model (ARIMA model).

[0115] Specifically, the server can be pre-deployed with a trained time series model. The dynamic blanking point set of the continuous multiple frames of images is input into the time series model, and the predicted position of the blanking point output by the time series model corresponding to the target frame image can be obtained. The pixel point at the predicted position of the blanking point is taken as the extrapolated blanking point.

[0116] Optionally, in some embodiments, taking the ARMA model as an example, the server can determine the horizontal coordinate of the extrapolated blanking point according to the following formula (and the vertical coordinate can also be determined according to the following formula):

[0117]

[0118] wherein x t may represent the horizontal coordinate of the extrapolated blanking point of the target frame image collected at the t-th frame. X t-1 may represent the horizontal coordinate of the dynamic blanking point corresponding to the image at the (t-1)-th frame. X t-p may represent the horizontal coordinate of the dynamic blanking point corresponding to the image at the (t-p)-th frame. p can represent the autoregressive order of the time series model, and q can represent the moving average order of the time series model.

[0119] At step S408, the preset pixel point is determined according to the average blanking point or the extrapolated blanking point.

[0120] Specifically, the server can use the average blanking point or the extrapolated blanking point corresponding to the continuous multiple frames of images as the preset pixel point when calculating the dynamic blanking point of the target frame image. For example, in the case where the continuous multiple frames of images include a preset number of frames of images, the extrapolated blanking point is used as the preset pixel point; in the case where the continuous multiple frames of images are less than the preset number of frames, the average blanking point is used as the preset pixel point. Alternatively, in the case where the dynamic blanking point positions of the continuous multiple frames of images differ by more than a preset error, the extrapolated blanking point is used as the preset pixel point; in the case where the preset error is not exceeded, the average blanking point is used as the preset pixel point.

[0121] In this embodiment, the average blanking point or the extrapolated blanking point corresponding to the dynamic blanking point of the continuous multiple frames of images collected before the target frame image is determined, and the average blanking point or the extrapolated blanking point is used as the preset pixel point for filtering the fitting result of the non-parallel target frame image, so that the determination accuracy of the preset pixel point is improved, and the accuracy of the target fitting result is improved.

[0122] In one embodiment, as shown in FIG. 8, in the case where the target fitting result does not include multiple linear fitting results (for example, there is no linear fitting result in the target fitting result, or the number of linear fitting results in the target fitting result is less than 2), the determination method of the dynamic blanking point includes the following steps. Figure 5

[0123] Step S502, determining the frame interval corresponding to the target frame image according to the image frames corresponding to the dynamic blanking point set.

[0124] Specifically, the server can determine the frame interval between the target frame image and the last image frame with a dynamic blanking point before the target frame image according to the frame sequence numbers of the image frames corresponding to the dynamic blanking point set and the frame sequence number of the target frame image. For example, in the case where the frame sequence numbers of the image frames corresponding to the dynamic blanking point set are {x1, x3, x5, x6}, and the frame sequence number of the target frame image is x8, it can be determined that the frame interval between the target frame image and the last image frame x6 with a dynamic blanking point is 2 frames.

[0125] Alternatively, the server can determine the last image frame with a dynamic blanking point corresponding to the target frame image according to the collection time of the image frames corresponding to the dynamic blanking point set, and then obtain the collection time difference between the image frame and the target frame image, that is, the frame interval.

[0126] Step S504, in response to the frame interval being greater than a first preset threshold, using the average blanking point as the dynamic blanking point of the target frame image.

[0127] Step S506, in response to the frame interval being less than the first preset threshold, using the extrapolated blanking point as the dynamic blanking point of the target frame image.

[0128] Specifically, in the case where the fitting result of the target frame image does not include a linear fitting result, the server can use the average blanking point of the continuous multiple frames of images as the dynamic blanking point of the current target frame image in response to the frame interval being greater than the first preset threshold. Alternatively, in response to the frame interval being less than the first preset threshold, the extrapolated blanking point corresponding to the target frame is calculated and processed as the dynamic blanking point of the target frame image.

[0129] ​In the embodiment, by using the frame interval between the target frame image and the previous image frame with the dynamic blanking point in the case that the fitting result of the target frame image does not include the linear fitting result, the corresponding average blanking point or extrapolated blanking point is selected as the dynamic blanking point of the target frame image, which can be applied to the scene in which the linear fitting result cannot be extracted in the curved road scene or the scene in which the linear fitting result is missing in the target frame image, so as to solve the problem that the target frame image blanking point cannot be solved due to the missing linear fitting result, thereby improving the robustness of the blanking point determination method.

[0130] In one embodiment, as shown in Figure 6 the determination method of the preset pixel point further includes:

[0131] In step S602, the image acquisition parameters of the target frame image are acquired.

[0132] In step S604, the static blanking point of the target frame image is determined by using the image acquisition parameters.

[0133] In step S606, the preset pixel point is determined according to the static blanking point.

[0134] The image acquisition parameters can include but are not limited to the focal length of the image acquisition device used to acquire the target frame image, the offset of the optical axis of the image acquisition device in the image coordinate system, the pitch angle, the yaw angle and the roll angle of the image acquisition device relative to the vehicle body. The image acquisition parameters can be obtained by calibrating the image acquisition device.

[0135] Specifically, the server can acquire the image acquisition parameters of the target frame image. The angle between the optical axis of the image acquisition device corresponding to the target frame image and the road surface is determined by using the image acquisition parameters. According to the relationship between the angle and the image coordinate system blanking point, the angle is calculated to obtain the static blanking point of the target frame image. In the case that there is no continuous multiple frames of images before the target frame image, the static blanking point is taken as the preset pixel point in the target frame image for filtering the non-parallel linear fitting result.

[0136] Optionally, in some embodiments, as shown in Figure 7 the server can also normalize the pixel points on the target frame image by using the image acquisition parameters according to the following formula, so as to facilitate subsequent processing of the target frame image by using the static blanking point:

[0137]

[0138] wherein (X, Y) can represent the normalized pixel point coordinates. (u, v) can represent the coordinates of the pixel points in the target frame image. f x and fy are the focal lengths of the image acquisition device. C xand Cy is an offset of the camera optical axis in the normalized coordinate system.

[0139] In this embodiment, by using the image acquisition parameters corresponding to the target frame image to determine the static blanking point of the target frame image in the case that there is no continuous multiple frame images before the target frame image, and using the static blanking point as the preset pixel point to solve the dynamic blanking point, the application scenarios of solving the dynamic blanking point can be enriched, so as to further improve the robustness of the blanking point determination method.

[0140] In one embodiment, as shown in Figure 8 Step S208, according to the coordinate difference between the intersection of the fitting result and the preset pixel point, the target fitting result is screened from the fitting result, including:

[0141] Step S802, according to the center line of the target frame image, the first fitting result adjacent to the center line is determined from the fitting result.

[0142] Specifically, the server can determine the center line of the target frame image according to the image width or image height of the target frame image. The distance between the fitting result and the center line is obtained. The fitting result with the smallest distance is taken as the first fitting result adjacent to the center line. It can be understood that the first fitting result can include the fitting result with the smallest distance on the left side of the center line, or the fitting result with the smallest distance on the right side of the center line.

[0143] Step S804, the intersection of the first fitting result and the second fitting result other than the first fitting result in the fitting result is obtained.

[0144] Step S806, in response to the coordinate difference between the intersection and the preset pixel point being less than a second preset threshold, the intersection is taken as a target intersection.

[0145] Step S808, the target fitting result corresponding to the target intersection is screened from the fitting result.

[0146] Specifically, the server can determine the second fitting result other than the first fitting result from the fitting result. The first fitting result intersects with each second fitting result to obtain a corresponding intersection. The coordinate difference between the preset pixel point and each intersection in the horizontal axis direction or the vertical axis direction is obtained. In response to the coordinate difference being less than a second preset threshold, the intersection corresponding to the coordinate difference is taken as a target intersection. A plurality of fitting results intersecting at the target intersection are screened from the fitting result, and the plurality of fitting results screened are directly taken as the target fitting result.

[0147] Optionally, in some embodiments, the server may further use the intersection points of the fitting results to form an intersection point set. In response to the coordinate difference being greater than a second preset threshold, the intersection point corresponding to the coordinate difference is removed from the intersection point set, thereby obtaining the target intersection point remaining after the removal.

[0148] In this embodiment, the fitting result is divided into a first fitting result and a second fitting result by the center line of the target frame image, and the first fitting result is sequentially intersected with each second fitting result to obtain corresponding intersection points, which can avoid missing intersection points. In response to the coordinate difference between the intersection point and the preset pixel point being less than a second preset threshold, the intersection point is used as the target intersection point, and then the target fitting result corresponding to the target intersection point is screened and obtained. The intersection points that are far away from the preset pixel point can be eliminated, thereby reducing the determination error of the target fitting result.

[0149] In one embodiment, Figure 9 As shown, in step S210, when the target fitting result includes multiple linear fitting results, the dynamic blanking point of the target frame image is obtained by intersecting the multiple linear fitting results, including:

[0150] Step S902: Obtain the linear identifier and linear length of each linear fitting result.

[0151] The linear length may represent the length of the linear fitting result extracted from the target frame image.

[0152] Specifically, the target frame image may include multiple road linear elements, such as virtual and real lane lines, road edges, curbs, guardrails, etc. The server can perform the following processing operations on each road linear element:

[0153] Determine the centerline of the target frame image based on its width. Assign linear identifiers of -1, -2, ..., -n to the road surface linear elements to the left of the centerline, starting from right to left. Assign linear identifiers of +1, +2, ..., +n to the road surface linear elements to the right of the centerline, starting from left to right. Obtain the linear length of each road surface linear element.

[0154] Step S904: A random model is established using the linear identifier and the linear length, and a linear weight of each linear fitting result is generated by the random model.

[0155] Specifically, the server may use the linear identifier and linear length of the linear fitting result to establish a random model with reference to the following formula, and generate a linear weight for each linear fitting result through the random model:

[0156]

[0157] Wherein, id can represent the linear identity of the linear fitting result. length can represent the linear length of the linear fitting result. σ 2 may represent the square of the standard deviation. a 2 may represent the square of the linear weight.

[0158] Step S906, each linear fitting result is weighted using the linear weight, and the processed linear fitting results are intersected to obtain the dynamic blanking point.

[0159] Specifically, the server can weight the pixel point coordinates in the linear fitting result using the linear weight of the linear fitting result, and intersect the processed linear fitting result. The intersection point of the processed linear fitting result obtained by least square fitting is the dynamic blanking point of the target frame image.

[0160] Optionally, in some embodiments, the server can refer to the following formula to obtain the dynamic blanking point:

[0161]

[0162] Wherein, Ax = b can represent the linear regression equation of the dynamic blanking point, x can represent the dynamic blanking point. (x, y) can represent the pixel point coordinates of the linear fitting result corresponding to the target intersection point. A can represent the fitting matrix. u n may represent the fitting parameter. b can represent the weight matrix of the linear fitting result corresponding to the target intersection point. p sn may represent the linear weight of the linear fitting result s with the linear identity n.

[0163] In this embodiment, the linear weight corresponding to the linear fitting result corresponding to each road surface linear element is generated by a random model, and the dynamic blanking point of the target frame image is obtained by weighted least square processing of the linear fitting result corresponding to the target intersection point using the linear weight. It can be applied to the scene where a single frame image contains multiple road surface linear elements, thereby improving the robustness of the blanking point determination method, and also improving the determination accuracy and accuracy of the dynamic blanking point.

[0164] In one embodiment, as Figure 10 shown, the determination method of the blanking point provided by the present application comprises:

[0165] Step S1002, obtaining a target frame image and an image acquisition parameter of the target frame image, and determining a static blanking point of the target frame image according to the image acquisition parameter.

[0166] Specifically, the server can obtain a target frame image and image acquisition parameters of a camera acquiring the target frame image, such as a focal length, a camera optical axis offset, a pitch angle, a yaw angle, and a roll angle of the camera relative to a vehicle body, and the like. The pitch angle of the camera relative to the vehicle body and the offset of the camera optical axis in the image coordinate system are used to determine a static blanking point of the target frame image.

[0167] At step S1004, a road surface linear element is extracted from the target frame image, and the road surface linear element is projected to a normalized coordinate system of the image acquisition device using the image acquisition parameters.

[0168] Specifically, the server can extract the road surface linear element from the target frame image using a deep learning network or a deep learning model. The focal length of the camera acquiring the target frame image and the camera optical axis offset are used to normalize the road surface linear element in the target frame image, and the road surface linear element is projected to the normalized coordinate system of the image acquisition device.

[0169] At step S1006, a segmentation point is determined according to the curvature of the road surface linear element, and the road surface linear element is segmented and fitted using the segmentation point to obtain a fitting result.

[0170] Specifically, the server can divide the road surface linear element into a first segment, a middle segment, and a tail segment according to the length of the road surface linear element. The road surface linear element is circularly fitted to determine the curvature of the road surface linear element. In response to the curvature being greater than a preset curvature threshold, two segmentation points are set in the middle segment and the tail segment, respectively. Alternatively, in response to the curvature being less than the preset curvature threshold, the two segmentation points are both set in the tail segment. The road surface linear element is segmented and fitted using the segmentation points, and the positions of the segmentation points are adjusted according to the fitting result. The adjustment is performed until the fitting result obtained by segmenting and fitting the road surface linear element using the adjusted segmentation points satisfies a minimum residual sum of squares.

[0171] At step S1008, an intersection of the first fitting result and each second fitting result is obtained according to the center line of the target frame image.

[0172] Specifically, the server can determine a first fitting result adjacent to the center line and a second fitting result other than the first fitting result according to the center line of the target frame image and the distance between each fitting result and the center line. The intersection of the first fitting result and each second fitting result is obtained to establish a set of intersections.

[0173] At step S1010, a preset pixel point of the target frame image is obtained, and a target fitting result is selected from the fitting results using a coordinate difference between the preset pixel point and the intersection.

[0174] Specifically, the server can adopt the static blanking point as the preset pixel point in the case that there are no continuous multiple frame images before the target frame image. In the case that there are continuous multiple frame images before the target frame image, the average blanking point or the extrapolated blanking point corresponding to the target frame is calculated by using the dynamic blanking point set of the continuous multiple frame images. The average blanking point or the extrapolated blanking point is adopted to replace the static blanking point, so as to update the preset pixel point.

[0175] The coordinate difference of the intersection point with the fitting result is determined by using the preset pixel point. In response to the coordinate difference being less than a second preset threshold, the intersection point corresponding to the coordinate difference is taken as a target intersection point. The target fitting result corresponding to the target intersection point is screened from the fitting result.

[0176] Step S1012, in the case that the target fitting result includes multiple linear fitting results, the multiple linear fitting results are weighted and fitted by using linear weights to obtain the dynamic blanking point of the target frame image.

[0177] Specifically, in the case that the target fitting result includes multiple linear fitting results, the server can establish a random model related to the linear fitting result by using the linear identifier and the linear length of the linear fitting result. The linear weight of the linear fitting result is output by the random model. The linear fitting result is weighted and least square fitted by using the linear weight to obtain the dynamic blanking point of the target frame image.

[0178] Step S1014, in the case that the target fitting result does not include multiple linear fitting results, the dynamic blanking point of the target frame image is determined by using the average blanking point or the extrapolated blanking point.

[0179] Specifically, in the case that the target fitting result does not include multiple linear fitting results, the server can obtain the dynamic blanking point set of the continuous multiple frame images collected before the target frame image. The frame interval between the target frame image and the last image frame with the dynamic blanking point is determined. In response to the frame interval being greater than a first preset threshold, the average blanking point is selected as the dynamic pixel point of the target frame image. In response to the frame interval being less than the first preset threshold, the extrapolated blanking point is selected as the dynamic pixel point of the target frame image.

[0180] In this embodiment, the road linear elements are fitted by segment fitting, which can improve the accuracy of the fitting result, solve the problem that the vanishing point cannot be extracted from the road linear elements or too many straight lines are extracted to solve the vanishing point in a curved road scene, and obtain the target fitting result by screening the fitting result by using the static vanishing point, the average vanishing point, or the extrapolated vanishing point, which can remove the error influence of non-parallel fitting results on subsequent vanishing point determination, and obtain the dynamic vanishing point by using the linear weight generated by the random model for weighted least square fitting, which can improve the determination accuracy of the vanishing point. By not including multiple linear fitting results in the target fitting result, the average vanishing point or the extrapolated vanishing point is selected as the dynamic vanishing point of the target frame image by using the frame interval, which can improve the robustness of the vanishing point determination method.

[0181] It should be understood that, although each step in the flowchart involved in each embodiment as described above is displayed in sequence according to the arrow, these steps are not necessarily executed in sequence according to the order of the arrow. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other orders. Moreover, at least part of the steps in the flowchart involved in each embodiment as described above can include multiple steps or stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily sequential, but can be alternately or alternately executed with at least part of other steps or steps or stages in other steps.

[0182] Based on the same inventive concept, the embodiments of the present application also provide a vanishing point determination apparatus for implementing the above-mentioned vanishing point determination method. The implementation scheme for solving the problem provided by the apparatus is similar to the implementation scheme described in the above method, and therefore the specific limitations in one or more vanishing point determination apparatus embodiments provided below can refer to the limitations of the vanishing point determination method described above, which will not be described here again.

[0183] In one embodiment, as shown in Figure 11 A vanishing point determination apparatus 1100 is provided, which includes an element extraction module 1102, a segmentation point determination module 1104, a segmented fitting module 1106, a result screening module 1108, and a vanishing point determination module 1110, wherein:

[0184] The element extraction module 1102 is configured to obtain a target frame image and extract road linear elements from the target frame image.

[0185] The segmentation point determination module 1104 is configured to determine the segmentation points of the road linear elements according to the curvatures of the road linear elements.

[0186] The segmented fitting module 1106 is configured to perform segmented fitting on the linear elements of the road surface by using the segmented points to obtain fitting results.

[0187] The result screening module 1108 is configured to screen target fitting results from the fitting results according to coordinate differences between intersection points of the fitting results and the preset pixel point.

[0188] The blanking point determination module 1110 is configured to, in a case where the target fitting results include a plurality of linear fitting results, obtain a dynamic blanking point of the target frame image by intersecting the plurality of linear fitting results.

[0189] In an embodiment, the blanking point determination apparatus 1100 further includes a pixel point determination module configured to: establish a dynamic blanking point set by using dynamic blanking points of a plurality of continuous frame images collected before the target frame image; determine an average blanking point corresponding to the target frame image by using mean coordinates of the dynamic blanking point set; process the dynamic blanking point set by using a time series model to obtain an extrapolated blanking point corresponding to the target frame image; and determine the preset pixel point according to the average blanking point or the extrapolated blanking point.

[0190] In an embodiment, the blanking point determination module 1110 is further configured to, in a case where the target fitting results do not include a plurality of linear fitting results, determine an inter-frame interval corresponding to the target frame image according to image frames corresponding to the dynamic blanking point set, and in response to the inter-frame interval being greater than a first preset threshold, use the average blanking point as the dynamic blanking point of the target frame image, or in response to the inter-frame interval being less than the first preset threshold, use the extrapolated blanking point as the dynamic blanking point of the target frame image.

[0191] In an embodiment, the pixel point determination module is further configured to: obtain an image acquisition parameter of the target frame image; determine a static blanking point of the target frame image by using the image acquisition parameter; and determine the preset pixel point according to the static blanking point.

[0192] In an embodiment, the result screening module 1108 is further configured to: determine a first fitting result adjacent to a center line of the target frame image from the fitting results; obtain an intersection point of the first fitting result and a second fitting result other than the first fitting result in the fitting results; in response to a coordinate difference between the intersection point and the preset pixel point being less than a second preset threshold, use the intersection point as a target intersection point; and screen target fitting results corresponding to the target intersection point from the fitting results.

[0193] In an embodiment, the blanking point determination module 1110 is further configured to: obtain a linear identifier and a linear length of each linear fitting result; establish a random model by using the linear identifier and the linear length; generate a linear weight of each linear fitting result by using the random model; perform weighted processing on each linear fitting result by using the linear weight; and obtain the dynamic blanking point by intersecting the processed linear fitting results.

[0194] Each module in the above-mentioned device for determining the vanishing point can be implemented in whole or in part through software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in the computer device in the form of software, so that the processor can call and execute the corresponding operations of each module.

[0195] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 12 As shown. The computer device includes a processor, a memory and a network interface connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data such as target frame images, road surface linear elements, curvature, segmentation points, fitting results, linear fitting results, intersection sets, dynamic vanishing points, etc. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a method for determining a vanishing point is implemented.

[0196] Those skilled in the art will understand that Figure 12 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0197] In one embodiment, a computer device is further provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps in the above method embodiments when executing the computer program.

[0198] In one embodiment, a computer-readable storage medium is further provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0199] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0200] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when the computer program is executed, the processes of the above-mentioned embodiments of the methods can be included. Any reference to memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (Read-Only Memory, ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive memory (ReRAM), magnetoresistive random access memory (Magnetoresistive Random Access Memory, MRAM), ferroelectric memory (Ferroelectric Random Access Memory, FRAM), phase change memory (Phase Change Memory, PCM), graphene memory, etc. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory, etc. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (Static Random Access Memory, SRAM) or dynamic random access memory (Dynamic Random Access Memory, DRAM), etc. The database involved in the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a block chain, etc., without being limited thereto. The processor involved in the embodiments provided in the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., without being limited thereto.

[0201] Any combination of the technical features of the above embodiments can be made. In order to make the description simple, all possible combinations of the technical features in the above embodiments are not described, however, as long as the combination of the technical features does not exist contradictory, it should be considered as the scope of the present application.

[0202] The above embodiments only express several implementation manners of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the scope of the patent of the present application. It should be pointed out that for ordinary skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are within the scope of protection of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A method of determining a blanking point, characterized in that, The method comprises: acquiring a target frame image, and extracting a road linear element from the target frame image; determining a segmentation point of the road linear element according to a curvature of the road linear element; segmenting and fitting the road linear element by using the segmentation point to obtain a fitting result; screening a target fitting result from the fitting result according to a coordinate difference between an intersection of the fitting result and a preset pixel point; in a case where the target fitting result comprises a plurality of linear fitting results, obtaining a dynamic blanking point of the target frame image by intersecting the plurality of linear fitting results; wherein, in a case where there are continuous multiple frame images before the target frame image, the method for determining the preset pixel point comprises: establishing a dynamic blanking point set by using dynamic blanking points of the continuous multiple frame images collected before the target frame image; determining an average blanking point corresponding to the target frame image by using a mean coordinate of the dynamic blanking point set; processing the dynamic blanking point set by using a time series model to obtain an extrapolated blanking point corresponding to the target frame image; determining the preset pixel point according to the average blanking point or the extrapolated blanking point; in a case where there are no continuous multiple frame images before the target frame image, the method for determining the preset pixel point comprises: acquiring an image acquisition parameter of the target frame image; determining a static blanking point of the target frame image by using the image acquisition parameter; determining the preset pixel point according to the static blanking point.

2. The method of claim 1, wherein, in a case where the target fitting result does not comprise a plurality of linear fitting results, the method further comprises: determining an interframe distance corresponding to the target frame image according to image frames corresponding to the dynamic blanking point set; in response to the interframe distance being greater than a first preset threshold, using the average blanking point as a dynamic blanking point of the target frame image; in response to the interframe distance being less than the first preset threshold, using the extrapolated blanking point as a dynamic blanking point of the target frame image.

3. The method of any one of claims 1-2, wherein, the screening of a target fitting result from the fitting result according to a coordinate difference between an intersection of the fitting result and a preset pixel point comprises: determining a first fitting result adjacent to a center line of the target frame image from the fitting result; acquiring an intersection of the first fitting result and a second fitting result other than the first fitting result in the fitting result; in response to a coordinate difference between the intersection and the preset pixel point being less than a second preset threshold, taking the intersection as a target intersection; screening the target fitting result corresponding to the target intersection from the fitting result.

4. The method of any one of claims 1-2, wherein, in a case where the target fitting result comprises a plurality of linear fitting results, the obtaining of a dynamic blanking point of the target frame image by intersecting the plurality of linear fitting results comprises: acquiring a linear identifier and a linear length of each linear fitting result; establishing a random model by using the linear identifier and the linear length, and generating a linear weight of each linear fitting result by using the random model; The linear fitting results are weighted by using the linear weights, and the processed linear fitting results are intersected to obtain the dynamic blanking point.

5. An apparatus for determining a blanking point, characterized in that The device comprises: An element extraction module is configured to acquire a target frame image and extract a road linear element from the target frame image; A segmentation point determination module is configured to determine a segmentation point of the road linear element according to a curvature of the road linear element; A segmentation fitting module is configured to segment and fit the road linear element by using the segmentation point to obtain a fitting result; A result screening module is configured to screen a target fitting result from the fitting result according to a coordinate difference between an intersection of the fitting result and a preset pixel point; A blanking point determination module is configured to, in a case where the target fitting result comprises a plurality of linear fitting results, obtain a dynamic blanking point of the target frame image by intersecting the plurality of linear fitting results. The blanking point determination device further comprises a pixel point determination module. The pixel point determination module is configured to, in a case where there are continuous multiple frames of images before the target frame image, establish a dynamic blanking point set by using dynamic blanking points of the continuous multiple frames of images acquired before the target frame image, determine an average blanking point corresponding to the target frame image by using mean coordinates of the dynamic blanking point set, process the dynamic blanking point set by using a time series model to obtain an extrapolated blanking point corresponding to the target frame image, and determine the preset pixel point according to the average blanking point or the extrapolated blanking point; and in a case where there are no continuous multiple frames of images before the target frame image, acquire an image acquisition parameter of the target frame image, determine a static blanking point of the target frame image by using the image acquisition parameter, and determine the preset pixel point according to the static blanking point.

6. The apparatus of claim 5, wherein, The blanking point determination module is further configured to determine an interframe distance corresponding to the target frame image according to image frames corresponding to the dynamic blanking point set; in response to the interframe distance being greater than a first preset threshold, use the average blanking point as the dynamic blanking point of the target frame image; and in response to the interframe distance being less than the first preset threshold, use the extrapolated blanking point as the dynamic blanking point of the target frame image.

7. The apparatus of any one of claims 5-6, wherein, The result screening module is further configured to determine a first fitting result adjacent to a center line of the target frame image from the fitting result, acquire an intersection of the first fitting result and a second fitting result other than the first fitting result in the fitting result, in response to a coordinate difference between the intersection and the preset pixel point being less than a second preset threshold, use the intersection as a target intersection, and screen the target fitting result corresponding to the target intersection from the fitting result. The processor implements the steps of the method of any one of claims 1 to 4 when executing the computer program. The computer program, when executed by the processor, implements the steps of the method of any one of claims 1 to 4.

8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that, The computer program, when executed by the processor, implements the steps of the method of any one of claims 1 to 4.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, ​ 10. A computer program product comprising a computer program, characterized in that, ​

Images