Optical flow point screening method and device, electronic equipment and storage medium

Abstract

The invention relates to the technical field of image processing, in particular to an optical flow point screening method and device, electronic equipment and a storage medium. Acquiring a current original image and acquiring a current optical flow point set corresponding to the current original image; for each current optical flow point in the current optical flow point set, searching a historical optical flow point corresponding to the current optical flow point, and determining a historical optical flow point set corresponding to the historical optical flow point and a historical original image; calculating a current statistical feature of each current optical flow point relative to the current original image; for the historical optical flow points corresponding to the current optical flow points, calculating historical statistical characteristics of the historical optical flow points relative to the historical original image; and screening each current optical flow point based on the current statistical characteristics and the historical statistical characteristics. Interference of factors such as illumination changes and low-texture areas can be effectively resisted, and therefore the accuracy of optical flow point screening is guaranteed. High-quality optical flow point data can be provided for subsequent work such as pose estimation, and the precision and effect of related work are improved.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and more specifically to optical flow point selection methods, apparatus, electronic devices, and storage media. Background Technology

[0002] Currently, in Simultaneous Localization and Mapping (SLAM) technology, accurate selection of optical flow points in images is a crucial step in ensuring the system's positioning accuracy and robustness. Existing SLAM schemes primarily rely on multi-stage collaborative methods for selecting optical flow points: first, preliminary screening is performed using corner detectors (such as FAST and Harris) to quickly extract candidate optical flow points with obvious local features, laying the foundation for subsequent processing; then, the RANSAC algorithm based on motion consistency is used to remove outliers, iteratively estimating the motion model and selecting interior points that conform to the model, reducing errors caused by abnormal motion or mismatches; finally, continuous screening based on tracking length is combined to retain optical flow points that can be stably tracked across multiple frames, further improving the reliability of the optical flow point set.

[0003] However, the traditional screening methods mentioned above still have many problems and shortcomings in practical applications, making it difficult to meet the high-precision requirements of SLAM systems in complex environments. The main issues are: erroneous matching is easily generated in low-texture regions. Due to the lack of local image features, optical flow points in low-texture regions not only fail to fully reflect environmental information, but some optical flow points also only possess texture at certain image layers. As the image layer decreases, the local sub-image of the corresponding optical flow point exhibits extremely weak texture, leading to a significant decrease in the stability of optical flow tracking and consequently generating a large number of erroneous matching points.

[0004] Therefore, ensuring the accuracy of optical flow point selection has become an urgent problem to be solved. Summary of the Invention

[0005] In view of this, the present invention provides a method, apparatus, electronic device and storage medium for optical flow point screening, in order to solve the problem of how to ensure the accuracy of optical flow point screening.

[0006] In a first aspect, the present invention provides a method for filtering optical flow points, the method comprising: acquiring a current original image and acquiring a current set of optical flow points corresponding to the current original image; for each current optical flow point in the current set of optical flow points, finding historical optical flow points corresponding to the current optical flow points, a set of historical optical flow points corresponding to the historical optical flow points, and a historical original image; calculating the current statistical features of each current optical flow point relative to the current original image; for each historical optical flow point corresponding to the current optical flow point, calculating the historical statistical features of the historical optical flow points relative to the historical original image; and filtering each current optical flow point based on the current statistical features and the historical statistical features.

[0007] The optical flow point filtering method provided in this application obtains the current original image and the corresponding current optical flow point set, providing a clear analysis object for subsequent statistical feature calculation and filtering. This ensures that all operations revolve around the current original image and the current optical flow point set that actually need to be processed, avoiding interference from irrelevant data. It searches for historical optical flow points corresponding to the current optical flow point, as well as historical optical flow point sets and historical original images. This provides a comparable historical reference for the current optical flow point. By comparing with historical data, the stability and reliability of the current optical flow point can be judged more accurately, thus helping to identify unreliable optical flow points that only appear in the current frame and lack historical consistency, reducing the impact of transient erroneous optical flow points on subsequent processing. By calculating the current statistical features of each current optical flow point relative to the current original image, the local image features of the current optical flow point are quantified. This objectively reflects the texture, brightness, and other conditions of the area where the current optical flow point is located, providing a quantifiable basis for filtering. For each historical optical flow point corresponding to the current optical flow point, the historical statistical features of the historical optical flow point relative to the historical original image are calculated, accumulating feature data of the historical optical flow point in historical frames. This process reflects the historical characteristics of optical flow points over time, such as the stability of brightness and texture richness of the region where the point was located. Based on current and historical statistical characteristics, each current optical flow point is filtered. This allows for a more comprehensive evaluation of the quality of optical flow points. For example, the current statistical characteristics of a particular optical flow point may appear normal, but a significant difference is found when compared with historical statistical characteristics. This step identifies it as an abnormal optical flow point and filters it out. Furthermore, it effectively resists interference from factors such as changes in illumination and low-texture areas. After this filtering step, the remaining current optical flow points have good stability and reliability, providing high-quality optical flow point data for subsequent work such as pose estimation, thus improving the accuracy and effectiveness of related tasks.

[0008] In one optional implementation, calculating the current statistical features of each current optical flow point relative to the current original image includes: downsampling and smoothing the current original image to construct a current multi-layer image pyramid corresponding to the current original image; determining the current optical flow block corresponding to each current optical flow point; determining the positional relationship of the current optical flow block in each level of the current multi-layer image pyramid based on the positional information of the current optical flow point in the current optical flow point set; and calculating the current statistical features of the current optical flow point relative to the current multi-layer image pyramid corresponding to the current original image based on the positional relationship of the current optical flow block in each level of the current multi-layer image pyramid.

[0009] The optical flow point filtering method provided in this application downsamples and smooths the current original image to construct a multi-layer image pyramid corresponding to the current original image. Different levels in the multi-layer image pyramid correspond to different resolutions, allowing for macroscopic to microscopic analysis of optical flow point features. For example, higher levels reflect the overall features of a larger area where the optical flow point is located, while lower levels capture local details, solving the problem that a single-resolution image cannot adequately capture features at different scales. For each current optical flow point, a corresponding current optical flow block is determined. The current optical flow block limits the analysis scope to the key area surrounding the current optical flow point, avoiding interference from irrelevant pixels. This ensures that statistical feature calculations only revolve around effective image information near the current optical flow point, improving feature specificity. Based on the location information of the current optical flow point within the current optical flow point set, the positional relationship of the current optical flow block in each level of the current multi-layer image pyramid is determined. Through positional relationship calculations, the corresponding region of the current optical flow block can be accurately found in each level of the current multi-layer image pyramid. This ensures that the macroscopic features of higher levels and the microscopic features of lower levels both originate from the same spatial location of the optical flow point, avoiding analysis errors caused by feature misalignment. Based on the positional relationship of the current optical flow block within each level of the current multi-layer image pyramid, the current statistical features of the current optical flow point corresponding to the current multi-layer image pyramid of the original image are calculated. This allows for the capture of feature changes of the current optical flow point at different scales, ensuring the accuracy of the calculated statistical features corresponding to the current optical flow point.

[0010] In one optional implementation, based on the positional relationship of the current optical flow block in each level of the current multi-layer image pyramid, the current statistical features of the current optical flow point corresponding to the current original image in the current multi-layer image pyramid are calculated, including: based on the positional relationship of the current optical flow block in each level of the current multi-layer image pyramid, the current pixel mean value corresponding to the current optical flow block in each level of the current multi-layer image pyramid is calculated.

[0011] The optical flow point selection method provided in this application calculates the mean value of the current pixel corresponding to each level of the current optical flow block in the current multi-layer image pyramid based on the positional relationship of the current optical flow block in each level of the current multi-layer image pyramid. The mean value of the current pixel is an intuitive statistical feature, and its calculation method is simple and its physical meaning is clear (reflecting the average brightness of the region). This ensures the stability of the selection criteria. At the same time, the multi-level mean calculation utilizes the hierarchical correlation of the image pyramid to ensure that the mean values ​​of different levels correspond to the same optical flow point region in spatial location, reducing feature errors caused by positional deviations and providing more reliable input for subsequent optical flow point pose optimization.

[0012] In one optional implementation, the current statistical features include the average current pixel value of the current optical flow block at each level in the current multi-layer image pyramid; the current optical flow block is generated based on the current optical flow point, and the current multi-layer image pyramid is generated based on the current original image corresponding to the current optical flow point; the historical statistical features include the average historical pixel value of the historical optical flow block at each level in the historical multi-layer image pyramid corresponding to the historical optical flow block; the historical optical flow block is generated based on the historical optical flow point corresponding to the current optical flow point; the historical multi-layer image pyramid is generated based on the historical original image corresponding to the historical optical flow block; based on the current statistical features and the historical statistical features, each current optical flow point is filtered, including: for each current optical flow block, calculating the average current pixel value of the current optical flow block at each level and the average historical pixel value of the historical optical flow block corresponding to the current optical flow block at the same level to obtain the absolute difference of the average pixel value at each level; and filtering each current optical flow point based on the calculated absolute difference of the average pixel value at each level.

[0013] The optical flow point filtering method provided in this application calculates the absolute difference between the current pixel mean of the current optical flow block at each level and the historical pixel mean of the corresponding historical optical flow block at the same level for each current optical flow block. By calculating the absolute difference between the pixel mean of the current optical flow block and the historical optical flow block at the same level, this brightness difference can be quantified. This method directly addresses the technical pain point of "reduced reliability of optical flow points due to changes in illumination" in existing technologies. Compared with traditional methods that do not consider level differences, it can more accurately filter out erroneous optical flow points caused by dynamic light sources, alternating light and dark conditions, etc. Based on the calculated absolute difference between the pixel mean of each level, each current optical flow point is filtered. By analyzing the distribution of the absolute difference between the pixel mean of each level, mismatched points caused by dynamic light sources (such as vehicle lights) can be accurately identified, thus ensuring the accuracy of filtering each current optical flow point.

[0014] In one optional implementation, each current optical flow point is filtered based on the calculated absolute difference of the pixel mean values ​​corresponding to each level, including: generating a first set based on the absolute difference of the pixel mean values ​​corresponding to each level; calculating the range value in the first set; comparing the range value with a preset range value threshold; and deleting the current optical flow point corresponding to the current optical flow block if the range value is greater than the preset range value threshold.

[0015] The optical flow point selection method provided in this application generates a first set based on the absolute difference of the pixel mean values ​​corresponding to each level, avoiding the one-sidedness of the absolute difference of the pixel mean values ​​at a single level. The integration of multiple levels of differences can comprehensively reflect the brightness variation characteristics of optical flow points at different scales. For example, higher-level differences reflect the brightness stability of large areas, while lower-level differences reflect the brightness stability of local details, providing a comprehensive data foundation for subsequent range calculation. The range values ​​in the first set are calculated. The range value (the difference between the maximum and minimum values) can intuitively reflect the overall fluctuation degree of the absolute difference of the pixel mean values ​​at multiple levels. The range value is compared with a preset range threshold. This comparison process transforms the abstract "brightness fluctuation" into a clear "numerical comparison," avoiding subjective judgment errors, ensuring a unified selection standard for different optical flow points, and guaranteeing the fairness and stability of the selection. If the range value is greater than the preset range threshold, the current optical flow point corresponding to the current optical flow block is deleted. Current optical flow points with excessively large range values ​​inevitably have texture differences caused by dynamic light sources, and these points will interfere with the accuracy of pose estimation. The deletion operation directly removes it from the optical flow point set, solving the core problem of "the reliability of optical flow points decreasing due to changes in illumination".

[0016] In an optional implementation, the current statistical features further include the current pixel variance corresponding to each level of the current optical flow block in the current multi-layer image pyramid; the historical statistical features further include the historical pixel variance of historical optical flow blocks in each level of the historical multi-layer image pyramid corresponding to historical optical flow blocks. The method further includes: if the range value is less than or equal to a preset range value threshold, then normalizing the current pixel variance corresponding to each level of the current optical flow block in the current multi-layer image pyramid to obtain the current standard pixel variance corresponding to each level of the current optical flow block; comparing the current standard pixel variance corresponding to each level with the corresponding standard pixel variance threshold in descending order of level; wherein, the standard pixel variance thresholds corresponding to different levels are different; if there exists a target level corresponding to the current... If the standard pixel variance is less than the corresponding standard pixel variance threshold, the current optical flow block is deleted from the target level and the next lower level, resulting in the retained current optical flow blocks. The historical optical flow blocks corresponding to the retained current optical flow blocks are then obtained. The historical pixel variances of the historical optical flow blocks in each level of the historical multi-layer image pyramid corresponding to the historical optical flow blocks are normalized to obtain the historical standard pixel variances of the historical optical flow blocks corresponding to the current optical flow blocks at each level. The absolute difference in variance between the current standard pixel variance of the current optical flow blocks at the same level and the historical standard pixel variance of the historical optical flow blocks at the same level is calculated. Based on the absolute difference in variance at each level, each current optical flow point is filtered.

[0017] The optical flow point filtering method provided in this application embodiment, if the range value is less than or equal to a preset range value threshold, normalizes the current pixel variance corresponding to each level of the current optical flow block in the current multi-layer image pyramid to obtain the current standard pixel variance corresponding to each level of the current optical flow block. Following the order of levels from high to low, the current standard pixel variance corresponding to each level is compared with the corresponding standard pixel variance threshold; if the current standard pixel variance corresponding to a target level is less than the corresponding standard pixel variance threshold, the current optical flow block is deleted from the target level and the next lower level corresponding to the target level, resulting in the retained target current optical flow blocks; thus, current optical flow blocks with smaller textures can be deleted. The target historical optical flow blocks corresponding to the retained target current optical flow blocks are obtained. The historical pixel variance of the target historical optical flow block in each level of the historical multi-layer image pyramid corresponding to the target historical optical flow block is normalized to obtain the historical standard pixel variance corresponding to each level of the target historical optical flow block corresponding to the target current optical flow block. Normalization eliminates differences in brightness scale, making texture features in different brightness regions comparable. For example, after normalization, the standard variance of a high-brightness region with a large variance will significantly decrease if the actual texture is simple, facilitating subsequent comparison with historical data. The absolute difference in variance between the current standard pixel variance of the current optical flow block corresponding to the target at the same level and the historical standard pixel variance of the target's historical optical flow blocks is calculated. Comparison at the same level ensures the spatial scale consistency of texture analysis (e.g., high-level comparisons of large-area texture changes, low-level comparisons of local detail texture changes). The smaller the difference, the more consistent the texture features of the current optical flow point are with history, and the higher the reliability; conversely, a large difference may indicate texture anomalies due to texture disappearance, occlusion, etc. By calculating the difference at different levels, the specific level at which texture anomalies occur can be located. For example, if an optical flow point has a small difference at a high level (overall texture stability) but a large difference at a low level (loss of local detail texture), its texture at high resolution is unreliable and requires targeted screening. Based on the absolute difference in variance corresponding to each level, each target current optical flow point is screened. This avoids redundant calculations at invalid levels, improves screening efficiency, and ensures that the retained optical flow points have stable textures at all valid levels, providing high-quality input for pose estimation.

[0018] In one optional implementation, the current optical flow points of each target are filtered based on the absolute variance difference value corresponding to each level, including: comparing the absolute variance difference value corresponding to each level with the corresponding absolute variance difference value threshold in descending order of level; wherein, the absolute variance difference value threshold is different for different levels; if there is an absolute variance difference value corresponding to a target level that is greater than the corresponding absolute variance difference value threshold, then the current optical flow block of the target is deleted from the target level and the next lower level corresponding to the target level, so as to complete the filtering of the current optical flow points.

[0019] The optical flow point filtering method provided in this application compares the absolute variance difference of each level with its corresponding absolute variance difference threshold in descending order of level. The absolute variance difference threshold differs for different levels. Setting different thresholds for different levels matches the texture feature patterns of each level. For example, higher levels allow for larger variance differences (to avoid accidental deletion due to slight changes in illumination over a large area), while lower levels strictly control the differences (to ensure the reliability of local detail textures), solving the problem that a single threshold cannot adapt to multi-level analysis. If the absolute variance difference of a target level exceeds the corresponding absolute variance difference threshold, the current optical flow block of the target is deleted from the target level and the next lower level, thus completing the filtering of the current optical flow points of each target. If the variance difference of a certain target level (such as the 3rd level) exceeds the limit, it indicates that the texture of that level and lower levels (the 2nd and 1st levels) is unreliable (because lower-level details are built on the overall texture of higher levels). The deletion operation can remove all invalid optical flow points at once, preventing these points from participating in pose calculations and causing errors. This step, through layer-by-layer deletion, can thoroughly filter out optical flow points that appear normal only at some layers but whose actual underlying texture is invalid (e.g., low-texture areas appear as "pseudo-textures" in higher layers due to smoothing, but are deleted after the true weak texture in the lower layers is exposed), significantly reducing the false matching rate.

[0020] In a second aspect, the present invention provides an optical flow point screening device, the device comprising:

[0021] The acquisition module is used to acquire the current original image and the current optical flow point set corresponding to the current original image;

[0022] The search module is used to search for historical optical flow points corresponding to each current optical flow point in the current optical flow point set, as well as the historical optical flow point set and the historical original image.

[0023] The first calculation module is used to calculate the current statistical features of each current optical flow point relative to the current original image;

[0024] The second calculation module is used to calculate the historical statistical features of the historical optical flow points relative to the historical original image for each current optical flow point.

[0025] The filtering module is used to filter each current optical flow point based on current and historical statistical characteristics.

[0026] The optical flow point filtering device provided in this application acquires the current original image and the corresponding current optical flow point set, providing a clear analysis object for subsequent statistical feature calculation and filtering. This ensures that all operations revolve around the current original image and the current optical flow point set that actually need to be processed, avoiding interference from irrelevant data. It searches for historical optical flow points corresponding to the current optical flow point, as well as historical optical flow point sets and historical original images. This provides a comparable historical reference for the current optical flow point. By comparing with historical data, the stability and reliability of the current optical flow point can be judged more accurately, thus helping to identify unreliable optical flow points that only appear in the current frame and lack historical consistency, reducing the impact of transient erroneous optical flow points on subsequent processing. By calculating the current statistical features of each current optical flow point relative to the current original image, the local image features of the current optical flow point are quantified. This objectively reflects the texture, brightness, and other conditions of the area where the current optical flow point is located, providing a quantifiable basis for filtering. For each historical optical flow point corresponding to the current optical flow point, the historical statistical features of the historical optical flow point relative to the historical original image are calculated, accumulating feature data of the historical optical flow point in historical frames. This process reflects the historical characteristics of optical flow points over time, such as the stability of brightness and texture richness of the region where the point was located. Based on current and historical statistical characteristics, each current optical flow point is filtered. This allows for a more comprehensive evaluation of the quality of optical flow points. For example, the current statistical characteristics of a particular optical flow point may appear normal, but a significant difference is found when compared with historical statistical characteristics. This step identifies it as an abnormal optical flow point and filters it out. Furthermore, it effectively resists interference from factors such as changes in illumination and low-texture areas. After this filtering step, the remaining current optical flow points have good stability and reliability, providing high-quality optical flow point data for subsequent work such as pose estimation, thus improving the accuracy and effectiveness of related tasks.

[0027] Thirdly, the present invention provides an electronic device, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the computer instructions to perform the optical flow point screening method described in the first aspect or any corresponding embodiment thereof.

[0028] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to perform the optical flow point screening method described in the first aspect or any corresponding embodiment thereof.

[0029] Fifthly, the present invention provides a computer program product, including computer instructions for causing a computer to execute the optical flow point screening method described in the first aspect or any corresponding embodiment thereof. Attached Figure Description

[0030] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0031] Figure 1 This is a flowchart illustrating the optical flow point selection method according to an embodiment of the present invention;

[0032] Figure 2 (a) is a schematic diagram of a historical optical flow point detected in a historical original image according to an embodiment of the present invention;

[0033] Figure 2 (b) is the current original image according to an embodiment of the present invention. Figure 2 (a) is a schematic diagram of the current optical flow point corresponding to the historical optical flow point;

[0034] Figure 3 This is a flowchart illustrating another optical flow point screening method according to an embodiment of the present invention;

[0035] Figure 4 This is a schematic diagram of the image optical flow points and their corresponding local image blocks in the image pyramids of layers 1-4 according to an embodiment of the present invention;

[0036] Figure 5 This is a flowchart illustrating another optical flow point screening method according to an embodiment of the present invention;

[0037] Figure 6 According to an embodiment of the present invention Figure 2 (b) is a schematic diagram of the local sub-images corresponding to the anomalous optical flow points under different pyramid levels;

[0038] Figure 7 (a) is a schematic diagram of another historical optical flow point detected in a historical original image according to an embodiment of the present invention;

[0039] Figure 7 (b) is the image in the current original image according to an embodiment of the present invention. Figure 7 (a) is a schematic diagram of another current optical flow point corresponding to a historical optical flow point;

[0040] Figure 8 for Figure 7 A schematic diagram of the local sub-images of the candidate flow points and the historical flow points corresponding to the candidate flow points under different pyramid levels;

[0041] Figure 9(a) is a schematic diagram of a rehistorical optical flow point detected in a historical original image according to an embodiment of the present invention;

[0042] Figure 9 (b) is the current original image according to an embodiment of the present invention. Figure 9 (a) is a schematic diagram of the current optical flow point corresponding to the historical optical flow point;

[0043] Figure 10 This is a schematic diagram illustrating the comparison between the current standard pixel variance of the current optical flow block at different levels and the historical standard pixel variance of the historical optical flow block at different levels, according to an embodiment of the present invention.

[0044] Figure 11 This is a structural block diagram of an optical flow point screening device according to an embodiment of the present invention;

[0045] Figure 12 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0046] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0047] It should be noted that the optical flow point screening method provided in this application can be executed by an optical flow point screening device. This device can be implemented as part or all of an electronic device through software, hardware, or a combination of both. The electronic device can be a control device in the target vehicle or an electronic device independent of the target vehicle. In the following method embodiments, the execution subject is always described using an electronic device as an example.

[0048] According to an embodiment of the present invention, an embodiment of an optical flow point screening method is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0049] This embodiment provides an optical flow point selection method, which can be used in the aforementioned electronic device. Figure 1 This is a flowchart of an optical flow point selection method according to an embodiment of the present invention, such as... Figure 1 As shown, the process includes the following steps:

[0050] Step S101: Obtain the current original image and obtain the current optical flow point set corresponding to the current original image.

[0051] Specifically, electronic devices can capture the latest image using an onboard camera, which serves as the current raw image. The image captured by the camera must meet the imaging requirements of intelligent driving scenarios (such as clearly reflecting environmental textures and having moderate brightness) to provide basic visual data for subsequent optical flow point extraction.

[0052] Electronic devices can use a SLAM algorithm based on the fusion of LiDAR and vision to project the 3D coordinates (x, y, z) of the spatial point cloud acquired by the LiDAR onto the image coordinate system of the current original image, obtaining the image coordinates (u, v) of the optical flow points. The distribution and position information of these optical flow points on the current original image constitute the current optical flow point set. The projection process is achieved through camera intrinsic parameters (K) and coordinate transformation matrices between the LiDAR and the camera (rotation matrix R, translation matrix t), ensuring the accuracy of the optical flow points' positioning on the image. The specific projection process is as follows:

[0053]

[0054] Where ρ represents the depth information of the image points, K represents the intrinsic parameters of the camera, R represents the rotation matrix for transforming the LiDAR coordinate system to the camera coordinate system, and t represents the translation matrix for transforming the LiDAR coordinate system to the camera coordinate system. (x, y, z) represent the three-dimensional coordinates corresponding to the LiDAR point cloud data.

[0055] Step S102: For each current optical flow point in the current optical flow point set, find the historical optical flow point corresponding to the current optical flow point, the historical optical flow point set corresponding to the historical optical flow point, and the historical original image.

[0056] Specifically, the electronic device can traverse each current optical flow point in the current optical flow point set and retrieve matching historical optical flow points from a pre-maintained map based on their spatial coordinates (x, y, z). The map stores the spatial coordinates of optical flow points in historical frames and their corresponding relationships. It should be noted that the historical optical flow points corresponding to each current optical flow point can be optical flow points corresponding to the same moment or optical flow points corresponding to different moments. The historical moment corresponding to each historical optical flow point is within a preset duration before the current moment. This preset duration can be 1 second, 2 seconds, or other durations; this embodiment does not specifically limit the preset duration.

[0057] While retrieving historical optical flow points, the electronic device can extract the corresponding historical optical flow point set (the distribution of historical optical flow points on historical images) and the original historical image (the original visual image where the historical optical flow point is located) from the map. The original historical image and the current original image are correlated through a time series to ensure that both reflect the different states of the same spatial area at different times.

[0058] For example, such as Figure 2 The optical flow points corresponding to (a) and 2(b), where Figure 2 (a) is a historical optical flow point detected in the original historical image. Figure 2 (b) is the image in the current original image that is related to... Figure 2 The current optical flow point corresponding to the historical optical flow point in (a).

[0059] Step S103: Calculate the current statistical features of each current optical flow point relative to the current original image.

[0060] Specifically, the electronic device can calculate the current statistical characteristics of the current optical flow point relative to the current original image based on the position of each current optical flow point in the current original image.

[0061] Among them, the current statistical features can be the current pixel mean and current pixel variance of the pixels occupied by each current optical flow point in the current original image.

[0062] This step will be explained in detail below.

[0063] Step S104: For each current optical flow point, calculate the historical statistical features of the historical optical flow points relative to the historical original image.

[0064] Specifically, the electronic device can calculate the historical statistical characteristics of the historical optical flow points relative to the historical original image based on the position of each historical optical flow point in the corresponding historical original image.

[0065] Among them, historical statistical features can be the historical pixel mean and historical pixel variance of each historical optical flow point in the current original image.

[0066] This step will be explained in detail below.

[0067] Step S105: Based on the current statistical characteristics and historical statistical characteristics, each current optical flow point is filtered.

[0068] Specifically, the electronic device can compare the current statistical characteristics with historical statistical characteristics, and filter each current optical flow point based on the comparison results.

[0069] For example, if the difference between the current statistical feature and the historical statistical feature is within a preset range, the current optical flow point is retained. If the difference between the current statistical feature and the historical statistical feature is not within the preset range, the current optical flow point is deleted.

[0070] The optical flow point filtering method provided in this embodiment acquires the current original image and the corresponding current optical flow point set, providing a clear analysis object for subsequent statistical feature calculation and filtering. This ensures that all operations revolve around the current original image and the current optical flow point set that actually need to be processed, avoiding interference from irrelevant data. It searches for historical optical flow points corresponding to the current optical flow point, as well as historical optical flow point sets and historical original images. This provides a comparable historical reference for the current optical flow point. By comparing with historical data, the stability and reliability of the current optical flow point can be judged more accurately, thus helping to identify unreliable optical flow points that only appear in the current frame and lack historical consistency, reducing the impact of transient erroneous optical flow points on subsequent processing. By calculating the current statistical features of each current optical flow point relative to the current original image, the local image features of the current optical flow point are quantified. This objectively reflects the texture, brightness, and other conditions of the area where the current optical flow point is located, providing a quantifiable basis for filtering. For each historical optical flow point corresponding to the current optical flow point, the historical statistical features of the historical optical flow point relative to the historical original image are calculated, accumulating the feature data of the historical optical flow point in historical frames. This process reflects the historical characteristics of optical flow points over time, such as the stability of brightness and texture richness of the region where the point was located. Based on current and historical statistical characteristics, each current optical flow point is filtered. This allows for a more comprehensive evaluation of the quality of optical flow points. For example, the current statistical characteristics of a particular optical flow point may appear normal, but a significant difference is found when compared with historical statistical characteristics. This step identifies it as an abnormal optical flow point and filters it out. Furthermore, it effectively resists interference from factors such as changes in illumination and low-texture areas. After this filtering step, the remaining current optical flow points have good stability and reliability, providing high-quality optical flow point data for subsequent work such as pose estimation, thus improving the accuracy and effectiveness of related tasks.

[0071] This embodiment provides an optical flow point selection method, which can be used in the aforementioned electronic device. Figure 3 This is a flowchart of an optical flow point selection method according to an embodiment of the present invention, such as... Figure 3 As shown, the process includes the following steps:

[0072] Step S201: Obtain the current original image and obtain the current optical flow point set corresponding to the current original image.

[0073] Please refer to the above description of step S101 for details on this step, which will not be repeated here.

[0074] Step S202: For each current optical flow point in the current optical flow point set, find the historical optical flow point corresponding to the current optical flow point, the historical optical flow point set corresponding to the historical optical flow point, and the historical original image.

[0075] Please refer to the above description of step S102 for details on this step, which will not be repeated here.

[0076] Step S203: Calculate the current statistical features of each current optical flow point relative to the current original image.

[0077] Specifically, step S203 above may include the following steps:

[0078] Step S2031: Downsample and smooth the current original image to construct the current multi-layer image pyramid corresponding to the current original image.

[0079] The purpose of constructing the current multi-layer image pyramid is to analyze the characteristics of optical flow points at different scales. High-level images can reflect the overall characteristics of a larger area where the optical flow point is located, while low-level images can capture the detailed features around the optical flow point, providing a multi-scale data foundation for subsequent statistical feature calculations.

[0080] Specifically, the current multi-layer image pyramid is a hierarchical structure generated by repeatedly downsampling and smoothing the current original image. The lowest layer (layer 1) is the current original image, with the resolution of the original image (denoted as W×H), and has the highest image resolution. The higher layers (layer 2 and above) are generated by downsampling (reducing the size) and smoothing (reducing the jagged effect caused by downsampling, which can be achieved through OpenCV's resize function) the images of the layers below it.

[0081] Resolution calculation rule: The resolution of the l-th layer image (W) l ×H l Images are obtained by downsampling lower-level images; the higher the level, the lower the resolution. For example, higher-level images are obtained by scaling down lower-level images proportionally (e.g., the size is half that of the previous level after each downsampling) to achieve image representation at different scales.

[0082] For example, the resolution (W) of the corresponding l-th layer image l ×H l The formula is:

[0083]

[0084] For example, Figure 4 It refers to the image optical flow points and their corresponding local image blocks in the 1-4 layer image pyramid according to embodiments of the present invention. Figure 4It shows a current optical flow point with relatively rich texture and the corresponding current optical flow blocks under the current image pyramid at different levels, where the image resolution gradually increases from high to low level.

[0085] Step S2032: For each current optical flow point, determine the current optical flow block corresponding to the current optical flow point.

[0086] Specifically, for each current optical flow point, the electronic device can determine the current optical flow block corresponding to the current optical flow point according to preset rules.

[0087] For example, the electronic device determines the current optical flow block as an N×N pixel local image block centered on the current optical flow point (N is a preset side length, such as 5×5, 7×7, etc.) based on a preset rule, and uses it to focus on the local image features around the optical flow point.

[0088] Specifically, the electronic device can extract an image region of N×N pixels centered at the coordinates (u,v) of each current optical flow point in the current original image and each level of the current multi-layer image pyramid, as the current optical flow block for that current optical flow point at the corresponding level. For example, if the coordinates of the current optical flow point are (u,v), then the pixel range of the current optical flow block is from... arrive (Adjustments need to be made according to the image boundaries to avoid exceeding the image range).

[0089] By extracting the current optical flow block, the analysis scope is limited to the key area around the current optical flow point, avoiding interference from irrelevant pixels, so that subsequent statistical features (such as mean and variance) can accurately reflect the local image characteristics of the optical flow point.

[0090] Step S2033: Determine the positional relationship of the current optical flow block in each level of the current multi-layer image pyramid based on the positional information of the current optical flow point in the current optical flow point set.

[0091] Specifically, the coordinates (u,v) of the current optical flow point in the current optical flow point set correspond to the position in the current original image (the first layer of the pyramid). The higher-level images in the current multi-layer image pyramid are generated by downsampling of the lower layers. Therefore, the position of the optical flow point in the higher-level images can be obtained by scaling the coordinates.

[0092] Electronic devices can quickly determine the position of the current optical flow block at each level by utilizing the pointer positions of image pixels. For example, the electronic device can determine the pointer position of the top-left pixel of the current optical flow block based on the coordinates (u, v) of the current optical flow point in the current optical flow point set, the corresponding level l of the image pyramid, and the side length N of the current optical flow block:

[0093]

[0094] Where *patch is the pointer position of the top-left pixel of the current local sub-image, and *img is the pointer position of the top-left pixel of the entire image. This indicates rounding down, and cols represents the total number of columns in the image (image width).

[0095] Correspondingly, the pointer position *p of the i-th column and j-th row (i,j) in the current optical flow block is:

[0096] * p = *patch + (j × 2) l-1 ×cols)+i×2 l-1

[0097] Location mapping ensures that the current optical flow block corresponds to a local region of the same spatial location in each level of the multi-layer image pyramid, guaranteeing spatial consistency of optical flow blocks at different levels and providing an accurate location basis for subsequent cross-level statistical feature comparisons (such as the mean and variance of different levels).

[0098] Step S2034: Based on the positional relationship of the current optical flow block in each level of the current multi-layer image pyramid, calculate the current statistical features of the current optical flow point corresponding to the current multi-layer image pyramid of the current original image.

[0099] Specifically, step S2034 above may include the following steps:

[0100] Based on the positional relationship of the current optical flow block in each level of the current multi-layer image pyramid, calculate the mean value of the current pixel corresponding to each level of the current optical flow block in the current multi-layer image pyramid.

[0101] Specifically, the electronic device can determine all pixels contained in the current optical flow block in each level based on the positional relationship of the current optical flow block in each level (such as the coordinates of the top left corner located by the pointer and the side length N of the optical flow block). For example, in the l-th level, the pixel coordinate range of the current optical flow block is from (u1_start, v1_start) to (u1 start+N-1, v1 start+N-1) (where u1_start and v1 start are the column and row coordinates of the top left corner of the current optical flow block in this level).

[0102] Then, for the current optical flow block in the l-th level, the electronic device can traverse all N×N pixels it contains and obtain the pixel value I(i,j) of each pixel (i is the column index, j is the row index, and both range from 1 to N).

[0103] The average pixel value is calculated using the following formula:

[0104]

[0105] Where μl is the average pixel value of the current optical flow block at level l, which reflects the average brightness level of the optical flow block region at that level.

[0106] Following the above method, the electronic device sequentially calculates the pixel mean for the current optical flow block at each level (e.g., levels 1 to 4) of the current multi-layer image pyramid. The current pixel mean for each level is recorded, forming multi-layer brightness feature data for the optical flow block. This provides a quantitative basis for subsequent comparison with historical optical flow blocks and for filtering out brightness anomalies.

[0107] Step S204: For each current optical flow point, calculate the historical statistical features of the historical optical flow points relative to the historical original image.

[0108] Specifically, for each historical optical flow point, the electronic device downsamples and smooths the corresponding historical original image to construct a historical multi-layer image pyramid. Then, for each historical optical flow point, the corresponding historical optical flow block is determined. Based on the location information of the historical optical flow points in the historical optical flow point set, the positional relationship of the historical optical flow blocks in each level of the historical multi-layer image pyramid is determined.

[0109] Based on the positional relationship of historical optical flow blocks in each level of the historical multi-layer image pyramid, the historical optical flow points are calculated to correspond to the average historical pixels in each level of the historical multi-layer image pyramid corresponding to the original historical image.

[0110] For a detailed explanation, please refer to the above article, "Calculating the current statistical characteristics of the current optical flow point relative to the current original image."

[0111] Step S205: Based on the current statistical characteristics and historical statistical characteristics, each current optical flow point is filtered.

[0112] Please refer to the above description of step S105 for details on this step, which will not be repeated here.

[0113] The optical flow point filtering method provided in this application downsamples and smooths the current original image to construct a multi-layer image pyramid corresponding to the current original image. Different levels in the multi-layer image pyramid correspond to different resolutions, allowing for macroscopic to microscopic analysis of optical flow point features. For example, higher levels reflect the overall features of a larger area where the optical flow point is located, while lower levels capture local details, solving the problem that a single-resolution image cannot adequately capture features at different scales. For each current optical flow point, a corresponding current optical flow block is determined. The current optical flow block limits the analysis scope to the key area surrounding the current optical flow point, avoiding interference from irrelevant pixels. This ensures that statistical feature calculations only revolve around effective image information near the current optical flow point, improving feature specificity. Based on the location information of the current optical flow point within the current optical flow point set, the positional relationship of the current optical flow block in each level of the current multi-layer image pyramid is determined. Through positional relationship calculations, the corresponding region of the current optical flow block can be accurately found in each level of the current multi-layer image pyramid. This ensures that the macroscopic features of higher levels and the microscopic features of lower levels both originate from the same spatial location of the optical flow point, avoiding analysis errors caused by feature misalignment. Based on the positional relationship of the current optical flow block within each level of the current multi-layer image pyramid, the mean value of the current pixel corresponding to the current optical flow block at each level of the current multi-layer image pyramid is calculated. The mean value of the current pixel is an intuitive statistical feature, and its calculation method is simple and its physical meaning is clear (reflecting the average brightness of the region). This ensures the stability of the selection criteria. Simultaneously, the multi-level mean calculation utilizes the hierarchical correlation of the image pyramid, ensuring that the mean values ​​at different levels correspond to the same optical flow point region in spatial location, reducing feature errors caused by positional deviations, and providing more reliable input for subsequent optical flow point pose optimization. This allows for the capture of feature changes of the current optical flow point at different scales, ensuring the accuracy of the calculated statistical features corresponding to the current optical flow point.

[0114] This embodiment provides an optical flow point selection method, which can be used in the aforementioned electronic device. Figure 5 This is a flowchart of an optical flow point selection method according to an embodiment of the present invention, such as... Figure 5 As shown, the process includes the following steps:

[0115] Step S301: Obtain the current original image and obtain the current optical flow point set corresponding to the current original image.

[0116] Please refer to the above description of step S201 for details on this step, which will not be repeated here.

[0117] Step S302: For each current optical flow point in the current optical flow point set, find the historical optical flow point corresponding to the current optical flow point, the historical optical flow point set corresponding to the historical optical flow point, and the historical original image.

[0118] Please refer to the above description of step S202 for details on this step, which will not be repeated here.

[0119] Step S303: Calculate the current statistical features of each current optical flow point relative to the current original image.

[0120] Please refer to the above description of step S203 for details on this step, which will not be repeated here.

[0121] Step S304: For each current optical flow point, calculate the historical statistical features of the historical optical flow points relative to the historical original image.

[0122] Please refer to the above description of step S204 for details on this step, which will not be repeated here.

[0123] Step S305: Based on the current statistical characteristics and historical statistical characteristics, each current optical flow point is filtered.

[0124] The current statistical features include the average current pixel value of the current optical flow block at each level of the current multi-layer image pyramid; the current optical flow block is generated based on the current optical flow point, and the current multi-layer image pyramid is generated based on the current original image corresponding to the current optical flow point; the historical statistical features include the average historical pixel value of the historical optical flow block at each level of the historical multi-layer image pyramid corresponding to the historical optical flow block; the historical optical flow block is generated based on the historical optical flow point corresponding to the current optical flow point; and the historical multi-layer image pyramid is generated based on the historical original image corresponding to the historical optical flow block.

[0125] Specifically, step S305 above may include the following steps:

[0126] Step S3051: For each current optical flow block, calculate the average value of the current pixel in each level and the average value of the historical pixel in the same level of the historical optical flow block corresponding to the current optical flow block to obtain the absolute difference of the average pixel value in each level.

[0127] Specifically, for each current optical flow block, according to the hierarchical order of the current image pyramid and the historical image pyramid (e.g., from layer 1 to layer 4, or from high layer to low layer), the current pixel mean μ of the current layer l is calculated. curr,l The historical pixel average μ of the historical optical flow block at the same level l as the current optical flow block hist,l Perform pairing.

[0128] Then, for each pair of current pixel mean and historical pixel mean at the same level, the absolute difference is calculated using the following formula: Δμ l =∣μ curr,l -μ hist,l |. Where, Δμ l This represents the absolute difference in the average pixel value at level l, reflecting the brightness difference between the current and historical optical flow blocks at that level.

[0129] For example: if the current average pixel value of the second layer is 99.235 and the historical average pixel value of the second layer is 132.062, then the absolute difference Δμ2 of this layer is |99.235-132.062| = 32.827.

[0130] The electronic device traverses all levels of the current optical flow block in the current multi-layer image pyramid, repeats the above steps to calculate the absolute difference of the pixel mean for each level.

[0131] Step S3052: Based on the calculated absolute difference of the pixel mean values ​​corresponding to each level, filter each current optical flow point.

[0132] Specifically, step S3052 above may include the following steps:

[0133] Step a1: Generate the first set based on the absolute difference of the mean pixel values ​​corresponding to each level.

[0134] Specifically, the electronic device can integrate the calculated absolute differences of the pixel mean values ​​(Δμ1, Δμ2, Δμ3, Δμ4) corresponding to each level into a first set {Δμ} in the order of the levels, with the set form being {Δμ1, Δμ2, Δμ3, Δμ4}.

[0135] Step a2: Calculate the range values ​​in the first set.

[0136] Specifically, the range is the difference between the maximum and minimum values ​​in the first set, and the calculation formula is: range = max({Δμ}) - min({Δμ}).

[0137] For example, if the first set {Δμ} = {8.790, 19.000, 32.827, 41.938}, then the range value = 41.938 - 8.790 = 33.148.

[0138] Step a3: Compare the range value with the preset range threshold value.

[0139] Specifically, electronic devices can identify the current driving scenario type from the original image. The current driving scenario type includes enclosed areas with stable lighting and entrance / exit areas with alternating light and dark conditions. Enclosed areas with stable lighting, such as the interior of an underground parking lot or areas with fixed lighting and no direct natural light, have small light fluctuations and few abnormal light flow points. Entrance / exit areas with alternating light and dark conditions, such as passageways connecting parking lots to the outside world, are affected by alternating natural light (daytime) and streetlights (nighttime), resulting in frequent light changes and more abnormal light flow points.

[0140] Then, the electronic device can determine the initial range threshold based on the current driving scenario type. For example, by statistically analyzing the historical optical flow point brightness characteristic data (such as the range distribution of the absolute difference of the mean in multiple levels) of the two current driving scenario types, the basic threshold for the closed area with stable lighting is determined to be 15 (because the lighting is stable, a small brightness fluctuation is allowed), and the basic threshold for the entrance and exit area with alternating light and dark is 25 (because the lighting is variable, a larger normal fluctuation needs to be tolerated).

[0141] Next, the electronic device can obtain the headlight on / off status of the target vehicle via the target vehicle's CAN bus. When the headlights are detected to be on, the dynamic light source significantly affects the brightness of the surrounding light flow points, requiring increased filtering sensitivity. The electronic device can reduce the initial range threshold of the current area by a first preset reduction threshold to obtain a preset range threshold. For example, the first preset reduction threshold is 10%, or it could be 15%. This application embodiment does not specifically limit the first preset reduction threshold. For example, if the first preset reduction threshold is 10%, the threshold for the closed area is adjusted from 15 to 13.5, and the threshold for the entrance / exit area is adjusted from 25 to 22.

[0142] Optionally, the electronic device can also detect the overall brightness distribution or exposure parameters of the current original image through the camera sensor, and then determine whether the target vehicle is currently in a backlit scene based on the overall brightness distribution or exposure parameters. If it is currently in a backlit scene (such as an entrance / exit area with strong sunlight during the day), the probability of abnormal brightness increases due to sudden changes in illumination. Similarly, the electronic device can reduce the initial range threshold of the current area by a second preset threshold to obtain a preset range threshold, avoiding missing erroneous optical flow points caused by backlighting. For example, the second preset reduction threshold is 10%, or it could be 15%. This application embodiment does not specifically limit the second preset reduction threshold.

[0143] If the electronic device detects that the vehicle lights are on and the target vehicle is currently in a backlit scene, the electronic device will lower the initial range threshold by the first preset threshold, then lower the second preset threshold, and finally obtain the preset range threshold.

[0144] If the electronic device detects that the headlights are not on and the target vehicle is not currently in a backlit scene, the electronic device will set the initial range threshold as the preset range threshold.

[0145] Finally, the electronic device compares the range value with a preset range threshold.

[0146] Step a4: If the range value is greater than the preset range value threshold, then delete the current optical flow point corresponding to the current optical flow block.

[0147] Specifically, if the range value is greater than the preset range value threshold, the electronic device can delete the current optical flow point corresponding to the current optical flow block.

[0148] For example, when Figure 2 (b) If the range value corresponding to the current optical flow point is greater than the preset range value threshold, then it is determined that... Figure 2 The current optical flow point in (b) is relative to Figure 2 The historical optical flow points in (a) are anomalous optical flow points. For example... Figure 6 As shown, Figure 6 The middle section shows Figure 2 (b) shows the local sub-images corresponding to the anomalous optical flow points at different pyramid levels. Figure 6 In the corresponding local sub-images under the four different pyramid levels, the right sub-image of the local sub-image is the image corresponding to the current optical flow point in the current original image under the corresponding pyramid level, and the left sub-image is the image corresponding to the historical optical flow point in the historical original image under the corresponding pyramid level. For example, as shown in Table 1, it shows the pixel mean and the corresponding absolute difference of the mean when the current optical flow point is an abnormal optical flow.

[0149] Table 1 shows the pixel mean and absolute difference between the mean values ​​when the current optical flow point is an abnormal optical flow point.

[0150] Image layers Historical pixel average Current pixel average Absolute difference between mean 4 80.395 89.185 8.790 3 88.741 107.741 19.000 2 99.235 132.062 32.827 1 111.827 153.765 41.938 Maximum value \ \ 41.938 Minimum value \ \ 8.790 range \ \ 33.148

[0151] If the range value is less than or equal to the preset range value threshold, the electronic device can determine the current optical flow point corresponding to the current optical flow block as a candidate flow point.

[0152] For example, such as Figure 7 The image shows a matching diagram for candidate streamer points whose range values ​​are less than or equal to a preset range threshold. Figure 7 In, such as Figure 7 The optical flow points corresponding to (a) and 7(b), where Figure 7 (a) is another historical optical flow point detected in the original historical image. Figure 7 (b) is the image in the current original image that is related to... Figure 7 (a) corresponds to another current optical flow point.

[0153] like Figure 8 As shown, Figure 8 for Figure 7 The corresponding candidate flow points and their corresponding historical flow points are local sub-images of the pyramid at different levels. Figure 8 In the corresponding local sub-images under the four different pyramid levels, the right sub-image of the local sub-image is the image corresponding to the current optical flow point in the current original image under the corresponding pyramid level, and the left sub-image is the image corresponding to the historical optical flow point in the historical original image under the corresponding pyramid level. Table 2 shows... Figure 8 The information includes the mean pixel value and the absolute difference between the mean values ​​when the current optical flow point is a candidate optical flow point.

[0154] Table 2 shows the pixel mean and absolute difference between the current optical flow point and the corresponding optical flow point when the current optical flow point is a candidate optical flow point.

[0155]

[0156]

[0157] In an optional embodiment of this application, the current statistical features further include the current pixel variance corresponding to each level of the current optical flow block in the current multi-layer image pyramid; the historical statistical features further include the historical pixel variance of historical optical flow blocks in each level of the historical multi-layer image pyramid corresponding to the historical optical flow blocks; the above method further includes:

[0158] Step a5: If the range value is less than or equal to the preset range value threshold, then the current pixel variance corresponding to each level of the current optical flow block in the current multi-layer image pyramid is normalized to obtain the current standard pixel variance corresponding to each level of the current optical flow block.

[0159] Specifically, if the range value is less than or equal to a preset range value threshold, the electronic device can calculate the current pixel variance corresponding to each level of the current optical flow block in the current multi-layer image pyramid.

[0160] For example, an electronic device can calculate the current pixel variance based on the following formula:

[0161]

[0162] in, Let I(i,j) be the current pixel variance corresponding to the l-th level of the current multi-layer image pyramid, and let I(i,j) be the pixel value of the current optical flow block in the i-th column and j-th row.

[0163] Then, the electronic device can normalize the current pixel variance of the current optical flow block at each level in the current multi-layer image pyramid to obtain the current standard pixel variance of the current optical flow block at each level.

[0164] For example, electronic devices can be normalized based on the following formula:

[0165]

[0166] Among them, I max This is the maximum pixel value in the current optical flow block corresponding to the l-th level of the current multi-layer image pyramid.

[0167] Step a6: In descending order of level, compare the current standard pixel variance of each level with the corresponding standard pixel variance threshold.

[0168] The standard pixel variance thresholds are different for different levels.

[0169] Specifically, the electronic device can compare the current standard pixel variance of each level with the corresponding standard pixel variance threshold in descending order of level.

[0170] Step a7: If the current standard pixel variance corresponding to a target level is less than the corresponding standard pixel variance threshold, then the current optical flow block is deleted from the target level and the lower level corresponding to the target level, and the remaining current optical flow blocks of each target are obtained.

[0171] Specifically, the electronic device compares the current standard pixel variance corresponding to the current optical flow block in each level with the standard pixel variance threshold corresponding to that level. If the current standard pixel variance corresponding to a target level is less than the corresponding standard pixel variance threshold, the current optical flow block is deleted from the target level and the next lower level corresponding to the target level, resulting in the retained target current optical flow blocks.

[0172] Step a8: Obtain the target historical optical flow blocks corresponding to the current optical flow blocks of each target that are retained.

[0173] Specifically, the electronic device can find the target historical optical flow block at the corresponding level based on the level of the current multi-layer image pyramid where each target current optical flow block is located.

[0174] Step a9: Normalize the historical pixel variance of the target historical optical flow block in each level of the historical multi-layer image pyramid corresponding to the target historical optical flow block to obtain the historical standard pixel variance of the target historical optical flow block corresponding to the current target optical flow block in each level.

[0175] Specifically, the electronic device can normalize the historical pixel variance of the target historical optical flow block in each level of the historical multi-layer image pyramid corresponding to the target historical optical flow block, so as to obtain the historical standard pixel variance of the target historical optical flow block corresponding to the current target optical flow block in each level.

[0176] Step a10: Calculate the absolute difference of variance between the current standard pixel variance and the historical standard pixel variance corresponding to the same level.

[0177] Specifically, electronic devices can calculate the absolute difference in variance between the current standard pixel variance and the historical standard pixel variance corresponding to the same level.

[0178] For example, for each level l, the absolute difference between the current standard pixel variance and the historical standard pixel variance is calculated using the formula: Absolute difference of variance = |current standard pixel variance_l - historical standard pixel variance_l|.

[0179] Step a11: Filter the current optical flow points of each target based on the absolute difference of variance corresponding to each level.

[0180] Specifically, step a11 above may include the following steps:

[0181] Step a111: According to the order of hierarchy from high to low, compare the absolute difference of variance corresponding to each hierarchy with the corresponding absolute difference of variance threshold.

[0182] The absolute difference thresholds for variance vary for different levels, with higher levels having larger absolute difference thresholds.

[0183] Specifically, the electronic device can compare the absolute variance difference of each level with the corresponding absolute variance difference threshold in descending order of level.

[0184] Step a112: If there is a target level whose absolute variance difference is greater than the corresponding absolute variance difference threshold, then the target current optical flow block is deleted from the target level and the lower level corresponding to the target level, so as to complete the filtering of each current optical flow point.

[0185] Specifically, if the absolute difference of variance corresponding to a target level is greater than the corresponding absolute difference threshold, the electronic device will delete the current optical flow block of the target from the target level and the lower level corresponding to the target level, so as to complete the filtering of each current optical flow point.

[0186] For example, if the absolute difference of variance corresponding to the third level is greater than the corresponding absolute difference of variance threshold, the electronic device will delete the target current optical flow block from the third level and the second and first levels below the third level to complete the filtering of each current optical flow point.

[0187] For example, such as Figure 9 The image shows a matching diagram for abnormal optical flow points where the absolute difference of variance corresponding to the target level is greater than the corresponding absolute difference of variance threshold. Figure 9 In, such as Figure 9 (a) and the corresponding optical flow points in 9(b), where Figure 9 (a) is another historical optical flow point detected in the original historical image. Figure 9 (b) is the image in the current original image that is related to... Figure 9 (a) corresponds to another current optical flow point.

[0188] like Figure 10 The diagram illustrates a comparison between the current standard pixel variance of the current optical flow block at different levels and the historical standard pixel variance of historical optical flow blocks at different levels. Specifically, in... Figure 9 In the corresponding local sub-images under the four different pyramid levels, the right sub-image of the local sub-image represents the current standard pixel variance of the current optical flow point in the current original image under the corresponding pyramid level, and the left sub-image represents the historical standard pixel variance of the historical optical flow point in the historical original image under the corresponding pyramid level. Figure 10 As can be observed from the local sub-images, the optical flow point exhibits good texture features in layers 4, 3, and 2, while the texture is not rich in layer 1. The corresponding statistical parameters are shown in Table 3. Table 3 presents the standard pixel variance and texture effectiveness statistics when the current optical flow point is an anomalous optical flow point. Table 4 presents the standard pixel variance and texture effectiveness statistics when the current optical flow point is a normal optical flow point.

[0189] Table 3 shows the standard pixel variance and texture validity statistics when the current optical flow point is an anomalous optical flow point.

[0190] Image layers Historical standard pixel variance Current standard pixel variance Texture validity 4 <![CDATA[4.657×10 -3 ]]> <![CDATA[5.693×10 -3 ]]> yes 3 <![CDATA[5.770×10 -3 ]]> <![CDATA[6.239×10 -3 ]]> yes 2 <![CDATA[4.248×10 -3 ]]> <![CDATA[5.218×10 -3 ]]> yes 1 <![CDATA[1.223×10 -4 ]]> <![CDATA[9.473×10 -5 ]]> no

[0191] Table 4 shows the standard pixel variance and texture validity statistics when the current optical flow point is a normal optical flow point.

[0192] Image layers Historical standard pixel variance Current standard pixel variance Texture validity 4 <![CDATA[1.295×10 -2 ]]> <![CDATA[1.374×10 -2 ]]> yes 3 <![CDATA[1.177×10 -2 ]]> <![CDATA[1.777×10 -2 ]]> yes 2 <![CDATA[1.586×10 -2 ]]> <![CDATA[1.735×10 -2 ]]> yes 1 <![CDATA[1.453×10 -2 ]]> <![CDATA[1.567×10 -2 ]]> yes

[0193] A comparison of the normalized variances in Tables 3 and 4 reveals that the normalized variance is significantly smaller at layers with invalid textures. Therefore, an empirical threshold for the normalized variance can be used to filter out optical flow points with weak textures, and these points can be further removed from the corresponding image layers.

[0194] The optical flow point filtering method provided in this application calculates the absolute difference between the average pixel value of the current optical flow block at each level and the average pixel value of the historical optical flow block at the same level for each current optical flow block. By calculating the absolute difference between the average pixel value of the current optical flow block and the historical optical flow block at the same level, this brightness difference can be quantified. This method directly addresses the technical pain point of "the reliability of optical flow points decreasing due to changes in illumination" in the prior art. Compared with traditional methods that do not consider level differences, it can more accurately filter out erroneous matching optical flow points caused by dynamic light sources, alternating light and dark conditions, etc.

[0195] Then, based on the absolute difference of the pixel mean at each level, a first set is generated, avoiding the one-sidedness of a single-level pixel mean absolute difference. The integration of multiple levels of differences can fully reflect the brightness variation characteristics of optical flow points at different scales. For example, higher-level differences reflect the brightness stability of large areas, while lower-level differences reflect the brightness stability of local details, providing a comprehensive data foundation for subsequent range calculations. The range values ​​in the first set are calculated. The range value (the difference between the maximum and minimum values) can intuitively reflect the overall fluctuation of the absolute difference of the pixel mean at multiple levels. The range value is compared with a preset range threshold. This comparison process transforms the abstract "brightness fluctuation" into a clear "numerical comparison," avoiding subjective judgment errors, ensuring a unified selection standard for different optical flow points, and guaranteeing the fairness and stability of the selection. If the range value is greater than the preset range threshold, the current optical flow point corresponding to the current optical flow block is deleted. Current optical flow points with excessively large range values ​​inevitably have texture differences caused by dynamic light sources, and these points will interfere with the accuracy of pose estimation. The deletion operation directly removes it from the optical flow point set, solving the core problem of "the reliability of optical flow points decreasing due to changes in illumination".

[0196] If the range value is less than or equal to a preset range threshold, the current pixel variance of the current optical flow block at each level of the current multi-layer image pyramid is normalized to obtain the current standard pixel variance of the current optical flow block at each level. Following the order of levels from high to low, the current standard pixel variance of each level is compared with the corresponding standard pixel variance threshold. If the current standard pixel variance of the target level is less than the corresponding standard pixel variance threshold, the current optical flow block is deleted from the target level and the next lower level, resulting in the retained target current optical flow blocks. This allows for the removal of current optical flow blocks with smaller textures. The target historical optical flow blocks corresponding to the retained target current optical flow blocks are then obtained. The historical pixel variance of the target historical optical flow blocks at each level of the historical multi-layer image pyramid corresponding to the target historical optical flow blocks is normalized to obtain the historical standard pixel variance of the target historical optical flow blocks at each level. Normalization eliminates brightness scale differences, making the texture features of different brightness regions comparable. For example, after normalization, the standard variance of a high-brightness region with large variance will significantly decrease if the actual texture is simple, facilitating subsequent comparison with historical data. The absolute difference in variance between the current standard pixel variance of the current optical flow block corresponding to the target at the same level and the historical standard pixel variance of the target historical optical flow block is calculated. Comparison within the same level ensures the spatial scale consistency of texture analysis (e.g., high-level comparisons of large-area texture changes, low-level comparisons of local detail texture changes). The smaller the difference, the more consistent the texture features of the current optical flow point are with history, and the higher the reliability; conversely, a larger difference may indicate texture anomalies due to texture disappearance, occlusion, etc. By calculating the difference layer by layer, the specific level at which texture anomalies occur can be located. For example, if an optical flow point has a small difference at a high level (overall texture stability) but a large difference at a low level (loss of local detail texture), its texture at high resolution is unreliable and requires targeted screening. Finally, in descending order of level, the absolute difference in variance for each level is compared with the corresponding absolute difference threshold; the absolute difference thresholds for different levels are different. Different threshold settings correspond to different layers, matching the texture feature patterns of each layer. For example, higher layers allow for larger variance differences (avoiding accidental deletion due to slight changes in illumination over a large area), while lower layers strictly control the differences (ensuring the reliability of local detail textures), solving the problem that a single threshold cannot adapt to multi-layer analysis. If the absolute variance difference value corresponding to a target layer is greater than the corresponding absolute variance difference threshold, the current optical flow block of the target is deleted from the target layer and the layers below it, thus completing the screening of the current optical flow points of each target. If the variance difference value of a certain target layer (such as layer 3) exceeds the limit, it means that the texture of that layer and the lower layers (layers 2 and 1) are unreliable (because the details of lower layers are built on the overall texture of higher layers). The deletion operation can remove all optical flow points of invalid layers at once, preventing these points from participating in pose calculations and causing errors.This step, through hierarchical linkage deletion, can thoroughly filter out optical flow points that appear normal only at some levels but whose actual underlying texture has failed (such as low-texture areas displaying "pseudo-textures" in higher levels due to smoothing, but being deleted after the true weak texture in the lower level is exposed), significantly reducing the false matching rate.

[0197] This embodiment also provides an optical flow point screening device for implementing the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, hardware implementations, or a combination of software and hardware, are also possible and contemplated.

[0198] This embodiment provides an optical flow point screening device, such as... Figure 11 As shown, it includes:

[0199] The acquisition module 401 is used to acquire the current original image and the current optical flow point set corresponding to the current original image;

[0200] The lookup module 402 is used to look up the historical optical flow points corresponding to the current optical flow points, the historical optical flow point sets corresponding to the historical optical flow points, and the historical original images for each current optical flow point in the current optical flow point set.

[0201] The first calculation module 403 is used to calculate the current statistical features of each current optical flow point relative to the current original image;

[0202] The second calculation module 404 is used to calculate the historical statistical features of the historical optical flow points relative to the historical original image for each current optical flow point.

[0203] The filtering module 405 is used to filter each current optical flow point based on current statistical characteristics and historical statistical characteristics.

[0204] In some optional implementations, the first calculation module 403 is specifically used to downsample and smooth the current original image to construct the current multi-layer image pyramid corresponding to the current original image; for each current optical flow point, determine the current optical flow block corresponding to the current optical flow point; determine the positional relationship of the current optical flow block in each level of the current multi-layer image pyramid according to the positional information of the current optical flow point in the current optical flow point set; and calculate the current statistical features of the current optical flow point corresponding to the current multi-layer image pyramid corresponding to the current original image based on the positional relationship of the current optical flow block in each level of the current multi-layer image pyramid.

[0205] In some optional implementations, the first calculation module 403 is specifically used to calculate the average value of the current pixel corresponding to each level of the current optical flow block in the current multi-layer image pyramid based on the positional relationship of the current optical flow block in each level of the current multi-layer image pyramid.

[0206] In some optional implementations, the current statistical features include the average current pixel value of the current optical flow block at each level in the current multi-layer image pyramid; the current optical flow block is generated based on the current optical flow point, and the current multi-layer image pyramid is generated based on the current original image corresponding to the current optical flow point; the historical statistical features include the average historical pixel value of the historical optical flow block at each level in the historical multi-layer image pyramid corresponding to the historical optical flow block; the historical optical flow block is generated based on the historical optical flow point corresponding to the current optical flow point; the historical multi-layer image pyramid is generated based on the historical original image corresponding to the historical optical flow block; the filtering module 405 is specifically used to calculate, for each current optical flow block, the average current pixel value of the current optical flow block at each level and the average historical pixel value of the historical optical flow block corresponding to the current optical flow block at the same level, to obtain the absolute difference of the average pixel value at each level; and to filter each current optical flow point based on the calculated absolute difference of the average pixel value at each level.

[0207] In some optional implementations, the filtering module 405 is specifically used to generate a first set based on the absolute difference of the mean pixel values ​​corresponding to each level; calculate the range value in the first set; compare the range value with a preset range value threshold; and if the range value is greater than the preset range value threshold, delete the current optical flow point corresponding to the current optical flow block.

[0208] In some optional implementations, the current statistical features also include the current pixel variance corresponding to each level of the current optical flow block in the current multi-layer image pyramid; the historical statistical features also include the historical pixel variance of the historical optical flow block in each level of the historical multi-layer image pyramid corresponding to the historical optical flow block. The filtering module 405 is further configured to: normalize the current pixel variance corresponding to each level of the current optical flow block in the current multi-layer image pyramid if the range value is less than or equal to a preset range value threshold, to obtain the current standard pixel variance corresponding to each level of the current optical flow block; normalize the historical pixel variance of the historical optical flow block in each level of the historical multi-layer image pyramid corresponding to the historical optical flow block, to obtain the historical standard pixel variance corresponding to each level of the historical optical flow block corresponding to the current optical flow block; calculate the absolute difference of variance between the current standard pixel variance and the historical standard pixel variance corresponding to the same level; and filter each current optical flow point according to the absolute difference of variance corresponding to each level.

[0209] In some optional implementations, the filtering module 405 is specifically used to compare the absolute variance difference value corresponding to each level with the corresponding absolute variance difference threshold in descending order of level; wherein, the absolute variance difference threshold values ​​corresponding to different levels are different; if there is an absolute variance difference value corresponding to the target level that is greater than the corresponding absolute variance difference threshold value, then the current optical flow block is deleted from the target level and the next lower level corresponding to the target level, so as to complete the filtering of each current optical flow point.

[0210] Further functional descriptions of the above modules and units are the same as those in the corresponding embodiments described above, and will not be repeated here.

[0211] In this embodiment, the optical flow point screening device is presented in the form of a functional unit. Here, a unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0212] This invention also provides an electronic device having the above-described features. Figure 11 The optical flow point screening device shown.

[0213] Please see Figure 12 , Figure 12 This is a schematic diagram of the structure of an electronic device provided in an optional embodiment of the present invention, such as... Figure 12 As shown, the electronic device includes one or more processors 10, memory 20, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components communicate with each other via different buses and can be mounted on a common motherboard or otherwise as required. The processors can process instructions executed within the electronic device, including instructions stored in or on memory to display graphical information of a GUI on external input / output devices (such as display devices coupled to the interfaces). In some alternative implementations, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple electronic devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). Figure 12 Take a processor 10 as an example.

[0214] The memory 20 stores instructions executable by at least one processor 10 to cause at least one processor 10 to perform the method shown in the above embodiments.

[0215] The memory 20 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the electronic device. Furthermore, the memory 20 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, the memory 20 may optionally include memory remotely located relative to the processor 10, and these remote memories may be connected to the electronic device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0216] The memory 20 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, hard disk or solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0217] The electronic device also includes a communication interface 30 for communicating with other devices or communication networks.

[0218] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code, which, when accessed and executed by the computer, processor, or hardware, implements the methods shown in the above embodiments.

[0219] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0220] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for selecting optical flow points, characterized in that, The method includes: Obtain the current original image and obtain the current optical flow point set corresponding to the current original image; For each current optical flow point in the current optical flow point set, find the historical optical flow point corresponding to the current optical flow point, the historical optical flow point set corresponding to the historical optical flow point, and the historical original image. Calculate the current statistical features of each current optical flow point relative to the current original image; For each current optical flow point corresponding to a historical optical flow point, calculate the historical statistical features of the historical optical flow point relative to the historical original image; Based on the current statistical characteristics and the historical statistical characteristics, each current optical flow point is filtered.

2. The method according to claim 1, characterized in that, The calculation of the current statistical features of each current optical flow point relative to the current original image includes: The current original image is downsampled and smoothed to construct the current multi-layer image pyramid corresponding to the current original image; For each current optical flow point, determine the current optical flow block corresponding to the current optical flow point; Based on the position information of the current optical flow point in the current optical flow point set, determine the positional relationship of the current optical flow block in each level of the current multi-layer image pyramid; Based on the positional relationship of the current optical flow block in each level of the current multi-layer image pyramid, the current statistical features of the current optical flow point corresponding to the current multi-layer image pyramid of the current original image are calculated.

3. The method according to claim 2, characterized in that, The step of calculating the current statistical features of the current optical flow point corresponding to the current multi-layer image pyramid based on the positional relationship of the current optical flow block in each layer of the current multi-layer image pyramid includes: Based on the positional relationship of the current optical flow block in each level of the current multi-layer image pyramid, the mean value of the current pixel corresponding to the current optical flow block in each level of the current multi-layer image pyramid is calculated.

4. The method according to claim 1, characterized in that, The current statistical features include the average current pixel value of the current optical flow block at each level of the current multi-layer image pyramid; the current optical flow block is generated based on the current optical flow point, and the current multi-layer image pyramid is generated based on the current original image corresponding to the current optical flow point; the historical statistical features include the average historical pixel value of the historical optical flow block at each level of the historical multi-layer image pyramid corresponding to the historical optical flow block; the historical optical flow block is generated based on the historical optical flow point corresponding to the current optical flow point; the historical multi-layer image pyramid is generated based on the historical original image corresponding to the historical optical flow block; the filtering of each current optical flow point based on the current statistical features and the historical statistical features includes: For each current optical flow block, the mean value of the current pixel in each layer corresponding to the current optical flow block is calculated with the mean value of the historical pixel in the same layer corresponding to the historical optical flow block corresponding to the current optical flow block to obtain the absolute difference of the mean pixel value in each layer. Based on the calculated absolute difference of the pixel mean values ​​corresponding to each level, the current optical flow points are filtered.

5. The method according to claim 4, characterized in that, The step of filtering each current optical flow point based on the calculated absolute difference of the pixel mean values ​​corresponding to each level includes: A first set is generated based on the absolute difference of the mean pixel values ​​corresponding to each level; Calculate the range in the first set; The range value is compared with a preset range threshold. If the range value is greater than the preset range value threshold, then the current optical flow point corresponding to the current optical flow block is deleted.

6. The method according to claim 5, characterized in that, The current statistical features also include the current pixel variance of the current optical flow block at each level of the current multi-layer image pyramid; the historical statistical features also include the historical pixel variance of the historical optical flow block at each level of the historical multi-layer image pyramid corresponding to the historical optical flow block, and the method further includes: If the range value is less than or equal to the preset range value threshold, then the current pixel variance corresponding to each level of the current optical flow block in the current multi-layer image pyramid is normalized to obtain the current standard pixel variance corresponding to each level of the current optical flow block. The standard pixel variance corresponding to each level is compared with the corresponding standard pixel variance threshold in descending order of level; wherein the standard pixel variance threshold is different for different levels. If the variance of the current standard pixel corresponding to a target level is less than the corresponding standard pixel variance threshold, then the current optical flow block is deleted from the target level and the next lower level corresponding to the target level, and the remaining target current optical flow blocks are obtained. Obtain the target historical optical flow blocks corresponding to each of the retained target current optical flow blocks; The historical pixel variance of the target historical optical flow block in each level of the historical multi-layer image pyramid corresponding to the target historical optical flow block is normalized to obtain the historical standard pixel variance of the target historical optical flow block corresponding to the target current optical flow block in each level. Calculate the absolute difference of variance between the current standard pixel variance of the target current optical flow block and the historical standard pixel variance of the target historical optical flow block at the same level; Based on the absolute difference of variance corresponding to each level, the current optical flow points of each target are filtered.

7. The method according to claim 6, characterized in that, The step of filtering the current optical flow points of each target based on the absolute difference of variance corresponding to each level includes: The absolute variance difference value corresponding to each level is compared with the corresponding absolute variance difference threshold in descending order of level; wherein the absolute variance difference threshold is different for different levels. If the absolute difference of variance corresponding to a target level is greater than the corresponding absolute difference threshold, then the target current optical flow block is deleted from the target level and the lower level corresponding to the target level, so as to complete the filtering of each current optical flow point.

8. An optical flow point screening device, characterized in that, The device includes: The acquisition module is used to acquire the current original image and acquire the current optical flow point set corresponding to the current original image; The search module is used to search for historical optical flow points corresponding to each current optical flow point in the current optical flow point set, as well as historical optical flow point sets and historical original images. The first calculation module is used to calculate the current statistical features of each current optical flow point relative to the current original image; The second calculation module is used to calculate the historical statistical features of the historical optical flow points relative to the historical original image for each current optical flow point corresponding to the historical optical flow points; The filtering module is used to filter each current optical flow point based on the current statistical features and the historical statistical features.

9. An electronic device, characterized in that, include: A memory and a processor are communicatively connected, the memory stores computer instructions, and the processor executes the optical flow point screening method according to any one of claims 1 to 7 by executing the computer instructions.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing a computer to perform the optical flow point screening method according to any one of claims 1 to 7.

Images