Automobile blind area early warning method and system based on image recognition and storage medium

Abstract

The invention discloses an automobile blind area early warning method and system based on image recognition and a storage medium, and relates to the technical field of automobile blind area recognition based on Internet security. The automobile blind area early warning method based on image recognition comprises the following steps of data acquisition, complete shielding evaluation, partial shielding detection and partial shielding evaluation. According to the method, the image data is obtained by performing image preprocessing on the blind area image, complete occlusion judgment is performed on the image data and the reference data interval, a first-level early warning measure is taken, the partial occlusion influence coefficient is obtained according to the image data and the reference image data, and whether partial occlusion evaluation is executed or not is judged according to the partial occlusion influence coefficient and the partial occlusion threshold value. And obtaining all partial shielding influence coefficients in a preset time period to obtain a partial shielding similarity coefficient, and judging whether to take a secondary early warning measure according to the partial shielding similarity coefficient and a similarity threshold value, thereby achieving the effect of improving the accuracy of the image-based automobile blind area early warning, and solving the problem of insufficient image analysis in the automobile blind area early warning in the prior art.

Description

Technical Field

[0001] The present invention relates to the technical field of automotive blind spot recognition based on Internet security, and particularly to an automotive blind spot warning method, system and storage medium based on image recognition. Background Art

[0002] With the increasing complexity of road traffic, blind spots have become an important hidden danger to driving safety. Traditional physical rearview mirrors have limited coverage and are difficult to effectively monitor the blind spots around the vehicle. Blind spot warning methods based on image recognition technology use cameras and artificial intelligence algorithms (such as object detection, deep learning) to achieve precise recognition and real-time alerts for obstacles in the blind spots. This system not only enhances the driver's perception ability but also provides technical support for intelligent transportation systems such as autonomous driving and ADAS (Advanced Driver Assistance Systems), promoting further improvement of road safety.

[0003] Existing automotive blind spot warning technologies based on image recognition mainly rely on cameras installed on the side or rear of the vehicle to collect image data around the vehicle in real time. Through image processing and object detection algorithms, the system can identify potential dangerous objects such as vehicles and pedestrians in the blind spots and improve the recognition accuracy by combining deep learning models. At the same time, computer vision technology is used to estimate the distance and orientation between the target object and the vehicle. When an obstacle is detected in the blind spot, the system will issue a warning to the driver in the form of sound, vision or vibration. The existing technology has been widely applied to Advanced Driver Assistance Systems (ADAS), significantly improving driving safety.

[0004] For example, a road detection and alarm system disclosed in a patent application with the publication number CN118212775A includes: detecting traffic scene data through a millimeter wave radar and a vision sensor, preprocessing the traffic scene data, training the preprocessed data with a complex road scene data set for deep learning, extracting features from the information after deep learning, predicting road traffic accidents and detecting automotive blind spots to obtain fusion features, iteratively classifying the fusion features to obtain a classification result, and analyzing the classification result to complete road traffic accident warning. The millimeter wave radar and the time sensor are used for time and space fusion to detect and identify road targets to ensure the consistency of the two sensors in time and space.

[0005] For example, the invention patent announcement of the intelligent blind area warning method, system and cloud platform for vehicles based on image recognition with the announcement number of CN117373248B includes: determining the vehicle road condition object space conversion matrix based on the vehicle road condition object cluster, and then being able to analyze the object space position characteristics between any two target vehicle road condition objects based on this vehicle road condition object space conversion matrix.

[0006] However, in the process of implementing the technical solution of the present invention in the embodiments of the present application, it is found that the above technology has at least the following technical problems:

[0007] In the prior art, due to the complex vehicle conditions and road conditions during the driving of the vehicle, and in extreme weather conditions (such as heavy rain, heavy snow, thick fog, etc.), the camera for collecting images is easily blocked, resulting in the problem of insufficient image analysis in vehicle blind area warning. Summary of the Invention

[0008] The embodiments of the present application provide an intelligent blind area warning method, system and storage medium for vehicles based on image recognition, solve the problem of insufficient image analysis in vehicle blind area warning in the prior art, and realize the improvement of the accuracy of vehicle blind area warning based on images.

[0009] The embodiments of the present application provide an intelligent blind area warning method for vehicles based on image recognition, including the following steps: S1, obtaining the blind area images during the driving of the vehicle within a preset time period through a blind area camera, and performing image preprocessing on the blind area images to obtain image data; S2, performing a complete occlusion judgment on the image data and the reference data interval obtained from the preset database and taking corresponding first-level warning measures; S3, obtaining the partial occlusion influence coefficient based on the image data and the reference image data, and judging whether to execute S4 based on the partial occlusion influence coefficient and the partial occlusion threshold, where the partial occlusion influence coefficient is used to quantify the possibility of the blind area camera being partially blocked; S4, obtaining the partial occlusion similarity coefficient based on the partial occlusion influence coefficient obtained within the preset time period, and judging whether to take corresponding second-level warning measures based on the partial occlusion similarity coefficient and the similarity threshold, where the partial occlusion similarity coefficient represents the probability of the blind area camera being continuously blocked.

[0010] Further, the specific process of the image preprocessing is as follows: extracting the feature points in the blind area image through a feature point detection algorithm and statistically obtaining the number of key feature points, extracting the texture features of the blind area image through local binary pattern and statistically obtaining the number of texture feature points; performing denoising processing and grayscale conversion on the blind area image to obtain a grayscale image; obtaining the basic image data from the grayscale image, where the basic image data includes the brightness mean, image contrast and the number of edge lines; the image data includes the basic image data, the number of key feature points and the number of texture feature points.

[0011] Further, the specific steps for completely occluding the determination of the image data and the reference data interval obtained from the preset database are as follows: A1. Obtain the reference data interval from the preset database, where the reference data interval includes a reference brightness interval and a reference contrast interval; A2. Compare the average brightness with the reference brightness interval. If the average brightness belongs to the reference brightness interval, record that the blind area camera is completely occluded; otherwise, execute A3; A3. Compare the image contrast with the reference contrast interval. If the image contrast belongs to the reference contrast interval, record that the blind area camera is completely occluded; otherwise, execute A4; A4. Determine whether the number of edge lines is 0. If the number of edge lines is 0, record that the blind area camera is completely occluded; otherwise, execute A5; A5. Determine whether the number of key feature points is 0. If the number of key feature points is 0, record that the blind area camera is completely occluded; otherwise, execute A6; A6. Determine whether the number of texture feature points is 0. If the number of texture feature points is 0, record that the blind area camera is completely occluded; otherwise, record that the blind area camera is not completely occluded.

[0012] Further, the specific content of the first-level warning measure is as follows: If the blind area camera is completely occluded, continue to obtain the image data within the next preset time period and perform a complete occlusion determination. If the blind area camera is still completely occluded, automatically switch to the radar blind area warning method; if the blind area camera is not completely occluded, perform a partial occlusion assessment.

[0013] Further, the specific process of obtaining the partial occlusion influence coefficient based on the image data and the reference image data and making a determination based on the partial occlusion influence coefficient and the partial occlusion threshold is as follows: Number the blind area images and obtain the image data, and obtain the partial occlusion influence coefficient based on the image data; Compare the partial occlusion influence coefficient with the partial occlusion threshold obtained from the preset database: If the partial occlusion influence coefficient is not less than the partial occlusion threshold, record that the blind area camera is not occluded; if the partial occlusion influence coefficient is less than the partial occlusion threshold, obtain all the partial occlusion influence coefficients within the preset time period to obtain the partial occlusion similarity coefficient.

[0014] Further, the process of obtaining the partial occlusion similarity coefficient is as follows: Obtain all the partial occlusion influence coefficients within the preset time period, perform a ratio operation on the partial occlusion influence coefficient and the previous partial occlusion influence coefficient to obtain the occlusion change ratio; Cumulatively sum all the occlusion change ratios within the preset time period to obtain the total occlusion change value, and perform a ratio operation on the total occlusion change value and the total number of blind area images to obtain the average occlusion change value, and obtain the partial occlusion similarity coefficient based on the average occlusion change value.

[0015] Further, the specific process of determining whether to take corresponding secondary warning measures based on the partial occlusion similarity coefficient and the similarity threshold is as follows: Compare the partial occlusion similarity coefficient with the similarity threshold obtained from the preset database. If the partial occlusion similarity coefficient is less than the similarity threshold, it is recorded that the blind area camera is not occluded. If the partial occlusion similarity coefficient is not less than the similarity threshold, it is recorded that the blind area camera is partially occluded, and secondary warning measures are taken. The secondary warning measures include dashboard warning, sound warning, occlusion detection, and vehicle networking warning. The dashboard warning is used to display the position of the occluded blind area camera on the dashboard of the vehicle. The occlusion detection means detecting the occlusion substance that occludes the blind area camera through a cleaning sensor and taking corresponding occlusion cleaning measures. The cleaning sensor includes a laser sensor, a humidity sensor, and a pressure sensor. The vehicle networking warning is used to prompt other vehicles to keep a distance from the risk vehicle. The risk vehicle refers to the vehicle with a partially occluded blind area camera.

[0016] Further, the vehicle networking warning further includes: transmitting the basic characteristics and position information of the risk vehicle to other vehicles through the vehicle networking and reminding other vehicles to keep a distance. The other vehicles are determined through the vehicle positioning system. When the risk vehicle conducts blind area occlusion warning interaction with other vehicles through the vehicle networking, the vehicle networking security communication method is enabled. The vehicle networking security communication method includes a transmission security method and an anti-invasion method. The transmission security method is used to ensure the data transmission security when the risk vehicle interacts with other vehicles. The anti-invasion method is used to cope with various network attacks.

[0017] The embodiment of the present application provides an image recognition-based vehicle blind area warning system. The image recognition-based vehicle blind area warning system includes: a data acquisition module, a complete occlusion evaluation module, a partial occlusion detection module, and a partial occlusion evaluation module. Among them, the complete occlusion evaluation module is used to perform a complete occlusion judgment on the image data and the reference data interval obtained from the preset database and take corresponding primary warning measures. The partial occlusion detection module is used to obtain a partial occlusion influence coefficient based on the image data and the reference image data, and determine whether to execute the partial occlusion evaluation module based on the partial occlusion influence coefficient and the partial occlusion threshold. The partial occlusion influence coefficient is used to quantify the possibility that the blind area camera is partially occluded. The partial occlusion evaluation module is used to obtain all the partial occlusion influence coefficients within a preset time period to obtain a partial occlusion similarity coefficient, and determine whether to take corresponding secondary warning measures based on the partial occlusion similarity coefficient and the similarity threshold. The partial occlusion similarity coefficient represents the probability that the blind area camera is continuously occluded.

[0018] The embodiment of the present application provides a computer-readable storage medium for storing a program, and the program implements the image recognition-based vehicle blind area warning method when executed by a processor.

[0019] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0020] 1. Image data is obtained by performing image preprocessing on the blind - area image. A complete occlusion judgment is made on the interval between the image data and the reference data interval, and a first - level early - warning measure is taken. The partial occlusion influence coefficient is obtained based on the image data and the reference image data, and it is judged whether to perform a partial occlusion assessment by comparing it with the partial occlusion threshold. The partial occlusion similarity coefficient is obtained by acquiring all the partial occlusion influence coefficients within a preset time period. Whether to take a second - level early - warning measure is judged based on this and the similarity threshold, thereby reducing the influence caused by the occlusion of the blind - area camera, and further improving the accuracy of vehicle blind - area warning based on images, effectively solving the problem of insufficient image analysis in vehicle blind - area warning in the prior art.

[0021] 2. The blind - area image is numbered and image data is obtained, and the partial occlusion influence coefficient is obtained based on the image data, thereby more comprehensively and objectively quantifying the possibility of partial occlusion of the blind - area camera, and further realizing a more accurate judgment of whether the blind - area camera is partially occluded.

[0022] 3. By acquiring all the partial occlusion influence coefficients within a preset time period, a ratio operation is performed on the partial occlusion influence coefficient and the previous partial occlusion influence coefficient to obtain the occlusion change ratio; the cumulative sum of all the occlusion change ratios within the preset time period is obtained to get the total occlusion change value, and the average occlusion change value is obtained through a ratio operation between the total occlusion change value and the total number of blind - area images. The partial occlusion similarity coefficient is obtained based on the average occlusion change value, thereby further judging whether there is partial occlusion of the blind - area camera, and further realizing the timely early - warning when the blind - area camera is partially occluded. Brief Description of the Drawings

[0023] Figure 1 It is a flowchart of the vehicle blind - area warning method based on image recognition provided by the embodiments of the present application;

[0024] Figure 2 It is a schematic diagram of the change of the partial occlusion influence coefficient provided by the embodiments of the present application;

[0025] Figure 3 It is a schematic structural diagram of the vehicle blind - area warning system based on image recognition provided by the embodiments of the present application. Detailed Embodiments

[0026] Embodiments of the present application provide a method, system, and storage medium for blind spot warning of vehicles based on image recognition, which solve the problem of insufficient image analysis in blind spot warning of vehicles in the prior art. A blind spot image is obtained through a blind spot camera, and the blind spot image is preprocessed to obtain image data; a complete occlusion judgment is made between the image data and the reference data interval, and corresponding first-level warning measures are taken; the blind spot images are numbered, and the image data is obtained. Based on the image data, a partial occlusion influence coefficient is obtained. Whether to perform a partial occlusion evaluation is judged based on the partial occlusion influence coefficient and the partial occlusion threshold. By obtaining all the partial occlusion influence coefficients within a preset time period, a ratio operation is performed on the partial occlusion influence coefficient and the previous partial occlusion influence coefficient to obtain an occlusion change ratio; the occlusion change ratios within the preset time period are cumulatively summed to obtain a total occlusion change value, and the average occlusion change value is obtained through a ratio operation between the total occlusion change value and the total number of blind spot images. Based on the average occlusion change value, a partial occlusion similarity coefficient is obtained, and whether to take corresponding second-level warning measures is judged based on the partial occlusion similarity coefficient and the similarity threshold, thereby improving the accuracy of blind spot warning of vehicles based on images.

[0027] The technical solution in the embodiments of the present application is to solve the problem of insufficient image analysis in the above-mentioned vehicle blind spot warning. The general idea is as follows:

[0028] A complete occlusion judgment is made between the image data and the reference data interval, and first-level warning measures are taken. Based on the image data and the reference image data, a partial occlusion influence coefficient is obtained, and whether to perform a partial occlusion evaluation is judged accordingly. By obtaining all the partial occlusion influence coefficients within a preset time period, a partial occlusion similarity coefficient is obtained to judge whether to take second-level warning measures, achieving the effect of improving the accuracy of vehicle blind spot warning based on images.

[0029] To better understand the above technical solution, the above technical solution will be described in detail below in combination with the accompanying drawings of the specification and specific implementation manners.

[0030] Such as Figure 1As shown in the figure, it is a flowchart of the vehicle blind spot warning method based on image recognition provided by the embodiment of the present application. The method includes the following steps: S1, obtaining blind spot images during the driving process of the vehicle within a preset time period through a blind spot camera, and performing image preprocessing on the blind spot images to obtain image data; S2, performing a complete occlusion judgment on the image data and a reference data interval obtained from a preset database and taking corresponding first-level warning measures. The first-level warning measures are used to eliminate the impact of complete occlusion on vehicle blind spot warning; S3, obtaining a partial occlusion influence coefficient based on the image data and reference image data, and judging whether to execute S4 based on the partial occlusion influence coefficient and a partial occlusion threshold. The partial occlusion influence coefficient is used to quantify the possibility that the blind spot camera is partially occluded; S4, obtaining a partial occlusion similarity coefficient based on the partial occlusion influence coefficients obtained within a preset time period, and judging whether to take corresponding second-level warning measures based on the partial occlusion similarity coefficient and a similarity threshold. The partial occlusion similarity coefficient represents the probability that the blind spot camera is continuously occluded.

[0031] In this embodiment, the preset time period represents the duration of a data acquisition cycle of the blind spot warning method, such as 10 minutes, 30 minutes, or 60 minutes; the blind spot camera is usually deployed below the left and right side mirrors of the vehicle to monitor the blind spots on the sides of the vehicle. In addition, some vehicles deploy cameras in the front bumper area to monitor the blind spots in front of the vehicle. For example, in the BMW X5 model of BMW, a high-definition camera is installed on the front bumper, which works together with the rear camera and the side cameras of the vehicle to form a bird's-eye view; by distinguishing between complete occlusion and partial occlusion of the blind spot images, it is possible to effectively identify whether the field of view of the blind spot camera is occluded, reduce false alarms, improve the reliability of the warning system, and thus improve the accuracy of vehicle blind spot warning based on images.

[0032] Furthermore, the specific process of image preprocessing is as follows: extracting feature points in the blind spot image through a feature point detection algorithm and counting to obtain the number of key feature points, extracting the texture features of the blind spot image through local binary patterns and counting to obtain the number of texture feature points. The feature point detection algorithm is used to obtain the number of key feature points, and local binary patterns are used to obtain the number of texture feature points; performing denoising processing and grayscale conversion on the blind spot image to obtain a grayscale image. The denoising processing is used to reduce the noise of the blind spot image, and the grayscale processing is used to convert the color image into a grayscale image; obtaining image basic data from the grayscale image. The image basic data includes the brightness mean, image contrast, and the number of edge lines. The brightness mean represents the mean pixel value of the grayscale image, the image contrast represents the standard deviation of the pixel values of the grayscale image, and the number of edge lines is obtained by extracting the edges of the grayscale image through an edge detection algorithm and counting the number of non-zero pixel edges; the image data includes the image basic data, the number of key feature points, and the number of texture feature points.

[0033] In this embodiment, the pixel values of the grayscale image are read, all the pixel values of the grayscale image are processed as a two-dimensional array, and the average value of all the pixel values of the two-dimensional array is calculated to obtain the brightness mean value; it can extract and process the data in the blind area image more efficiently, improve the perception and processing ability of the vehicle for occlusion situations, and thus enhance the reliability and accuracy of the vehicle blind area warning system.

[0034] Local Binary Pattern (LBP) is a method for texture feature extraction. It generates a binary number by performing binary comparison on the neighborhood around each pixel, representing the relationship between the pixel and its neighboring pixels. The specific steps are as follows: Select a central pixel and consider its surrounding neighboring pixels (usually a 3x3 window); Compare the neighboring pixels with the central pixel. If the value of the neighboring pixel is greater than or equal to the central pixel, mark it as 1, otherwise mark it as 0; Convert the obtained binary number to a decimal number to form a new feature value, which is a texture feature point.

[0035] Feature point detection algorithms are used to extract significant points from images. The specific feature point detection algorithm is SURF (Speeded-Up Robust Features): SURF is an improved version of SIFT (Scale-Invariant Feature Transform), focusing on improving the calculation speed. It uses the Hessian matrix for fast feature point detection and adopts an approximate method to accelerate the calculation and matching process of feature points.

[0036] Denoising processing aims to reduce the noise in the image and improve the image quality. The specific denoising technique is Gaussian filtering, that is, using the Gaussian function to weight the neighboring pixels to reduce high-frequency noise while retaining the basic structure of the image.

[0037] Grayscale conversion is the process of converting a color image into a grayscale image. It converts the RGB (red, green, blue) values of each pixel into a single grayscale value. The specific method is the weighted average method, that is, according to the sensitivity of the human eye to different colors, different weights are used to calculate the grayscale value.

[0038] Further, the specific steps for completely occluding the judgment of the image data and the reference data interval obtained from the preset database are as follows: A1. Obtain the reference data interval from the preset database, where the reference data interval includes the reference brightness interval and the reference contrast interval; A2. Compare the average brightness with the reference brightness interval. If the average brightness belongs to the reference brightness interval, it is recorded that the blind area camera is completely occluded. Otherwise, execute A3; A3. Compare the image contrast with the reference contrast interval. If the image contrast belongs to the reference contrast interval, it is recorded that the blind area camera is completely occluded. Otherwise, execute A4; A4. Judge whether the number of edge lines is 0. If the number of edge lines is 0, it is recorded that the blind area camera is completely occluded. Otherwise, execute A5; A5. Judge whether the number of key feature points is 0. If the number of key feature points is 0, it is recorded that the blind area camera is completely occluded. Otherwise, execute A6; A6. Judge whether the number of texture feature points is 0. If the number of texture feature points is 0, it is recorded that the blind area camera is completely occluded. Otherwise, it is recorded that the blind area camera is not completely occluded.

[0039] In this embodiment, by sequentially checking the average brightness, image contrast, number of edge lines, number of key feature points, and number of texture feature points, it is ensured that multiple blind area image features are comprehensively analyzed, avoiding misjudgment caused by a single feature; improving the intelligence and fineness of the overall blind area warning, and being able to more accurately identify the complete occlusion of the blind area camera.

[0040] Specifically, the reference brightness interval is obtained from the preset database. In a specific embodiment, the reference brightness interval is obtained by comparing the average brightness interval, and the interval where the image appears black is [0, 50], and the interval where the image appears white is [231, 255]. Then the reference brightness interval is [0, 50] ∪ [231, 255].

[0041] Specifically, the reference contrast interval is obtained from the preset database. In a specific embodiment, the reference contrast interval is obtained by contrast comparison, and the interval reflecting that the difference between the bright and dark areas of the image is not obvious is [0, 50]. Then the reference contrast interval is [0, 50].

[0042] Further, the specific content of the first-level warning measure is as follows: If the blind area camera is completely occluded, continue to obtain the image data within the next preset time period and perform a complete occlusion judgment. If the blind area camera is still completely occluded, automatically switch to the radar blind area warning method and prompt the driver of the vehicle that the blind area warning method has been switched. The radar blind area warning method refers to the blind area warning method using radar for blind area detection; if the blind area camera is not completely occluded, perform a partial occlusion assessment.

[0043] In this embodiment, when the blind spot camera is completely blocked, the radar blind spot warning method is automatically switched to ensure the continuity and reliability of the blind spot detection function. Specifically, the radar blind spot warning method aims to identify and warn potential obstacles or other vehicles on the side and rear of the vehicle, ensuring that the driver of the vehicle can obtain sufficient safety information during operations such as lane change and overtaking; it ensures that even when the blind spot camera fails to work properly, the vehicle can still obtain warning information through the radar, and at the same time prompts the driver of the switching of the detection method, enhancing driving safety and the intelligent response ability of the system.

[0044] Furthermore, the specific process of judging based on the partial occlusion influence coefficient and the partial occlusion threshold according to the image data and the reference image data is as follows: Number the blind spot images and obtain the image data, and obtain the partial occlusion influence coefficient based on the image data; Compare the partial occlusion influence coefficient with the partial occlusion threshold obtained from the preset database: If the partial occlusion influence coefficient is not less than the partial occlusion threshold, it is recorded that the blind spot camera is not occluded; If the partial occlusion influence coefficient is less than the partial occlusion threshold, obtain the partial occlusion similarity coefficient by obtaining all the partial occlusion influence coefficients within the preset time period. The specific limiting expression of the partial occlusion influence coefficient is:

[0045]

[0046] In the formula, n represents the number of the blind spot image, n = 1, 2,..., N, N represents the total number of blind spot images, AB n represents the average brightness of the nth blind spot image, AB n-1 represents the average brightness of the (n - 1)th blind spot image, IC n represents the image contrast of the nth blind spot image, IC n-1 represents the image contrast of the (n - 1)th blind spot image, LN n represents the number of edge lines of the nth blind spot image, KP n represents the number of key feature points of the nth blind spot image, FP n represents the number of texture feature points of the nth blind spot image, PEC n represents the partial occlusion influence coefficient of the nth blind spot image, and e represents the natural constant.

[0047] In this embodiment, the algorithm comprehensively analyzes the average brightness, image contrast, number of edge lines, number of key feature points, and number of texture feature points to obtain the partial occlusion influence coefficient. In the algorithm, as the absolute value of the difference between the ratio of the average brightness to the previous average brightness and 1 increases, the partial occlusion influence coefficient also increases. Similarly, as the absolute value of the difference between the ratio of the image contrast to the previous image contrast and 1 increases, the partial occlusion influence coefficient also increases, indicating that there are image changes in the blind area image, and it is highly likely that the blind area camera is not occluded. In addition, the number of edge lines, number of key feature points, and number of texture feature points are positively correlated with the partial occlusion influence coefficient. As shown in Figure 2 which is a schematic diagram of the change of the partial occlusion influence coefficient provided in this embodiment. It can be seen from the figure that as the number of edge lines, number of key feature points, and number of texture feature points increase, the image shows an upward trend, indicating that the more data elements there are in the blind area image, the greater the probability that the blind area camera is not occluded. The algorithm provided in this embodiment effectively quantifies the probability that the blind area camera may be occluded as reflected by the blind area image, which is beneficial to timely detecting the condition of the blind area camera and avoiding traffic safety problems.

[0048] Specifically, the partial occlusion threshold is obtained from a preset database. In a specific embodiment, the partial occlusion threshold represents the average value of the corresponding data set obtained by substituting historical image data into the partial occlusion influence coefficient limit expression.

[0049] Furthermore, the process of obtaining the partial occlusion similarity coefficient is as follows: Obtain all the partial occlusion influence coefficients within a preset time period, perform a ratio operation on the partial occlusion influence coefficient and the previous partial occlusion influence coefficient to obtain the occlusion change ratio; Cumulatively sum all the occlusion change ratios within the preset time period to obtain the total occlusion change value, and perform a ratio operation on the total occlusion change value and the total number of blind area images to obtain the average occlusion change value. Based on the average occlusion change value, obtain the partial occlusion similarity coefficient. The specific limit expression of the partial occlusion similarity coefficient is:

[0050]

[0051] In the formula, n represents the number of the blind area image, n = 1, 2,..., N, N represents the total number of blind area images, PEC n represents the partial occlusion influence coefficient of the nth blind area image, PEC n+1 represents the partial occlusion influence coefficient of the (n + 1)th blind area image, and PSC represents the partial occlusion similarity coefficient.

[0052] In this embodiment, the algorithm comprehensively analyzes the total number of blind area images and the partial occlusion influence coefficient to obtain the partial occlusion similarity coefficient. For the sake of simplifying the analysis, let PJ be the average occlusion change value, and let FC is the variance occlusion change value. By combining the average occlusion change value and the variance occlusion change value, a partial occlusion similarity coefficient change table is obtained, as shown in Table 1 specifically:

[0053] Table 1 Partial Occlusion Similarity Coefficient Change Table

[0054] Average occlusion change value PJ Variance occlusion change value FC Partial occlusion similarity coefficient PSC 2.0 3.0 1.06 8.0 3.0 1.10 4.0 2.0 1.27 4.0 10.0 1.00 5.0 5.0 1.01 ...... ...... ......

[0055] It can be seen from Table 1 that when the variance occlusion change value remains unchanged, as the average occlusion change value increases, the value of the partial occlusion similarity coefficient also increases. For example, for the data in the first row and the second row, when the variance occlusion change value remains unchanged at 3.0, the average occlusion change value increases from 2.0 to 8.0, and the partial occlusion similarity coefficient also increases from 1.06 to 1.10. Similarly, when the average occlusion change value remains unchanged, as the variance occlusion change value increases, the value of the partial occlusion similarity coefficient decreases. For example, for the data in the third row and the fourth row, the average occlusion change value remains unchanged at 4.0, and the variance occlusion change value increases from 2.0 to 10.0, and the partial occlusion similarity coefficient decreases from 1.27 to 1.00. Thus, it can be known that when the average occlusion change value is larger and the variance occlusion change value is smaller, the partial occlusion similarity coefficient is larger, indicating that when it is detected that there may be partial occlusion of a blind area image, all partial occlusion influence coefficients within a preset time period are analyzed, avoiding false alarms caused by accident and effectively reducing the false alarm rate of vehicle blind area warning.

[0056] Furthermore, the specific process of determining whether to take corresponding secondary warning measures based on the partial occlusion similarity coefficient and the similarity threshold is as follows: The partial occlusion similarity coefficient is judged against the similarity threshold obtained from the preset database. If the partial occlusion similarity coefficient is less than the similarity threshold, it is recorded that the blind area camera is not occluded. If the partial occlusion similarity coefficient is not less than the similarity threshold, it is recorded that the blind area camera is partially occluded, and secondary warning measures are taken. The secondary warning measures include dashboard warning, sound warning, occlusion detection, and vehicle networking warning. The dashboard warning is used to display the position of the occluded blind area camera on the vehicle's dashboard. The sound warning means informing the vehicle driver through sound that the blind area camera is occluded and driving carefully. The occlusion detection means detecting the occlusion substance that occludes the blind area camera through a cleaning sensor and taking corresponding occlusion cleaning measures. The cleaning sensor includes a laser sensor, a humidity sensor, and a pressure sensor. The occlusion cleaning measures are used to remove the occlusion substance on the blind area camera. The vehicle networking warning is used to prompt other vehicles to keep a distance from the risk vehicle, and the risk vehicle refers to the vehicle with a partially occluded blind area camera.

[0057] In this embodiment, by introducing a judgment mechanism for the partial occlusion similarity coefficient and the similarity threshold, the vehicle can evaluate different degrees of occlusion, avoid overwarning triggered by momentary occlusion, thereby improving the accuracy and rationality of the warning. At the same time, a two-level warning measure of dashboard warning, sound warning and occlusion detection is adopted, providing an automatic feedback combining vision and audition to ensure that the vehicle driver can promptly detect the occlusion of the blind spot camera, effectively enhancing driving safety. In addition, combined with laser, humidity, and pressure sensors, the vehicle can automatically detect and identify the occluder on the blind spot camera and take corresponding cleaning measures, which reduces human intervention, improves the automation level of the system, and ensures clear camera vision at critical moments, enhancing the reliability and automation level of the image-based blind spot warning method.

[0058] Specifically, when the humidity sensor continuously detects the presence of water molecules on the blind spot camera within a preset time period, the heating mode is automatically enabled to eliminate the water on the blind spot camera. When the laser sensor detects clothing fibers or plastic products, a command to remove the occluding substance on the camera is sent to the vehicle driver to remind the vehicle driver to clean the obstacle. When the pressure sensor continuously detects pressure on the mirror surface of the blind spot camera within a preset time period, the cleaning mode is automatically started to remove the occluder attached to the mirror surface.

[0059] Specifically, the similarity threshold is obtained from a preset database. In a specific embodiment, the similarity threshold represents the average value of the corresponding data set obtained by substituting the partial occlusion influence coefficient obtained from historical image data into the partial occlusion similarity coefficient limit expression.

[0060] Furthermore, the vehicle networking warning further includes: transmitting the basic features and location information of the risk vehicle to other vehicles through the vehicle networking and reminding other vehicles to keep a distance, which is determined by the vehicle positioning system of the other vehicles. When the risk vehicle conducts blind spot occlusion warning interaction with other vehicles through the vehicle networking, the vehicle networking security communication method is enabled. The vehicle networking security communication method includes a transmission security method and an anti-invasion method. The transmission security method is used to ensure the transmission security of data when the risk vehicle interacts with other vehicles, and the anti-invasion method is used to cope with various network attacks.

[0061] In this embodiment, the basic features include the model, brand, and color of the vehicle. The other vehicles refer to the vehicles whose distance from the risk vehicle gradually becomes no greater than the braking distance as displayed by the vehicle positioning system. The braking distance is obtained through the following numerical expression:

[0062]

[0063] In the formula, V represents the vehicle speed of the risk vehicle, μ represents the friction coefficient of the road surface, g represents the acceleration due to gravity, and XD represents the braking distance; the vehicle networking warning method provided in this embodiment is conducive to ensuring data transmission security and preventing network attacks, thereby improving driving safety.

[0064] Specifically, the transmission security method uses Internet security technologies to ensure that the interactive data (basic features and location information of the risk vehicle) is not stolen or tampered with during transmission, such as the SSL (Secure Sockets Layer) / TLS (Transport Layer Security) encryption protocol; at the same time, a secure communication protocol is adopted to ensure the security of interactive data transmission. For example, the V2X (Vehicle to Everything) communication technology is used to ensure the security of the communication protocol, preventing hackers from interfering with the normal use of the vehicle blind spot warning method by forging or tampering with communication data; in addition, the anti-invasion method enables the vehicle blind spot warning method to have intrusion detection and defense capabilities to cope with various network attacks, such as deploying firewalls, intrusion detection systems (IDS, Intrusion Detection System), and intrusion prevention systems (IPS, Intrusion Prevention System) to detect and block existing Internet security threats in a timely manner.

[0065] As Figure 3 shown, it is a schematic structural diagram of the vehicle blind spot warning system based on image recognition provided in the embodiment of the present application. The vehicle blind spot warning system based on image recognition provided in the embodiment of the present application includes: a data acquisition module, a complete occlusion evaluation module, a partial occlusion detection module, and a partial occlusion evaluation module; among them, the data acquisition module is used to acquire blind spot images during the vehicle driving process within a preset time period through a blind spot camera, and perform image preprocessing on the blind spot images to obtain image data; the complete occlusion evaluation module is used to perform a complete occlusion judgment on the image data and the reference data interval obtained from the preset database and take corresponding first-level warning measures, and the first-level warning measures are used to eliminate the influence of complete occlusion on the vehicle blind spot warning; the partial occlusion detection module is used to obtain a partial occlusion influence coefficient based on the image data and the reference image data, and judge whether to execute the partial occlusion evaluation module based on the partial occlusion influence coefficient and the partial occlusion threshold, and the partial occlusion influence coefficient is used to quantify the possibility of the blind spot camera being partially occluded; the partial occlusion evaluation module is used to obtain a partial occlusion similarity coefficient based on the partial occlusion influence coefficient obtained within a preset time period, and judge whether to take corresponding second-level warning measures based on the partial occlusion similarity coefficient and the similarity threshold, and the partial occlusion similarity coefficient represents the possibility of the blind spot camera being continuously occluded.

[0066] In this embodiment, the classification measures of the primary warning and the secondary warning avoid over-warning and ensure timely response in case of severe occlusion. Through automated blind spot camera occlusion detection and warning, the system can perform real-time monitoring and processing without relying on the attention of the vehicle driver, reducing the workload of the vehicle driver and enhancing the driving experience.

[0067] Moreover, an embodiment of the present application further provides a computer-readable storage medium for storing a program, which when executed by a processor implements a vehicle blind spot warning method based on image recognition.

[0068] In this embodiment, by storing the program in the computer-readable storage medium, when the processor executes the program, automated blind spot detection and warning are achieved. When the program is executed by the processor, it can analyze image data in real time, continuously monitor the vehicle blind spot, reduce human intervention, automatically identify the occlusion of the blind spot camera, improve driving safety, and through the computer-readable storage medium, an efficient and intelligent blind spot monitoring and warning system is implemented, significantly enhancing the driving safety and convenience.

[0069] Specifically, in the hardware design of the readable storage medium, the physical ports of the readable storage medium (such as USB, Universal Serial Bus, debugging interface) need to be prevented from becoming an entry point for attackers. For example, by sealing the physical ports of the readable storage medium and setting access permissions to prevent physical access. For example, the vehicle blind spot warning system allows unauthorized access to be blocked by locking the OBD-II (On-Board Diagnostics II) interface.

[0070] In summary, in the embodiment of the present application, image data is obtained by performing image preprocessing on the blind spot image, a full occlusion judgment is made on the interval between the image data and the reference data interval and primary warning measures are taken. The partial occlusion influence coefficient is obtained based on the image data and the reference image data, and it is judged whether to perform a partial occlusion evaluation by comparing it with the partial occlusion threshold. The partial occlusion similarity coefficient is obtained by acquiring all the partial occlusion influence coefficients within a preset time period, and it is judged whether to take secondary warning measures by comparing it with the similarity threshold, thereby reducing the impact caused by the occlusion of the blind spot camera, and further improving the accuracy of vehicle blind spot warning based on images, effectively solving the problem of insufficient image analysis in vehicle blind spot warning in the prior art.

[0071] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) that contain computer-usable program code.

[0072] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one or more of the processes or multiple processes and / or blocks Figure 1 one or more of the blocks or multiple blocks.

[0073] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that implement the functions specified in Figure 1 one or more of the processes or multiple processes and / or blocks Figure 1 one or more of the blocks or multiple blocks.

[0074] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in Figure 1 one or more of the processes or multiple processes and / or blocks Figure 1 one or more of the blocks or multiple blocks.

[0075] Although the preferred embodiments of the present invention have been described, those skilled in the art can make additional changes and modifications once they learn the basic creative concepts. Therefore, the appended claims are intended to be construed to include the preferred embodiments as well as all changes and modifications falling within the scope of the present invention.

[0076] Obviously, those skilled in the art can make various modifications and variations to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention also intends to include these modifications and variations.

Claims

1. A method for blind spot warning of a vehicle based on image recognition, characterized in that, Including the following steps: S1. Obtain blind area images during vehicle driving within a preset time period through a blind area camera, and perform image preprocessing on the blind area images to obtain image data; S2. Perform a complete occlusion judgment on the image data and a reference data range obtained from a preset database, and take corresponding first-level warning measures; S3. Obtain a partial occlusion influence coefficient based on the image data and reference image data, and judge whether to execute S4 based on the partial occlusion influence coefficient and a partial occlusion threshold. The partial occlusion influence coefficient is used to quantify the possibility of the blind area camera being partially occluded; S4. Obtain all partial occlusion influence coefficients within a preset time period to obtain a partial occlusion similarity coefficient, and judge whether to take corresponding second-level warning measures based on the partial occlusion similarity coefficient and a similarity threshold. The partial occlusion similarity coefficient represents the probability of the blind area camera being continuously occluded.

2. The method for warning of vehicle blind spots based on image recognition according to claim 1, wherein: The specific process of the image preprocessing is as follows: Extract feature points in the blind area image through a feature point detection algorithm and count to obtain the number of key feature points, and extract the texture features of the blind area image through local binary patterns and count to obtain the number of texture feature points; Perform denoising processing and grayscale conversion on the blind area image to obtain a grayscale image; Obtain image basic data from the grayscale image. The image basic data includes brightness mean, image contrast, and the number of edge lines; The image data includes image basic data, the number of key feature points, and the number of texture feature points.

3. The method for warning of vehicle blind spots based on image recognition according to claim 2, characterized in that: The specific steps of performing a complete occlusion judgment on the image data and a reference data range obtained from a preset database are as follows: A1. Obtain a reference data range from the preset database. The reference data range includes a reference brightness range and a reference contrast range; A2. Compare the brightness mean with the reference brightness range. If the brightness mean belongs to the reference brightness range, record that the blind area camera is completely occluded, otherwise execute A3; A3. Compare the image contrast with the reference contrast range. If the image contrast belongs to the reference contrast range, record that the blind area camera is completely occluded, otherwise execute A4; A4. Judge whether the number of edge lines is 0. If the number of edge lines is 0, record that the blind area camera is completely occluded, otherwise execute A5; A5. Judge whether the number of key feature points is 0. If the number of key feature points is 0, record that the blind area camera is completely occluded, otherwise execute A6; A6. Judge whether the number of texture feature points is 0. If the number of texture feature points is 0, record that the blind area camera is completely occluded, otherwise record that the blind area camera is not completely occluded.

4. The method for warning of blind spots of a vehicle based on image recognition according to claim 3, wherein: The specific content of the first-level warning measures is as follows: If the blind area camera is completely occluded, continue to obtain image data in the next preset time period and perform a complete occlusion judgment. If the blind area camera is still completely occluded, automatically switch to the radar blind area warning method; If the blind area camera is not completely occluded, perform a partial occlusion assessment.

5. The method for warning of vehicle blind spots based on image recognition according to claim 2, characterized in that: The specific process of obtaining a partial occlusion influence coefficient based on the image data and reference image data, and judging based on the partial occlusion influence coefficient and a partial occlusion threshold is as follows: Number the blind area images and obtain image data, and obtain a partial occlusion influence coefficient based on the image data; Compare the partial occlusion influence coefficient with the partial occlusion threshold obtained from the preset database: If the partial occlusion influence coefficient is not less than the partial occlusion threshold, it is recorded that the blind area camera is not occluded; If the partial occlusion influence coefficient is less than the partial occlusion threshold, obtain all the partial occlusion influence coefficients within the preset time period to obtain the partial occlusion similarity coefficient.

6. The method for warning of vehicle blind spots based on image recognition according to claim 5, characterized in that: The process of obtaining the partial occlusion similarity coefficient is as follows: Obtain all the partial occlusion influence coefficients within the preset time period, and perform a ratio operation on the partial occlusion influence coefficient and the previous partial occlusion influence coefficient to obtain the occlusion change ratio; Accumulatively sum all the occlusion change ratios within the preset time period to obtain the total occlusion change value, and perform a ratio operation on the total occlusion change value and the total number of blind area images to obtain the average occlusion change value, and obtain the partial occlusion similarity coefficient based on the average occlusion change value.

7. The method for warning of blind spots of a vehicle based on image recognition according to claim 6, wherein: The specific process of judging whether to take corresponding secondary warning measures based on the partial occlusion similarity coefficient and the similarity threshold is as follows: Judge the partial occlusion similarity coefficient with the similarity threshold obtained from the preset database: If the partial occlusion similarity coefficient is less than the similarity threshold, it is recorded that the blind area camera is not occluded; If the partial occlusion similarity coefficient is not less than the similarity threshold, it is recorded that the blind area camera is partially occluded, and secondary warning measures are taken. The secondary warning measures include dashboard warning, sound warning, occlusion detection, and vehicle networking warning; The dashboard warning is used to display the position of the occluded blind area camera on the dashboard of the vehicle; The occlusion detection means detecting the occlusion substance that occludes the blind area camera through a cleaning sensor and taking corresponding occlusion cleaning measures. The cleaning sensor includes a laser sensor, a humidity sensor, and a pressure sensor; The vehicle networking warning is used to prompt other vehicles to keep a distance from the risk vehicle. The windward vehicle refers to the vehicle with a partially occluded blind area camera.

8. The method for warning of vehicle blind spots based on image recognition according to claim 7, wherein: The vehicle networking warning also includes: Transmit the basic characteristics and position information of the risk vehicle to other vehicles through the vehicle networking and remind other vehicles to keep a distance. The other vehicles are determined through the vehicle positioning system; When the risk vehicle conducts blind area occlusion warning interaction with other vehicles through the vehicle networking, enable the vehicle networking security communication method. The vehicle networking security communication method includes a transmission security method and an anti-intrusion method. The transmission security method is used to ensure the data transmission security when the risk vehicle interacts with other vehicles, and the anti-intrusion method is used to cope with various network attacks.

9. Blind spot warning system for vehicles based on image recognition, characterized in that, Including: A data acquisition module, a complete occlusion evaluation module, a partial occlusion detection module, and a partial occlusion evaluation module; Among them, the data acquisition module is used to obtain the blind area images during the vehicle driving process within the preset time period through the blind area camera, and perform image preprocessing on the blind area images to obtain image data; The complete occlusion evaluation module is used to perform a complete occlusion judgment on the image data and the reference data interval obtained from the preset database and take corresponding primary warning measures; The partial occlusion detection module is used to obtain a partial occlusion influence coefficient based on the image data and the reference image data, and determine whether to execute the partial occlusion evaluation module based on the partial occlusion influence coefficient and the partial occlusion threshold. The partial occlusion influence coefficient is used to quantify the possibility that the blind area camera is partially occluded; The partial occlusion evaluation module is used to obtain all the partial occlusion influence coefficients within a preset time period to obtain a partial occlusion similarity coefficient, and determine whether to take corresponding secondary warning measures based on the partial occlusion similarity coefficient and the similarity threshold. The partial occlusion similarity coefficient represents the probability that the blind area camera is continuously occluded.

10. A computer-readable storage medium for storing a program, which when executed by a processor implements the method for warning of vehicle blind areas based on image recognition according to any one of claims 1 to 8.

Images