A visual recognition system for automotive assisted driving

Abstract

The application relates to the technical field of automobile auxiliary driving, in particular to a visual identification system for automobile auxiliary driving, which comprises an image collecting device arranged outside an automobile and used for collecting road condition images shot by the automobile in real time, a processor connected with the image collecting device and a display device and used for receiving the road condition images sent by the image collecting device, performing image enhancement on the road condition images, obtaining enhanced road condition images, and sending the enhanced road condition images to the display device, and the display device arranged on an automobile instrument desk and facing a driver and used for receiving and displaying the enhanced road condition images so as to assist the driver in driving and improve the road condition identification accuracy and driving safety of the driver.

Description

Technical Field

[0001] This invention relates to the field of automotive driver assistance technology, and more particularly to a visual recognition system for automotive driver assistance. Background Technology

[0002] With the development of automotive driver assistance technology, visual perception systems have become one of the key technologies for realizing intelligent driving. Currently, mainstream radar and vision fusion systems employ multi-sensor fusion, specifically combining sensors including but not limited to high-resolution cameras, millimeter-wave radar, and lidar to achieve environmental perception and driver assistance functions, demonstrating a certain degree of reliability. However, this system also has some drawbacks:

[0003] 1. Its multi-sensor configuration leads to high hardware costs, and requires complex hardware interfaces and system integration solutions. In addition, high-precision sensors such as LiDAR are expensive and have high repair and maintenance costs, making it difficult for radar and vision fusion systems to be widely used in low- and mid-range vehicles.

[0004] 2. Radar and vision fusion systems need to process heterogeneous data from multiple sources simultaneously, including but not limited to distance information from radar and image data from cameras. These heterogeneous data typically have different sampling frequencies, resolutions, and error characteristics, leading to frequent information conflicts between multi-sensor data. For example, the distance to an object detected by radar may not match the position of an object identified by the camera, making it difficult for the system to correctly judge the road environment. Secondly, when fusing multi-sensor data, the fusion algorithm relies on a large amount of computing resources. Especially in real-time environments, high-performance processors are needed to complete data integration, filtering, and decision-making, increasing the computational burden and response time of the system. Therefore, complex time synchronization and spatial correction are required when fusing multi-sensor data.

[0005] 3. In severe weather conditions such as heavy rain, snow, and sandstorms, the performance of radar and cameras will be significantly affected. For example, radar signals may be interfered with by raindrops or snowflakes, and cameras are prone to overexposure or underexposure in scenes with drastic changes in light (such as tunnel entrances and exits). This can make the system prone to making incorrect judgments due to abnormal data, thereby increasing driving risks.

[0006] In the prior art, in order to overcome the shortcomings of radar and vision fusion systems, a pure vision assistance system is provided. This system only uses cameras to collect external images outside the vehicle during driving to achieve environmental perception and driving assistance functions. However, during driving, due to the influence of weather conditions, such as rain, snow, wind, sand and sunlight, the quality of the acquired external images will decrease. Moreover, processing high-resolution images requires a large amount of computing power, which makes it prone to processing delays in complex environments. This reduces the accuracy of road condition analysis and makes it impossible to provide effective driving assistance information in real time, thus affecting driving safety. Summary of the Invention

[0007] In view of this, embodiments of the present invention provide a visual recognition system for assisted driving of automobiles, in order to solve the problems of poor environmental adaptability and real-time performance of pure vision-assisted systems in the prior art.

[0008] This invention provides a visual recognition system for assisted driving in automobiles, the system comprising:

[0009] An image acquisition device, installed on the exterior of the vehicle, is used to acquire real-time images of road conditions taken by the vehicle.

[0010] The processor, connected to the image acquisition device and the display device, is used to receive the road condition image sent by the image acquisition device, perform image enhancement on the road condition image to obtain an enhanced road condition image, and send the enhanced road condition image to the display device.

[0011] A display device, mounted on the vehicle's dashboard and facing the driver, is used to receive and display the enhanced road condition images to assist the driver.

[0012] Further, the step of enhancing the road condition image to obtain an enhanced road condition image includes:

[0013] The road condition image is divided into blocks to obtain different road condition regions;

[0014] Different grayscale ranges are obtained based on the pixel count variation characteristics in the grayscale histogram of the road condition area; grayscale distribution feature values ​​are obtained based on the pixel count difference characteristics of different grayscale values ​​within the grayscale range; contrast feature values ​​are obtained based on the grayscale characteristics of the road condition area, the number characteristics of the grayscale range, and the grayscale distribution feature values.

[0015] Clustering is performed on the pixels within the road condition area to obtain two pixel clusters; texture confidence is obtained based on the gradient difference features of the pixel clusters and the distance features between pixels in the pixel clusters; enhancement coefficient is obtained based on the contrast feature value and the texture confidence; and adaptive contrast limit value is obtained based on the enhancement coefficient and the number of pixels with different gray values ​​in the road condition area.

[0016] The road condition region is enhanced using a contrast-limited adaptive histogram equalization algorithm based on the adaptive contrast limit value to obtain an enhanced road condition region; an enhanced road condition image is obtained based on all enhanced road condition regions.

[0017] Furthermore, the step of obtaining different grayscale ranges based on the pixel count variation characteristics in the grayscale histogram of the road condition area includes:

[0018] A curve is fitted based on the number of pixels corresponding to all gray levels in the gray-scale histogram to obtain a pixel count representation curve; the gray-scale range is constructed based on the gray-scale values ​​of the adjacent inflection points before and after any peak point in the pixel count representation curve.

[0019] Further, the step of obtaining grayscale distribution feature values ​​based on the difference in the number of pixels of different grayscale values ​​within the grayscale range includes:

[0020] Calculate the super-kurtosis of the number of pixels in the grayscale range and perform positive correlation mapping to obtain the grayscale distribution feature value of the grayscale range.

[0021] Further, the step of obtaining the contrast feature value based on the grayscale features of the road condition area, the quantity features of the grayscale range, and the grayscale distribution feature value includes:

[0022] Calculate the average value of grayscale distribution feature values ​​for all grayscale ranges in the road condition area to obtain the average grayscale distribution value; calculate the product of the number of grayscale value types and the number of grayscale ranges in the road condition area to obtain the first factor; calculate the product of the reciprocal of the average grayscale distribution value and the first factor to obtain the contrast feature value of the road condition area.

[0023] Furthermore, the step of clustering pixels based on gradient features within the road condition area to obtain two pixel clusters includes:

[0024] Based on the gradient values ​​of the pixels in the road condition area, the pixels are clustered using the K-means clustering algorithm to obtain two pixel clusters.

[0025] Further, the step of obtaining texture confidence based on the gradient difference features of the pixel clusters and the pixel distance features within the pixel clusters includes:

[0026] Calculate the average gradient value of the pixels in the pixel cluster to obtain the average gradient value; calculate the average Euclidean distance between all pixels and the nearest pixel in the pixel cluster corresponding to the maximum average gradient value and perform negative correlation mapping to obtain the neighboring feature value; calculate the product of the absolute value of the difference between the average gradient values ​​of the pixel cluster and the neighboring feature value to obtain the texture confidence of the road condition area.

[0027] Further, the step of obtaining the enhancement coefficient based on the contrast feature value and the texture confidence includes:

[0028] The enhancement coefficient of the road condition region is obtained by multiplying the reciprocal of the contrast feature value with the texture confidence.

[0029] Further, the step of obtaining the adaptive contrast limit value based on the enhancement coefficient and the number of pixels with different grayscale values ​​in the road condition area includes:

[0030] Calculate the ratio of the number of pixels in the road condition area to the maximum number of pixels in the grayscale histogram to obtain the pixel number difference value; calculate the product of the pixel number difference value and the enhancement coefficient and normalize it to obtain the adaptive contrast limit value of the road condition area.

[0031] The beneficial effects of the embodiments of the present invention compared with the prior art are as follows:

[0032] The road condition images acquired by the image acquisition device are enhanced by the processor, resulting in higher comparison accuracy and more obvious road condition features. The enhanced road condition images are then transmitted to the display device, which can improve the driver's accuracy in recognizing road conditions and improve driving safety. Attached Figure Description

[0033] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0034] Figure 1 This is a system block diagram of a vision recognition system for assisted driving of automobiles provided in Embodiment 1 of the present invention;

[0035] Figure 2This is an installation location diagram of a visual recognition display device for assisted driving of automobiles provided in Embodiment 1 of the present invention. Detailed Implementation

[0036] Embodiments of this disclosure are described in detail below, with examples of these embodiments illustrated in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this disclosure, and should not be construed as limiting it.

[0037] It should be noted that the terms "first," "second," etc., used in this disclosure and the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure.

[0038] To illustrate the technical solution of the present invention, specific embodiments are described below.

[0039] See Figure 1 This is a system block diagram of a vision recognition system for assisted driving of automobiles provided in Embodiment 1 of the present invention, as shown below. Figure 1 As shown, the system includes:

[0040] Image acquisition device S1 is installed outside the vehicle and is used to acquire real-time images of road conditions taken by the vehicle.

[0041] During vehicle operation, images of road conditions are captured in real time by an image acquisition device installed on the outside of the vehicle, including but not limited to a camera. Figure 2 The diagram shows the installation location of a visual recognition display device for automotive driver assistance. Figure 2 The part pointed to by the number 2 is an external camera on the vehicle. This diagram is only for illustrating the camera's installation location. The actual camera is not limited to the location shown; it can be installed on the hood, headlights, roof, bumper, etc., depending on the vehicle model. Because images captured in inclement weather have low clarity, they can easily affect the driver's and vehicle's judgment, reducing driving safety. Therefore, image enhancement is necessary to improve clarity and contrast.

[0042] The processor S2 is connected to the image acquisition device and the display device. It is used to receive road condition images sent by the image acquisition device, perform image enhancement on the road condition images to obtain enhanced road condition images, and send the enhanced road condition images to the display device.

[0043] The road condition images acquired by the image acquisition device are transmitted to the processor, where image enhancement processing is performed. Existing methods typically use a contrast-limited adaptive histogram enhancement algorithm to improve image contrast. This algorithm sets a contrast limit parameter during image enhancement to limit local enhancement and prevent over-enhancement. However, due to factors such as inclement weather, the clarity of images acquired at different times and locations may be affected to varying degrees while the vehicle is in motion. Using a fixed contrast limit parameter can lead to over-enhancement or poor enhancement in some images, affecting image quality and reducing the accuracy of image information recognition. Therefore, it is necessary to improve the contrast-limited adaptive histogram enhancement algorithm to improve image enhancement quality.

[0044] Furthermore, the contrast-limited adaptive histogram equalization divides an image into blocks for local image enhancement, improving enhancement quality. Therefore, the road condition image is divided into blocks to obtain different road condition regions. In this embodiment, the algorithm's default 8×8 block method is used to divide the image into different road condition regions. Since the display contrast of each road condition region may differ, the contrast limit value selected for enhancement may also differ. When the display contrast of a road condition region is low, a larger contrast limit value is needed to improve the enhancement effect; if the display contrast of a road condition region is high, a smaller contrast limit value is needed to avoid over-enhancement. The contrast of the road condition region can then be analyzed based on its grayscale histogram. When the number of pixels on either side of the peak point in the grayscale histogram exhibits a normal distribution, it indicates that the grayscale values ​​of most pixels in the road condition region are similar, lacking rich tonal variations and resulting in low contrast. Therefore, different grayscale ranges are obtained based on the pixel count variation characteristics in the grayscale histogram of the road condition region.

[0045] Preferably, in this embodiment of the invention, the step of obtaining different grayscale ranges includes: performing curve fitting based on the number of pixels corresponding to all grayscale levels in the grayscale histogram to obtain a pixel count representation curve; the pixel count representation curve represents the approximate number of pixels corresponding to different grayscale levels in the road condition area, mitigating the influence of small differences in the number of pixels corresponding to some adjacent grayscale values ​​in the grayscale histogram on the selection of the grayscale range. A grayscale range is constructed based on the grayscale values ​​of the inflection points adjacent to any peak point in the pixel count representation curve. The grayscale range is used for subsequent analysis of the distribution characteristics of the number of pixels at different grayscale levels, and each peak point in the pixel count representation curve corresponds to a grayscale range.

[0046] Furthermore, if the number of pixels corresponding to different gray levels within a grayscale range exhibits a more normal distribution, it means that a large number of pixels are concentrated in similar gray levels, resulting in lower contrast; conversely, if the number of pixels corresponding to different gray levels is more similar, the contrast is greater. Therefore, grayscale distribution feature values ​​are obtained based on the difference in the number of pixels at different gray levels within a grayscale range. Preferably, in this embodiment of the invention, the step of obtaining grayscale distribution feature values ​​includes: calculating the kurtosis of the number of pixels within a grayscale range and performing a positive correlation mapping to obtain the grayscale distribution feature values ​​for that grayscale range. It should be noted that kurtosis is a prior art technique. When the kurtosis is a constant of 0, it means that the distribution of the number of pixels within that grayscale range exhibits a normal distribution. If the kurtosis is larger, it means that the peak shape is sharper, resulting in more pixels having the same grayscale value, and thus lower contrast in the road condition image. If the kurtosis is less than 0, it means that the peak shape is flatter, resulting in more similar numbers of pixels corresponding to different gray levels, and thus higher contrast in the road condition image. Therefore, the larger the grayscale distribution characteristic value of the grayscale range, the lower the contrast.

[0047] The fewer the types of grayscale values ​​of pixels in a road condition area, the more concentrated the grayscale distribution range of pixels in that road condition area, and the lower the contrast. Conversely, the fewer the number of grayscale ranges obtained in a road condition area, the fewer the peak points, meaning the grayscale values ​​of pixels in that road condition area are more concentrated, and the lower the contrast. Furthermore, the larger the grayscale distribution characteristic values ​​of all grayscale ranges in a road condition area, the lower the contrast of that road condition area. A contrast characteristic value is then obtained based on the grayscale characteristics of the road condition area, the number of grayscale ranges, and the grayscale distribution characteristic values. Preferably, in this embodiment of the invention, the step of obtaining the contrast characteristic value includes: calculating the average value of the grayscale distribution characteristic values ​​of all grayscale ranges in the road condition area to obtain the average grayscale distribution value; the smaller the average grayscale distribution value, the higher the contrast of the road condition area. A first factor is obtained by calculating the product of the number of grayscale value types and the number of grayscale ranges in the road condition area; the larger the first factor, the more grayscale value types and grayscale ranges the road condition area has, and the higher the contrast of the road condition area. The contrast characteristic value of the road condition area is obtained by multiplying the reciprocal of the average grayscale distribution value by the first factor. A larger contrast characteristic value indicates higher contrast in the road condition area, requiring a smaller contrast limit value to avoid over-enhancement; conversely, a smaller contrast characteristic value indicates lower contrast in the road condition area, requiring a larger contrast limit value to increase the enhancement level. The formulas for obtaining the contrast characteristic value include:

[0048]

[0049] In the formula, D represents the contrast feature value of the road condition area, G represents the average gray level distribution, R represents the number of gray level types in the road condition area, H represents the number of gray level ranges, and R*H represents the first factor.

[0050] Furthermore, during vehicle movement, the camera captures a large amount of sky background data. If the road area is displayed as a solid-color sky or clouds, the obtained contrast feature value is small. However, such images lack effective information. Using a large contrast limit value would lead to over-enhancement, highlighting noise features and reducing image quality. Conversely, when the effective information in the road area is affected by clouds or light intensity, the obtained contrast feature value is also small, but a greater degree of enhancement is needed to highlight the effective information. Solid-color areas such as the sky background lack edge texture, while road areas affected by clouds and light intensity exhibit obvious edge texture. Therefore, clustering is performed based on the gradient features of pixels within the road area to obtain two pixel clusters. Preferably, in this embodiment, the step of obtaining two pixel clusters includes: clustering pixels using a K-means clustering algorithm based on the gradient values ​​of pixels in the road area to obtain two pixel clusters. It should be noted that the K-means clustering algorithm is existing technology, and the specific steps will not be elaborated further.

[0051] Furthermore, if there is a significant difference in the gradient values ​​of two pixel clusters, it means that there are solid background areas and near-field road condition information in the road condition region. The gradient values ​​of the solid background areas are smaller, while the gradient values ​​of the near-field road condition information are larger due to edge texture. If the pixels in a pixel cluster with a larger gradient value are closer together, it is more likely to indicate continuous edge texture rather than noise pixels, thus indicating the presence of valid information in the road condition region. Therefore, texture confidence is obtained based on the gradient difference characteristics of pixel clusters and the pixel distance characteristics within pixel clusters.

[0052] Preferably, in this embodiment of the invention, the step of obtaining texture confidence includes: calculating the average gradient value of pixels in a pixel cluster to obtain an average gradient value; calculating the average Euclidean distance between all pixels and their nearest pixel in the pixel cluster corresponding to the maximum average gradient value and performing negative correlation mapping to obtain adjacent feature values; the larger the adjacent feature value, the closer the pixels in the pixel cluster are, and the more likely it is to be edge texture, which better indicates the presence of near-field road condition information in the road condition area. Calculating the product of the absolute value of the difference between the average gradient values ​​of pixel clusters and the adjacent feature values ​​to obtain the texture confidence of the road condition area; the smaller the absolute value of the difference between the average gradient values ​​of two pixel clusters, the more likely the road condition area is to be a solid color background area with similar pixel gradient values; conversely, the larger the absolute value of the difference between the average gradient values, the greater the possibility that the road condition area represents different road condition obstacles, the more edge textures there are, and the larger the adjacent feature values, the better it further indicates the presence of edge textures. Therefore, the higher the texture confidence, the more likely there is valid road condition information in that road condition area, and the more the enhancement level should be increased; when the texture confidence is lower, the less likely there is valid road condition information in that road condition area, and the less the enhancement level should be increased.

[0053] Furthermore, the enhancement coefficient can be obtained based on the contrast feature value and texture confidence. Preferably, in this embodiment of the invention, the step of obtaining the enhancement coefficient includes: calculating the product of the reciprocal of the contrast feature value and the texture confidence to obtain the enhancement coefficient of the road condition area. The larger the enhancement coefficient, the larger the contrast limit value is required to improve the enhancement level and highlight the contrast of the road condition area.

[0054] Furthermore, if the value of the maximum peak point in the grayscale histogram of the road condition area is larger and the number of pixels is greater, the contrast limit value should be appropriately reduced to avoid over-enhancing the area and causing contrast imbalance or noise amplification. Reducing the contrast limit value parameter can limit the height of the peak in the grayscale histogram, crop the grayscale levels with excessively high pixel counts and redistribute them to other grayscale levels, thereby smoothing the histogram distribution and reducing the problem of excessively strong local contrast in overly bright or dark areas, making the equalization effect more natural, while preventing loss of detail or amplification of noise. Therefore, an adaptive contrast limit value is obtained based on the enhancement degree coefficient and the pixel count characteristics of different grayscale values ​​in the road condition area; preferably, in this embodiment of the invention, the step of obtaining the adaptive contrast limit value includes: calculating the ratio of the number of pixels in the road condition area to the maximum number of pixels in the grayscale histogram to obtain a pixel count difference value; the larger the pixel count difference value, the larger the value of the maximum peak point in the grayscale histogram and the greater the number of pixels, so the contrast limit value should be appropriately reduced. The product of the pixel count difference and the enhancement coefficient is calculated and normalized to obtain the adaptive contrast limit value for the road condition area; in this embodiment of the invention, the normalization function is: y represents the adaptive contrast limit value, e represents the natural constant, k is 2.9 in this embodiment of the invention, b is 1.4, and x is the independent variable to be normalized. A larger adaptive contrast limit value means that a greater degree of enhancement is needed to highlight the effective information in the road condition area; conversely, a smaller adaptive contrast limit value means that the contrast of the road condition area is greater, or there is no solid color area with effective information, so excessive enhancement is not necessary.

[0055] After obtaining the adaptive contrast limit value for each road condition area, the road condition area can be enhanced using a contrast-limited adaptive histogram equalization algorithm based on the adaptive contrast limit value, resulting in enhanced road condition areas. It should be noted that the contrast-limited adaptive histogram equalization algorithm is existing technology, and its specific steps will not be elaborated further. The enhanced road condition areas obtain adaptive contrast limit values ​​based on their own image information features, making the enhancement level more closely match the image features, effectively improving enhancement quality compared to fixed contrast limit parameters. Then, an enhanced road condition image is obtained from all enhanced road condition areas. This enhanced road condition image significantly enhances the contrast of different areas compared to the captured road condition image, allowing drivers to obtain more accurate and effective information.

[0056] Display device S3, located on the vehicle dashboard and facing the driver, is used to receive and display enhanced road condition images to assist the driver.

[0057] The enhanced road condition images are transmitted to the vehicle's dashboard and onto a display facing the driver. Figure 2The part pointed to by the number 1 is the display device on the dashboard inside the vehicle. This picture only shows one scenario where the display device is installed inside the car. In reality, depending on the car model, the display device can also be installed in the rearview mirror, the central control display screen, the driver's left front side, etc. During driving, the driver can more accurately judge the road conditions by using enhanced road condition images, thereby improving the driver's road condition recognition accuracy and driving safety.

[0058] To achieve the long-term development and wider application of this technical solution, the intelligent scene detection function will be continuously optimized and expanded based on the existing foundation, further enhancing the system's adaptability and safety in complex driving environments. Specifically, the overall technical solution will integrate more intelligent environmental perception capabilities, combining deep learning with continuous training on large-scale environmental data, enabling the system to intelligently identify and classify various driving scenarios, particularly in adverse weather conditions, strong light changes, and complex road conditions. Through this optimization, the system can automatically adjust its image processing strategy, thereby more effectively coping with the interference of various environmental factors on visual information.

[0059] The intelligent detection and classification system will be able to perform real-time analysis of various driving environments, including but not limited to different weather conditions (such as heavy rain, snow, fog, sandstorms, etc.), changes in light intensity (such as direct sunlight, sunset, or rapid changes in light at tunnel entrances and exits), and dynamic monitoring of external objects (such as pedestrians, other vehicles, obstacles, etc.). The system will utilize advanced object detection and classification technology to accurately identify these external factors and provide the driver with optimized visual feedback in a timely manner, thereby enhancing the driver's perception of complex environments.

[0060] This process will continuously optimize image processing algorithms through the ongoing training of deep learning models with a large amount of environmental data, improving their adaptability and recognition accuracy in different scenarios. With technological iteration, the system will be able to more intelligently identify various environmental changes and potential risks, automatically adjusting image output to ensure that drivers always obtain clear and accurate road information, whether in extreme weather, complex lighting conditions, or in heavy traffic.

[0061] Furthermore, the intelligent detection and classification capabilities of the technical solution involved in this invention will further expand application scenarios. In addition to conventional driving environments, the system will also demonstrate strong adaptability in nighttime driving, extreme weather conditions, complex road conditions (such as highways, bridges, and tunnels), and multi-lane traffic scenarios. Through real-time environmental perception and image enhancement, the system will effectively reduce the driver's visual burden, improve reaction speed, and significantly reduce judgment errors caused by environmental interference, thereby further enhancing driving safety.

[0062] To ensure continuous system optimization and long-term stability, future innovations will focus on data acquisition and processing capabilities. This will be achieved through enhanced deep learning training of multi-dimensional environmental data, further improving the system's intelligence level. As technology evolves, the system's algorithms will continue to evolve, enabling it to cope with more complex environmental changes and driving scenarios, gradually moving towards a fully adaptive intelligent driver assistance system.

[0063] In summary, the technical solution of this invention not only addresses the practical needs of the current environment but also maintains a forward-looking perspective on technological development. It plans to comprehensively improve the accuracy, intelligence, and safety of driving assistance systems by continuously optimizing intelligent detection, object classification, and image processing algorithms. The objective of this application is to provide drivers with a more accurate, efficient, and reliable assisted driving experience, while simultaneously promoting the application of intelligent driving technology in more scenarios, ultimately achieving broader traffic safety assurance.

[0064] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A visual recognition system for assisted driving in automobiles, characterized in that, The visual recognition system includes: An image acquisition device, installed on the exterior of the vehicle, is used to acquire real-time images of road conditions taken by the vehicle. The processor, connected to the image acquisition device and the display device, is used to receive the road condition image sent by the image acquisition device, perform image enhancement on the road condition image to obtain an enhanced road condition image, and send the enhanced road condition image to the display device. A display device, mounted on the vehicle's dashboard and facing the driver, is used to receive and display the enhanced road condition images to assist the driver. The step of enhancing the road condition image to obtain an enhanced road condition image includes: The road condition image is divided into blocks to obtain different road condition regions; Different grayscale ranges are obtained based on the pixel count variation characteristics in the grayscale histogram of the road condition area; grayscale distribution feature values ​​are obtained based on the pixel count difference characteristics of different grayscale values ​​within the grayscale range; contrast feature values ​​are obtained based on the grayscale characteristics of the road condition area, the number characteristics of the grayscale range, and the grayscale distribution feature values. Clustering is performed on the pixels within the road condition area to obtain two pixel clusters; texture confidence is obtained based on the gradient difference features of the pixel clusters and the distance features between pixels in the pixel clusters; enhancement coefficient is obtained based on the contrast feature value and the texture confidence; and adaptive contrast limit value is obtained based on the enhancement coefficient and the number of pixels with different gray values ​​in the road condition area. The road condition region is enhanced using a contrast-limited adaptive histogram equalization algorithm based on the adaptive contrast limit value to obtain an enhanced road condition region; an enhanced road condition image is obtained based on all enhanced road condition regions. The step of obtaining the contrast feature value based on the grayscale features of the road condition area, the quantity features of the grayscale range, and the grayscale distribution feature value includes: Calculate the average value of grayscale distribution feature values ​​for all grayscale ranges in the road condition area to obtain the average grayscale distribution value; calculate the product of the number of grayscale value types and the number of grayscale ranges in the road condition area to obtain the first factor; calculate the product of the reciprocal of the average grayscale distribution value and the first factor to obtain the contrast feature value of the road condition area. The step of obtaining texture confidence based on the gradient difference features of the pixel clusters and the pixel distance features in the pixel clusters includes: Calculate the average gradient value of the pixels in the pixel cluster to obtain the average gradient value; calculate the average Euclidean distance between all pixels and the nearest pixel in the pixel cluster corresponding to the maximum average gradient value and perform negative correlation mapping to obtain the neighboring feature value; calculate the product of the absolute value of the difference between the average gradient values ​​of the pixel cluster and the neighboring feature value to obtain the texture confidence of the road condition area.

2. The visual recognition system for assisted driving of automobiles according to claim 1, characterized in that, The step of obtaining different grayscale ranges based on the pixel count variation characteristics in the grayscale histogram of the road condition area includes: A curve is fitted based on the number of pixels corresponding to all gray levels in the gray-scale histogram to obtain a pixel count representation curve; the gray-scale range is constructed based on the gray-scale values ​​of the adjacent inflection points before and after any peak point in the pixel count representation curve.

3. The visual recognition system for assisted driving of automobiles according to claim 1, characterized in that, The step of obtaining grayscale distribution feature values ​​based on the difference in the number of pixels of different grayscale values ​​within the grayscale range includes: Calculate the super-kurtosis of the number of pixels in the grayscale range and perform positive correlation mapping to obtain the grayscale distribution feature value of the grayscale range.

4. A visual recognition system for assisted driving of automobiles according to claim 1, characterized in that, The step of clustering pixels based on the gradient features of pixels within the road condition area to obtain two pixel clusters includes: Based on the gradient values ​​of the pixels in the road condition area, the pixels are clustered using the K-means clustering algorithm to obtain two pixel clusters.

5. A visual recognition system for assisted driving of automobiles according to claim 1, characterized in that, The step of obtaining the enhancement coefficient based on the contrast feature value and the texture confidence includes: The enhancement coefficient of the road condition region is obtained by multiplying the reciprocal of the contrast feature value with the texture confidence.

6. A visual recognition system for assisted driving of automobiles according to claim 1, characterized in that, The step of obtaining the adaptive contrast limit value based on the enhancement coefficient and the number of pixels with different gray values ​​in the road condition area includes: Calculate the ratio of the number of pixels in the road condition area to the maximum number of pixels in the grayscale histogram to obtain the pixel number difference value; calculate the product of the pixel number difference value and the enhancement coefficient and normalize it to obtain the adaptive contrast limit value of the road condition area.

Images