Auxiliary driving image reconstruction method and system in severe weather

Abstract

The invention discloses an auxiliary driving image reconstruction method and system in severe weather, and relates to the technical field of machine learning, and the method comprises the steps: constructing a multi-source detection database based on historical collection data; connecting a weather platform to obtain weather information; carrying out meteorological factor decomposition to obtain environmental factor characteristics; carrying out correlation analysis by utilizing the environmental factor characteristics and the influence parameters of the detection equipment, and carrying out matching sorting on multiple detection means; generating an execution instruction, calling a plurality of detection means to carry out detection and acquisition, carrying out image recognition on a detection result, and determining a recognition marking result; and overlapping and splicing the identification marking results to obtain an image construction result, and stopping executing the instruction when the image construction result reaches a preset target requirement. According to the invention, the technical problems of limited perception capability and low accuracy of the auxiliary driving image in severe weather in the prior art are solved, and the technical effects of improving the accuracy of the auxiliary driving image and improving the driving safety of the vehicle in severe weather are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of machine learning, and particularly relates to an auxiliary driving image reconstruction method and system in bad weather. Background Art

[0002] With the continuous development of vehicle intelligence and advanced driver assistance systems (ADAS), the construction of auxiliary driving images has become an important development direction in the automotive industry. Auxiliary driving images can obtain detailed information about the vehicle's surrounding environment and generate real-time driving images required by the vehicle. These images can not only help drivers better understand the situation around the vehicle but also provide necessary information for the autonomous driving system, playing an important role in driving safety.

[0003] Currently, in-vehicle cameras and sensors are the main means of obtaining information about the vehicle's surrounding environment. At the same time, advanced algorithms and data processing technologies are also widely used in the construction of auxiliary driving images. However, bad weather and complex environments can affect the perception ability of in-vehicle sensors, resulting in a decline in the quality of the obtained environmental data, and further affecting the accuracy and reliability of auxiliary driving images. The existing technology has the technical problems that the perception ability of auxiliary driving images is limited and the accuracy is low in bad weather. Summary of the Invention

[0004] By providing an auxiliary driving image reconstruction method and system in bad weather, the present application effectively solves the technical problems existing in the prior art, namely, the limited perception ability and low accuracy of auxiliary driving images in bad weather, and achieves the technical effects of improving the accuracy of auxiliary driving images and the driving safety of vehicles in bad weather.

[0005] The present application provides an auxiliary driving image reconstruction method and system in bad weather, and the technical solution is as follows:

[0006] In a first aspect, an embodiment of the present application provides an auxiliary driving image reconstruction method in bad weather, and the method includes:

[0007] Construct a multi-source detection database based on historical acquisition data, where the multi-source detection database includes various detection means and detection device influence parameters;

[0008] Connect to a weather platform and obtain local weather information in real time based on positioning;

[0009] Perform meteorological factor decomposition based on the local weather information to obtain environmental factor characteristics;

[0010] Perform correlation analysis using the environmental factor characteristics and the detection device influence parameters, and perform matching and sorting on various detection means based on the correlation analysis results;

[0011] Generate execution instructions according to the matching sorting, call various detection means to perform detection and acquisition according to the matching sorting sequence, and perform image recognition on the detection results in turn to determine the recognition marker results;

[0012] Overlay and splice the recognition marker results of multiple detection results to obtain an image construction result. When the image construction result meets the preset target requirements, stop executing the instructions.

[0013] In a second aspect, an auxiliary driving image reconstruction system in bad weather provided by an embodiment of the present application includes:

[0014] A multi-source detection database construction module, which is used to construct a multi-source detection database based on historical acquisition data. The multi-source detection database includes various detection means and detection device influence parameters;

[0015] A local weather information acquisition module, which is used to connect to a weather platform and obtain local weather information in real time based on positioning;

[0016] An environmental factor feature acquisition module, which is used to decompose meteorological factors based on the local weather information to obtain environmental factor features;

[0017] A matching sorting module, which is used to perform correlation analysis on the environmental factor features and the detection device influence parameters, and perform matching sorting on various detection means based on the correlation analysis results;

[0018] A recognition marker result determination module, which is used to generate execution instructions according to the matching sorting, call various detection means to perform detection and acquisition according to the matching sorting sequence, and perform image recognition on the detection results in turn to determine the recognition marker results;

[0019] An image construction module, which is used to overlay and splice the recognition marker results of multiple detection results to obtain an image construction result. When the image construction result meets the preset target requirements, stop executing the instructions.

[0020] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0021] This application constructs a multi-source detection database based on historical acquisition data. The multi-source detection database includes various detection means and detection equipment influence parameters. Connect to the weather platform, obtain local weather information in real time based on positioning, decompose meteorological factors based on the local weather information to obtain environmental factor characteristics, and then perform correlation analysis using the environmental factor characteristics and the detection equipment influence parameters. Based on the correlation analysis results, perform matching and sorting on various detection means, generate execution instructions according to the matching and sorting, then call various detection means to perform detection and acquisition according to the matching and sorting sequence, perform image recognition on the detection results in sequence to determine the recognition mark results, superimpose and splice the recognition mark results of multiple detection results to obtain an image construction result. When the image construction result meets the preset target requirements, stop the execution instruction and output the image construction result as an auxiliary driving image. It effectively solves the technical problems of limited perception ability and low accuracy of auxiliary driving images in bad weather in the prior art, and achieves the technical effects of improving the accuracy of auxiliary driving images and improving the driving safety of vehicles in bad weather. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0023] Figure 1 Schematic flowchart of a method for reconstructing an auxiliary driving image in bad weather provided by an embodiment of this application;

[0024] Figure 2 Schematic flowchart of an image construction method in a method for reconstructing an auxiliary driving image in bad weather provided by an embodiment of this application;

[0025] Figure 3 Schematic structural diagram of a system for reconstructing an auxiliary driving image in bad weather provided by an embodiment of this application.

[0026] Explanation of reference numerals: Multi-source detection database construction module 1, local weather information acquisition module 2, environmental factor characteristic acquisition module 3, matching and sorting module 4, recognition mark result determination module 5, image construction module 6. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0027] This application provides a method and system for reconstructing an auxiliary driving image in bad weather, which is used to solve the technical problems of limited perception ability and low accuracy of auxiliary driving images in bad weather in the prior art.

[0028] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative efforts belong to the scope protected by the present application.

[0029] It should be noted that the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or server including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or modules not clearly listed or inherent to these processes, methods, products or devices.

[0030] Embodiment 1

[0031] As Figure 1 shown, the present invention provides an assisted driving image reconstruction method under bad weather conditions, which is used to improve the driving safety of vehicles under bad weather conditions. The method includes:

[0032] Construct a multi-source detection database based on historical acquisition data. Among them, the historical acquisition data refers to the data collected during a certain period in the past. The multi-source detection database includes various detection means and detection device influence parameters. Among them, the various detection means refer to a series of technologies and methods for identifying, measuring and tracking the information of the vehicle's surrounding environment, including detection means such as lidar, cameras, millimeter wave radars, ultrasonic sensors, and infrared sensors. These detection means are used to provide information such as the position and speed of objects to help the driver better understand the surrounding environment. The detection device influence parameters refer to the factors that affect the performance and accuracy of the detection device under bad weather conditions, including factors such as rainfall, snowfall, fog, sandstorm, strong light, and temperature. By analyzing the data collected by various detection means over a period of time in the past, it is possible to obtain which environmental factors will affect the performance and accuracy of the detection device and the degree of influence. Furthermore, a multi-source detection database can be established based on the original data collected by the detection device, various analysis result data, etc., which is used to provide basic data for assisted driving image reconstruction.

[0033] Connect to the weather platform to obtain local weather information in real time based on positioning. Among them, the weather platform refers to a system or service that provides real-time weather information and is used to provide the weather conditions at the current location and driving direction of the vehicle. The local weather information refers to the real-time weather conditions at a specific location, including parameters such as temperature, humidity, wind speed, rainfall, visibility, air pressure, cloud cover, and solar radiation. The GPS system equipped on the vehicle locates the vehicle. Through network communication, the vehicle sends its location and driving direction to the weather platform to request corresponding weather information. The weather platform obtains real-time weather information through meteorological stations, meteorological satellites, radars, etc. and feeds it back to the vehicle, so that the vehicle can obtain real-time local weather information.

[0034] Perform meteorological factor decomposition based on the local weather information to obtain environmental factor characteristics. Among them, the meteorological factor decomposition is to extract and analyze and decode each meteorological factor in the local weather information. Among them, the meteorological factors include temperature, humidity, wind speed, rainfall, air pressure, etc. The environmental factor characteristics refer to the characteristics and change laws of various environmental factors, including meteorological environment, road environment, etc., and are used to characterize various factors and attributes in the driving environment. By extracting, analyzing, and decoding the meteorological factors of the local weather, the environmental factor characteristics of driving can be obtained. Based on these environmental factor characteristics, the assisted driving system can make more scientific and reasonable driving decisions and operations, such as adjusting the vehicle speed, maintaining a safe distance, etc. For the sake of understanding, a possible example is given below for explanation:

[0035] On a rainy day, when the vehicle is driving, it obtains some information about the local weather through the weather platform, including temperature, humidity, rainfall, etc. By decomposing these meteorological factors such as temperature, humidity, and rainfall, the assisted driving system of the vehicle obtains the environmental factor characteristics of rainy weather, including environmental characteristics such as low temperature, low visibility, high road surface humidity, and small friction coefficient. Furthermore, the assisted driving system generates instructions to reduce the vehicle speed and increase the following distance.

[0036] Perform a correlation analysis using the environmental factor characteristics and the detection device impact parameters. Specifically, the accuracy, sensitivity, measurement range, etc. of the detection device are affected by weather conditions. Therefore, by performing a correlation analysis on the environmental factor characteristics and the detection device impact parameters, the response characteristics and accuracy performance of the sensor under different weather conditions can be understood. Furthermore, the configuration and parameters of the sensor can be optimized for different environmental conditions to improve the perception and decision-making capabilities of the assisted driving system. For example, on a sunny day, the camera can provide high-definition images. However, on a rainy day, a single raindrop or a grain of sand may block the camera's field of view, resulting in its inability to correctly perceive the surrounding environment. Therefore, under rainy conditions, the detection accuracy and sensitivity of the camera are greatly affected. However, since millimeter-wave radar has strong penetration ability and can better penetrate bad weather such as rain and fog, the detection ability of millimeter-wave radar is relatively stable on rainy days, and the detection accuracy and sensitivity are less affected.

[0037] Match and sort multiple detection methods based on the correlation analysis results. Specifically, according to the correlation between the environmental factor characteristics and the detection device impact parameters, the applicability and advantages of different detection methods under different environmental conditions can be determined. Then, they can be sorted according to the size of the applicability and advantages. The greater the advantage, the higher the ranking. For example, on a rainy day, since the detection ability of the camera may be greatly affected while the millimeter-wave radar is relatively stable, the millimeter-wave radar ranks first and the camera ranks second on rainy days.

[0038] Generate an execution instruction according to the match sorting. Among them, the execution instruction is used to call multiple detection methods to perform detection and acquisition according to the match sorting sequence, and perform image recognition on the detection results in turn to determine the recognition marker results. For example, the match sorting sequence of multiple detection methods is: 1 lidar, 2 camera, 3 ultrasonic radar. First, use the lidar for detection and acquisition. For the collected data, use an image recognition algorithm for recognition and generate corresponding labels, and then obtain information such as the category, position, and speed of the target. Determine the recognition marker results based on this information. Next, perform the same operation steps as above using the camera and ultrasonic radar according to the sorting sequence. Finally, obtain 3 recognition marker results.

[0039] Overlay and splice the recognition marker results of multiple detection results to obtain an image construction result. Specifically, collect the results of recognition markers from multiple detection means, and these results include information such as the position, size, speed, and category of the target. Align and synchronize these results, overlay and splice the aligned results, and use the results after overlay and splicing to construct an image of the vehicle's surrounding environment. When the image construction result meets the preset target requirements, stop executing the instruction. Among them, the preset target requirements are the targets or standards for the results set in advance, such as accuracy, resolution, etc. Specifically, after constructing the image, evaluate it to determine whether it meets the preset target requirements. For example, the preset target is to identify all pedestrians around the vehicle, and the recognition accuracy rate needs to reach 99%. Then the evaluation process is to check whether all pedestrians are correctly identified. If the image construction result meets the preset target requirements, such as the recognition accuracy rate is 100%, then stop executing the instruction, that is, no longer call the detection means for detection and collection according to the matching sorting sequence, and output the constructed image to the assisted driving system. Otherwise, continue to execute the instruction, that is, perform detection and collection according to the next detection means in the matching sorting sequence until the requirements are met. It achieves the technical effect of improving the driving safety of the vehicle in bad weather.

[0040] In a preferred implementation manner provided by the embodiment of the present application, a multi-source detection database is constructed based on historical acquisition data, including:

[0041] Based on multiple detection means, obtain the multi-source historical acquisition data of each detection means. In other words, first determine the detection means to be used, which may include lidar, camera, ultrasonic radar, etc., and retrieve the multi-source historical acquisition data of each detection means from the local or cloud database. The multi-source historical acquisition data includes multiple weather environments, that is, the multi-source historical acquisition data contains the measurement result data of multiple detection means in multiple weather environments.

[0042] Conduct acquisition image feature analysis according to the multi-source historical acquisition data, and establish an image-environment relationship list. First, preprocess the multi-source historical acquisition data, including operations such as data cleaning, format conversion, and coordinate conversion, to ensure data quality and consistency. Extract image features related to the environment from the preprocessed data. These features include features such as the color, texture, and shape of the image, as well as features such as distance and speed extracted from sensors such as lidar and ultrasonic radar. Analyze the extracted features. For example, analyze the changes in features such as the color, texture, and shape of the vehicle's surrounding environment under different weather conditions. According to the results of the feature analysis, establish an image-environment relationship list. The following is an example of a possible image-environment relationship list:

[0043]

[0044] Based on the image-environment relationship list, perform influence relationship analysis to obtain the influence parameters of the detection device. Specifically, different weather conditions have different degrees of influence on the detection device. For example, through the analysis of the image-environment relationship list of the same detection device, it is found that color features are related to factors such as weather and light. Then the influence parameters of this detection device include weather type, light intensity, etc. Therefore, by performing influence relationship analysis on the image-environment relationship list, the influence parameters of the detection device can be obtained. Furthermore, the influence degree of changing the type or size of a certain image parameter on the corresponding image feature can be studied.

[0045] Establish the mapping relationships among the multiple detection means, the image-environment relationship list, and the influence parameters of the detection device, and construct the multi-source detection database. For the multiple detection means, establish their mapping relationships with the image-environment relationship list. Then, in the image-environment relationship list, analyze and summarize the relationships between various image features and environmental factors. Use these relationships to establish mapping relationships between the image-environment relationship list and the influence parameters of the detection device. Finally, integrate the above mapping relationships into a database to construct a multi-source detection database. This database includes information such as data of multiple detection means, the image-environment relationship list, and the influence parameters of the detection device. For the sake of understanding, the following lists two possible mapping relationships for explanation:

[0046] 1. The distance data of the lidar corresponds to the distance feature in the image-environment relationship list and also corresponds to the object speed in the influence parameters of the detection device;

[0047] 2. The color data of the camera corresponds to the color feature in the image-environment relationship list and also corresponds to the light intensity in the influence parameters of the detection device.

[0048] Then when constructing the multi-source detection database, it is necessary to integrate all the information such as the distance data of the lidar, the color data of the camera, the distance and color features in the image-environment relationship list, and the object speed and light intensity in the influence parameters of the detection device together to form a complete multi-source detection database. It achieves the technical effect of improving the intelligent level of the assisted driving system.

[0049] In another preferred implementation manner provided by the embodiments of the present application, perform acquisition image feature analysis according to the multi-source historical acquisition data and establish an image-environment relationship list, including:

[0050] Image evaluation is performed on the multi-source historical acquisition data according to preset image evaluation parameters to determine the acquisition image feature set. Among them, the preset image evaluation parameters are some pre-set image evaluation parameters, such as color, texture, shape, sharpness, etc., which are used to evaluate the characteristics of the acquisition images. The acquisition image feature set refers to a set of feature vectors extracted from the multi-source historical acquisition data, which are used to describe various characteristics of the acquisition images, including characteristics such as color, texture, shape, and their corresponding parameters. According to the preset image evaluation parameters, computer vision and image processing technologies are used to perform image evaluation on the multi-source historical acquisition data. By performing image evaluation and analysis on the multi-source historical acquisition data, the acquisition image feature set is determined.

[0051] Image feature trend fitting is performed on the acquisition image feature set according to different preset image evaluation parameters to obtain the image feature trend. Specifically, according to the preset image evaluation parameters, image processing, machine learning, deep learning and other technologies are used to fit the image features of each acquisition image, obtain their trends and analyze them. By performing image feature trend fitting and analysis on the acquisition image feature set, the image feature trend is obtained, which is used to characterize the changes of each image feature under the influence of time or other factors.

[0052] Feature impact degree analysis is performed based on the curve slope of the image feature trend to determine the image impact information. Among them, the curve slope of the image feature trend reflects the change speed and trend of the feature. The feature impact degree analysis refers to evaluating and analyzing the curve slope of the image feature trend to determine the impact degree of each feature on the image. If the curve slope of a certain feature is large, it means that the impact degree of this feature on the image is high. The image impact information refers to the information such as the impact degree, importance and relationship of each feature on the image obtained through feature impact degree analysis.

[0053] The environmental factors are determined according to the weather environment of the multi-source historical acquisition data. Among them, the environmental factors refer to the environmental information related to the images obtained from the multi-source historical acquisition data, such as weather, light, etc. These environmental factors will affect the changes and manifestations of image features. For example: if the weather condition at the time of data acquisition is cloudy, then the weather type is used as an environmental factor.

[0054] Finally, establish the influence correspondence between the image influence information and the environmental factors to obtain the image - environment relationship list. In other words, establish the connection and correspondence between the image features and the environmental factors. For example, if the trend of a certain image feature is closely related to the weather type, then by correlating and analyzing the influence information of this feature with the weather type, establish the correspondence between them. Furthermore, based on the analysis of the feature influence degree and the environmental factors, establish the image - environment relationship list, which includes the mutual influence relationships between various image features and environmental factors, as well as information such as their influence degrees and trends. It achieves the technical effect of enabling the vehicle to better adapt to various different weather conditions.

[0055] In another preferred implementation manner provided in the embodiments of the present application, perform a correlation analysis on the environmental factor features and the detection device influence parameters, and perform matching and sorting on multiple detection means based on the correlation analysis results, including:

[0056] Based on the multi-source detection database, perform an influence relationship analysis according to the environmental factor features to determine the detection devices and image influence information affected. Specifically, input the real-time environmental factor features into the multi-source detection database, and by traversing all the data related to this environmental factor in the database, find the detection devices and image influence information affected by this environmental factor feature.

[0057] According to the detection devices and image influence information affected, establish a multi-correspondence between the multi-source environmental factors and multiple detection devices, that is, construct a mapping relationship to connect the interactions and influences between different environmental factors and multiple detection devices, and fit and evaluate the fitness function with the minimum influence degree as the goal based on the multi-correspondence. That is, in order to evaluate the quality of the multi-correspondence, define an evaluation fitness function. Among them, the multi-source environmental factor set F = {F1, F2, F3...}, the multiple detection device set D = {D1, D2, D3...}, and the evaluation fitness function formula is as follows:

[0058]

[0059] Among them, D i represents multiple detection devices, f(D i ) represents the influence degree score of multiple detection devices, ∑R ij represents the total response of multiple detection devices D i to all environmental factors, ∑R ji represents the total response of all devices to the environmental factor F j .

[0060] This function can measure the accuracy of multiple correspondence relationships, with the minimum influence score as the index. Here, the influence degree is the degree of interference of environmental factors on the detection device. The fitness function is optimized with the goal of minimizing the influence degree. By adjusting the parameters or rules in the multiple correspondence relationships, the value of the evaluation fitness function is minimized, thereby fitting out the optimal multiple correspondence relationship.

[0061] Based on the evaluation fitness function, optimize the analysis of the sorting scheme to determine the matching sorting of multiple detection methods. Sort the influence scores of all devices. The smaller the influence score, the more suitable it is to monitor the environmental factor. According to the above multiple correspondence relationships, different detection schemes are designed for different environmental factors and detection tasks. These schemes include different detection methods and sequences. For each scheme, calculate the value of its fitness function according to the defined evaluation fitness function. Sort according to the calculated values of the fitness functions of each scheme, and select the optimal scheme as the sorting scheme. This scheme determines multiple detection methods and the order of multiple detection methods. It achieves the technical effects of improving detection efficiency and enhancing the reliability of detection results.

[0062] In another preferred implementation provided in the embodiments of the present application, the multiple detection methods are called to perform detection and acquisition according to the matching sorting sequence, and the detection results are sequentially subjected to image recognition to determine the recognition marking results, including:

[0063] According to the matching sorting sequence, obtain the first execution acquisition device, and use the first execution acquisition device to perform detection to obtain the first detection result. For example, in the matching sorting sequence, the camera is ranked first, then the first execution acquisition device is the camera, and the camera is used to detect a specific area to obtain the first detection result.

[0064] Perform detection image recognition on the first detection result according to the recognition algorithm of the first execution acquisition device to determine the first recognition image, that is, use a specific recognition algorithm to perform image recognition on the first detection result to identify specific image features. Based on the result of the image recognition, determine the first recognition image, and the first recognition image is the key feature extracted from the first detection result.

[0065] Based on the recognition algorithm of the first execution acquisition device and the first recognition image, determine the core recognition points. Here, the core recognition points are specific objects or areas to be recognized, etc., including key features of the image, target objects, abnormal areas, etc. Use the selected recognition algorithm to analyze the first recognition image to determine the core recognition points in the image, and mark the core recognition points, including drawing frames, circles, arrows, etc. on the image to highlight these core recognition points, and then obtain the recognition marking results. It achieves the technical effects of improving the stability and reliability of the system.

[0066] As Figure 2 shown, in another preferred implementation provided by the embodiments of the present application, the recognition marker results of multiple detection results are superimposed and spliced to obtain an image construction result. When the image construction result meets the preset target requirements, the instruction execution is stopped, including:

[0067] According to the matching sorting sequence, a first recognition marker result and a second recognition marker result are obtained. The first recognition marker result corresponds to the detection device ranked first, and the second recognition marker result corresponds to the detection device ranked second. That is, according to the matching sorting sequence, the first two ranked detection devices are sequentially detected and collected. For the collected data, an image recognition algorithm is used for recognition to generate corresponding labels, and then information such as the category, position, and speed of the target is obtained. According to this information, the first recognition marker result and the second recognition marker result are respectively determined.

[0068] Align and superimpose according to the recognition marker results, splice the recognition marker points to obtain the image construction result, that is, align and synchronize the first recognition marker result and the second recognition marker result, superimpose and splice the two aligned results, and use the superimposed and spliced result to construct an image of the vehicle surrounding environment.

[0069] Judge whether the image construction result meets the preset target requirements. When it meets, output the image construction result as an assisted driving image, that is, when the image construction result meets the preset target requirements, output the image construction result. Specifically, after the image is constructed, it is evaluated to determine whether it meets the preset target requirements. For example, the preset target is that the resolution must reach 1828×948. Then the evaluation process is to detect whether the resolution of the constructed image reaches this preset value. If the result of the image construction reaches the preset target requirements, such as the resolution is 3840×2160, then the constructed image is output to the assisted driving system.

[0070] When it does not meet, continue to perform the detection operation of the next execution collection device according to the matching sorting sequence to obtain a third recognition marker result. Align and superimpose the marker points of the third recognition marker result with the image construction result, and reconstruct the image construction result based on the superimposed result, that is, continue to perform the detection and collection according to the next detection method in the matching sorting sequence, and superimpose the collection result with the previously constructed image result. Judge whether the reconstruction result meets the preset target requirements. When it meets, output it as an assisted driving image. When it does not meet, continue to execute the next collection device until the requirements are met. It achieves the technical effects of increasing the accuracy, efficiency, and reliability of image construction.

[0071] Embodiment 2

[0072] Based on the same inventive concept as the method for reconstructing auxiliary driving images in bad weather in the foregoing embodiments, as Figure 3 shown, this application provides a system for reconstructing auxiliary driving images in bad weather. The system in the embodiments of this application and the method embodiments are based on the same inventive concept. Among them, the system includes:

[0073] A multi-source detection database construction module 1, which is used to construct a multi-source detection database based on historical acquisition data. The multi-source detection database includes various detection means and detection device influence parameters;

[0074] A local weather information acquisition module 2, which is used to connect to a weather platform and obtain local weather information in real time based on positioning;

[0075] An environmental factor feature acquisition module 3, which is used to decompose meteorological factors based on the local weather information to obtain environmental factor features;

[0076] A matching and sorting module 4, which is used to perform correlation analysis using the environmental factor features and the detection device influence parameters, and perform matching and sorting on various detection means based on the correlation analysis results;

[0077] An identification marking result determination module 5, which is used to generate an execution instruction according to the matching and sorting, call various detection means to perform detection and acquisition according to the matching and sorting sequence, and perform image recognition on the detection results in turn to determine the identification marking result;

[0078] An image construction module 6, which is used to superimpose and splice the identification marking results of multiple detection results to obtain an image construction result. When the image construction result reaches the preset target requirements, the execution instruction is stopped.

[0079] Furthermore, the multi-source detection database construction module 1 is used to execute the following method:

[0080] Based on various detection means, obtain the multi-source historical acquisition data of each detection means, where the multi-source historical acquisition data includes various weather environments;

[0081] Perform acquisition image feature analysis according to the multi-source historical acquisition data to establish an image-environment relationship list;

[0082] Perform influence relationship analysis based on the image-environment relationship list to obtain the influence parameters of the detection device, establish the mapping relationship of the various detection means, the image-environment relationship list, and the influence parameters of the detection device, and construct the multi-source detection database.

[0083] Further, the multi-source detection database construction module 1 is used to execute the following method:

[0084] Evaluate the multi-source historical acquisition data according to preset image evaluation parameters to determine the acquisition image feature set;

[0085] Perform image feature trend fitting on the acquisition image feature set according to different preset image evaluation parameters to obtain the image feature trend;

[0086] Based on the curve slope of the image feature trend, perform feature influence degree analysis to determine the image influence information. According to the weather environment of the multi-source historical acquisition data, determine the environmental factors, establish the influence correspondence relationship between the image influence information and the environmental factors, and obtain the image-environment relationship list.

[0087] Further, the matching and sorting module 4 is used to execute the following method:

[0088] Based on the multi-source detection database, perform influence relationship analysis according to the environmental factor characteristics to determine the influence detection device and the image influence information;

[0089] According to the influence detection device and the image influence information, establish a multi-correspondence relationship between the multi-source environmental factors and multiple detection devices, and based on the multi-correspondence relationship, fit the evaluation fitness function with the minimum influence degree as the goal;

[0090] Based on the evaluation fitness function, perform sorting scheme optimization analysis to determine the matching and sorting of multiple detection means.

[0091] Further, the recognition marker result determination module 5 is used to execute the following method:

[0092] Obtain the first execution acquisition device according to the matching sorting sequence, use the first execution acquisition device for detection, and obtain the first detection result;

[0093] Perform detection image recognition on the first detection result according to the recognition algorithm of the first execution acquisition device to determine the first recognition image;

[0094] Based on the recognition algorithm of the first execution acquisition device and the first recognition image, determine the core recognition points, and mark the core recognition points to obtain the recognition marker result.

[0095] Further, the image construction module 6 is used to execute the following method:

[0096] Obtain the first recognition marker result and the second recognition marker result according to the matching sorting sequence. The first recognition marker result corresponds to the detection device ranked first, and the second recognition marker result corresponds to the detection device ranked second;

[0097] Align and stack according to the recognition marker results, splice the recognition marker points, and obtain the image construction result;

[0098] Judge whether the image construction result meets the preset target requirements. When it meets, output the image construction result as the auxiliary driving image;

[0099] When it does not meet, continue to perform the next execution of the acquisition device detection operation according to the matching sorting sequence, obtain the third recognition marker result, align and stack the marker points of the third recognition marker result with the image construction result, and reconstruct the image construction result based on the stacking result;

[0100] Judge whether the reconstruction result meets the preset target requirements. When it meets, output it as the auxiliary driving image. When it does not meet, continue to execute the next acquisition device until the requirements are met.

[0101] It should be noted that the above order of the embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments. And the above specific embodiments of this specification have been described. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments and still achieve the desired result. Additionally, the processes depicted in the figures do not necessarily require the particular order or sequential order shown to achieve the desired result. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0102] The above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0103] This specification and the drawings are only exemplary descriptions of the present application and are considered to have covered any and all modifications, variations, combinations, or equivalents within the scope of the present application. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is intended to include these changes and modifications.

Claims

1. An assisted driving image reconstruction method under bad weather, characterized in that, The method includes: Constructing a multi-source detection database based on historical acquisition data, where the multi-source detection database includes various detection means and detection device influence parameters; Connecting to a weather platform to obtain local weather information in real time based on positioning; Decomposing meteorological factors based on the local weather information to obtain environmental factor characteristics; Performing correlation analysis using the environmental factor characteristics and the detection device influence parameters, and performing matching sorting on various detection means based on the correlation analysis results; Generating an execution instruction according to the matching sorting, calling various detection means to perform detection and acquisition according to the matching sorting sequence, and performing image recognition on the detection results in sequence to determine the recognition marking result; Overlaying and splicing the recognition marking results of multiple detection results to obtain an image construction result, and stopping the execution instruction when the image construction result meets the preset target requirements.

2. The method according to claim 1, characterized in that The constructing of the multi-source detection database based on historical acquisition data includes: Based on various detection means, obtaining multi-source historical acquisition data of each detection means, where the multi-source historical acquisition data includes various weather environments; Performing analysis on the characteristics of the acquired images according to the multi-source historical acquisition data to establish an image-environment relationship list; Performing influence relationship analysis based on the image-environment relationship list to obtain the influence parameters of the detection device, establishing a mapping relationship among the various detection means, the image-environment relationship list, and the influence parameters of the detection device, and constructing the multi-source detection database.

3. The method according to claim 2, wherein Performing analysis on the characteristics of the acquired images according to the multi-source historical acquisition data to establish an image-environment relationship list, including: Performing image evaluation on the multi-source historical acquisition data according to preset image evaluation parameters to determine an acquired image feature set; Performing fitting of the image feature trend on the acquired image feature set according to different preset image evaluation parameters to obtain an image feature trend; Performing feature influence degree analysis based on the curve slope of the image feature trend to determine image influence information, determining environmental factors according to the weather environment of the multi-source historical acquisition data, establishing an influence correspondence relationship between the image influence information and the environmental factors, and obtaining the image-environment relationship list.

4. The method according to claim 3, wherein Performing correlation analysis using the environmental factor characteristics and the detection device influence parameters, and performing matching sorting on various detection means based on the correlation analysis results, including: Based on the multi-source detection database, performing influence relationship analysis according to the environmental factor characteristics to determine the influence detection device and image influence information; According to the influence detection device and image influence information, establishing a multi-correspondence relationship between multi-source environmental factors and various detection devices, and fitting and evaluating a fitness function with the minimum influence degree as the goal based on the multi-correspondence relationship; Performing optimization analysis of the sorting scheme based on the evaluation fitness function to determine the matching sorting of various detection means.

5. The method according to claim 1, characterized in that, The calling of various detection means to perform detection and acquisition according to the matching sorting sequence, and performing image recognition on the detection results in sequence to determine the recognition marking result, includes: According to the matching sorting sequence, obtaining a first execution acquisition device, and using the first execution acquisition device to perform detection to obtain a first detection result; Perform detection image recognition on the first detection result according to the recognition algorithm of the first execution acquisition device to determine the first recognition image; Based on the recognition algorithm of the first execution acquisition device and the first recognition image, determine the core recognition points, mark the core recognition points, and obtain the recognition marking result.

6. The method according to claim 5, wherein Superimpose and splice the recognition marking results of multiple detection results to obtain the image construction result. When the image construction result meets the preset target requirements, stop executing the instruction, including: According to the matching sorting sequence, obtain the first recognition marking result and the second recognition marking result. The first recognition marking result corresponds to the detection device ranked first, and the second recognition marking result corresponds to the detection device ranked second; Align and superimpose according to the recognition marking result, splice the recognition marking points, and obtain the image construction result; Judge whether the image construction result meets the preset target requirements. When it meets, output the image construction result as the auxiliary driving image; When it does not meet, continue to perform the detection operation of the next execution acquisition device according to the matching sorting sequence, obtain the third recognition marking result, align and superimpose the marking points of the third recognition marking result with the image construction result, and reconstruct the image construction result based on the superimposed result; Judge whether the reconstruction result meets the preset target requirements. When it meets, output it as the auxiliary driving image. When it does not meet, continue to execute the next acquisition device until the requirements are met.

7. An assisted driving image reconstruction system under bad weather, characterized in that, The system includes: A multi-source detection database construction module, which is used to construct a multi-source detection database based on historical acquisition data. The multi-source detection database includes various detection means and detection device influence parameters; A local weather information acquisition module, which is used to connect to the weather platform and acquire local weather information in real time based on the location; An environmental factor feature acquisition module, which is used to decompose meteorological factors based on the local weather information to obtain environmental factor features; A matching sorting module, which is used to perform correlation analysis using the environmental factor features and the detection device influence parameters, and perform matching sorting on various detection means based on the correlation analysis result; A recognition marking result determination module, which is used to generate an execution instruction according to the matching sorting, call various detection means to perform detection and acquisition according to the matching sorting sequence, perform image recognition on the detection results in sequence, and determine the recognition marking result; An image construction module, which is used to superimpose and splice the recognition marking results of multiple detection results to obtain the image construction result. When the image construction result reaches the preset target requirements, stop executing the instruction.

Images