An embedded system and method for vehicle-mounted cameras for nighttime intelligent driving

Abstract

This invention relates to the field of autonomous driving technology and discloses an in-vehicle camera embedded system and method for nighttime intelligent driving. The system includes an image acquisition module, a target detection module, a data verification module, a human-machine interaction module, and a feedback module. Through the organic linkage of multiple modules, a complete closed-loop architecture with risk prediction, credibility verification, human-machine fusion decision-making, and emergency execution capabilities is constructed. Compared to existing driver assistance systems that rely solely on image recognition or single judgment logic, this system significantly improves the accuracy of target recognition, the stability of system judgment, and the timeliness of takeover response in low-light and complex traffic environments at night. It effectively overcomes the technical shortcomings of traditional solutions, such as misjudgment, slow response, or improper control at night.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving technology, specifically to an embedded system and method for nighttime intelligent driving using an in-vehicle camera. Background Technology

[0002] With the continuous development of technology, autonomous driving technology, as an important application of artificial intelligence, has been widely used in various modes of transportation and has become an important component of intelligent transportation systems. The core objective of autonomous driving technology is to achieve perception and understanding of the road environment through computer vision, sensor fusion, and other means, thereby replacing or assisting manual driving. Especially in complex driving environments, ensuring driving safety has become a key focus of research in this technology.

[0003] Despite significant advancements in automotive camera technology in recent years, limitations remain in their performance at night or in low-light conditions. Many current automotive cameras rely primarily on traditional visible light imaging techniques, resulting in images with high noise levels and poor contrast in dark or dimly lit environments, making it difficult to clearly identify pedestrians, obstacles, or other potential hazards on the road. While some systems enhance nighttime vision using infrared cameras and LiDAR, their processing technologies and algorithms have not yet fully resolved issues such as image blurring and low target recognition accuracy in low-light environments. Furthermore, data fusion and real-time decision-making between different sensors still face technical challenges such as data latency and information asymmetry, affecting the system's reaction speed and accuracy.

[0004] Therefore, we propose an embedded system and method for vehicle cameras for nighttime intelligent driving to solve the problems mentioned above. Summary of the Invention

[0005] The purpose of this invention is to provide an embedded system and method for vehicle cameras for intelligent driving at night, in order to solve the problem that before the above-mentioned background technology, many vehicle cameras mainly relied on traditional visible light imaging technology. This resulted in the images captured by the cameras often having a lot of noise and poor contrast in dark or dim environments, making it impossible to clearly identify pedestrians, obstacles or other potential dangers on the road.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an embedded system for vehicle-mounted cameras for intelligent driving at night, characterized in that it includes an image acquisition module, a target detection module, a data verification module, a human-computer interaction module, and a feedback module;

[0007] The image acquisition module is used to collect and preprocess data on the surrounding environment and vehicle conditions, and organize the data into a first data group and a second data group.

[0008] The target detection module is used to calculate the first data group and the second data group to generate the target monitoring reference coefficient MBJ, and to analyze the target monitoring reference coefficient MBJ.

[0009] The data verification module is used to calculate the first data group, the second data group, and the target monitoring reference coefficient to generate the data verification coefficient MBY, and to analyze the data verification coefficient MBY.

[0010] The human-computer interaction module is used to calculate the target monitoring reference coefficient MBJ and the data verification coefficient MBY, thereby generating the human-computer interaction coefficient JHX, and to analyze the human-computer interaction coefficient JHX.

[0011] The feedback module is used to execute takeover and alarm commands.

[0012] Preferably, the image acquisition module includes a data acquisition unit and a data preprocessing unit;

[0013] The data acquisition unit is used to collect data on the surrounding environment and vehicle conditions, including ambient light intensity, camera exposure, driving speed, obstacle distance, relative position angle of obstacles, lane deviation, camera clarity, and road curvature.

[0014] The data preprocessing unit is used to preprocess and dimensionless the collected parameters, and reorganize them into a first data group and a second data group.

[0015] The first data set includes ambient light intensity (A), obstacle distance (B), obstacle relative position angle (C), lane departure (D), and road curvature (E).

[0016] The second data set includes camera exposure (F), camera resolution (G), and driving speed (H).

[0017] Preferably, the target detection module includes a target coefficient generation unit and a target coefficient analysis unit;

[0018] The target coefficient generation unit is used to couple the first data group and the second data group by integrating ambient light intensity A, obstacle distance B, obstacle relative position angle C, lane deviation D, road curvature E, camera exposure F, camera clarity G, and driving speed H. Ambient light intensity A is calculated independently, obstacle distance B and driving speed H are coupled through an exponential function, obstacle relative position angle C, lane deviation D and road curvature E are coupled, and camera exposure F and camera clarity G are coupled. The multiple sets of calculations are integrated to finally generate the target monitoring reference coefficient MBJ.

[0019] The specific calculation formula is as follows:

[0020]

[0021] In the formula: A is the ambient light intensity, B is the distance to the obstacle, C is the relative position angle of the obstacle, D is the lane departure degree, E is the road curvature, F is the camera exposure, G is the camera clarity, and H is the driving speed;

[0022] The target coefficient analysis unit is used to analyze the target monitoring reference coefficient MBJ, so as to generate a first analysis result. According to the first analysis result, it is judged whether there are suspicious obstacles in the area.

[0023] Preferably, the first analysis result is specifically as follows:

[0024] When MBJ < Y, it means that there are no potential obstacles in the current visible area;

[0025] When MBJ = Y, it means that there are potential obstacles in the current visible area and secondary analysis and calibration are required;

[0026] When MBJ > Y, it means that there are visible obstacles in the current visible area and early warning is required.

[0027] Preferably, the data verification module includes a verification coefficient generation unit and a verification coefficient analysis unit;

[0028] The verification coefficient generation unit is used to couple the first data group, the second data group and the target monitoring reference coefficient MBJ. By coupling the light intensity A, the image clarity G and the exposure F, and then coupling the obstacle distance B, the relative position angle C of the obstacle, the lane departure degree D, the road curvature E and the camera exposure F, and then integrating and calculating the two groups of coupled data to finally generate the data verification coefficient MBY;

[0029] The specific calculation formula is as follows:

[0030]

[0031] In the formula: A is the ambient light intensity, B is the distance to the obstacle, C is the relative position angle of the obstacle, D is the lane departure degree, E is the road curvature, F is the camera exposure, G is the camera clarity, H is the driving speed, and MBJ is the target monitoring reference coefficient;

[0032] The verification coefficient analysis unit is used to analyze the data verification coefficient MBY, generate a second analysis result, and judge the rationality of the existence of suspicious objects according to the second analysis result.

[0033] Preferably, the second analysis result is specifically as follows:

[0034] When MBY < R, it means that the current data is not within the credible interval and no marking is performed;

[0035] When MBY = R, it means that the current data is in the low credible interval, marking is performed and ready for re-analysis;

[0036] When MBY > R, it means that the current data is in the high credible interval and ready for alarm.

[0037] Preferably, the human - machine interaction module includes a takeover coefficient generation unit and a takeover coefficient analysis unit;

[0038] The takeover coefficient generation unit is used to couple the target monitoring reference coefficient MBJ and the data verification coefficient MBY, and finally generate the human - machine interaction coefficient JHX;

[0039] The specific calculation formula is as follows:

[0040] JHX = a1×MBJ + a2×MBY + a3×(MBJ + MBY);

[0041] In the formula: MBJ is the target monitoring reference coefficient, MBY is the data verification coefficient, a1, a2, and a3 are weight values, and the values of a1, a2, and a3 are adjusted and set by the user;

[0042] The takeover coefficient analysis unit is used to perform data analysis, generate the third analysis result, and judge whether alarm and emergency takeover are needed according to the third analysis result.

[0043] Preferably, the third analysis result is specifically as follows:

[0044] When JHX < 0.4, it means that the current operation is stable, the risk is reduced, and the system does not perform intervention and alarm;

[0045] When 0.4 ≤ JHX < 0.8, it means that the current operation risk is increased, there is a primary risk, the system gives a voice prompt alarm and takes over after 10 seconds;

[0046] When JHX ≥ 0.8, it means that the current operation risk is extremely high, there is a secondary risk, the system immediately performs takeover and sends a message to the emergency contact.

[0047] This application also includes an in - vehicle camera embedding method for night intelligent driving, and the specific steps are as follows: [[ID=4​​​​S2. The target detection module calculates the first data group and the second data group to generate the target monitoring reference coefficient MBJ, and then analyzes the target monitoring reference coefficient MBJ.

[0050] S3. The data verification module calculates the first data group, the second data group, and the target monitoring reference coefficient to generate the data verification coefficient MBY, and then analyzes the data verification coefficient MBY.

[0051] S4. Calculate the target monitoring reference coefficient MBJ and data verification coefficient MBY through the human-computer interaction module to generate the human-computer interaction coefficient JHX, and analyze the human-computer interaction coefficient JHX.

[0052] S5. Execute takeover and alarm commands through the feedback module.

[0053] Compared with the prior art, the beneficial effects of the present invention are:

[0054] 1. This invention provides an embedded vehicle camera system for nighttime intelligent driving. Through the organic linkage of multiple modules, it constructs a complete closed-loop architecture with capabilities for risk prediction, reliability verification, human-machine fusion decision-making, and emergency execution. Compared to existing driver assistance systems that rely solely on image recognition or single judgment logic, this system significantly improves the accuracy of target recognition, the stability of system judgment, and the timeliness of takeover response in low-light and complex traffic environments at night. It effectively overcomes the technical shortcomings of traditional solutions, such as misjudgment, slow response, or improper control at night.

[0055] 2. This system divides the response range into four levels based on the JHX value, which correspond to the progressive response logic from "no intervention required" to "delayed takeover" and "immediate takeover". This can avoid unnecessary excessive system intervention and quickly ensure driving safety in high-risk situations, thereby improving the system's scenario adaptability and human-machine coordination. Attached Figure Description

[0056] Figure 1 This is a system flowchart of the present invention.

[0057] Figure 2 This is a diagram illustrating the method steps of the present invention.

[0058] In the diagram: 1. Image acquisition module; 11. Data acquisition unit; 12. Data preprocessing unit; 2. Target detection module; 21. Target coefficient generation unit; 22. Target coefficient analysis unit; 3. Data verification module; 31. Verification coefficient generation unit; 32. Verification coefficient analysis unit; 4. Human-computer interaction module; 41. Takeover coefficient generation unit; 42. Takeover coefficient analysis unit; 5. Feedback module. Detailed Implementation

[0059] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] Example 1: Please refer to Figure 1 An embedded system for vehicle-mounted cameras for nighttime intelligent driving includes an image acquisition module 1, a target detection module 2, a data verification module 3, a human-computer interaction module 4, and a feedback module 5.

[0061] Image acquisition module 1 is used to collect and preprocess data on the surrounding environment and vehicle conditions, and organize the data into a first data group and a second data group.

[0062] The target detection module 2 is used to calculate the first data group and the second data group to generate the target monitoring reference coefficient MBJ, and to analyze the target monitoring reference coefficient MBJ.

[0063] The data verification module 3 is used to calculate the first data group, the second data group, and the target monitoring reference coefficient to generate the data verification coefficient MBY, and to analyze the data verification coefficient MBY.

[0064] The human-computer interaction module 4 is used to calculate the target monitoring reference coefficient MBJ and the data verification coefficient MBY, thereby generating the human-computer interaction coefficient JHX, and analyzing the human-computer interaction coefficient JHX;

[0065] Feedback module 5 is used to execute takeover and alarm commands.

[0066] In this embodiment, the image acquisition module 1 is used to acquire images and data of the vehicle's surrounding environment and its own operating status, and to preprocess the raw data to improve the efficiency of subsequent recognition and analysis. Specifically, this module acquires information including ambient light intensity, obstacle distance, obstacle relative angle, lane departure, and road curvature, and organizes them into a first data group; simultaneously, it acquires information such as camera exposure, clarity, and vehicle speed, and organizes them into a second data group. Through grouped and structured processing, the image acquisition module 1 effectively improves data manageability and processing speed, enhances image quality and data integrity under complex lighting conditions at night, and provides high-quality input for subsequent module analysis and processing.

[0067] Target detection module 2 is used to perform fusion analysis on the first and second data sets to identify potential surrounding targets and assess their impact on driving safety. Based on a preset multi-parameter weighted model, this module comprehensively considers key parameters such as obstacle distance, angle, light intensity, and image quality to calculate and generate the target monitoring reference coefficient (MBJ). The MBJ is then evaluated and analyzed to preliminarily determine the threat level and plausibility of the target. By introducing multi-dimensional fusion judgment logic, target detection module 2 effectively improves the accuracy of target recognition at night, especially maintaining stable detection capabilities in low-light or image degradation scenarios, thereby enhancing the system's perception robustness in nighttime environments.

[0068] Data verification module 3 is used to perform cross-validation analysis on the first data group, the second data group, and the target monitoring reference coefficient MBJ to generate data verification coefficient MBY, and then further analyzes it. This module establishes a judgment model based on data consistency, completeness, and rationality to identify whether there are deviations, redundancies, or distortions in the camera data, thereby determining the credibility of the target detection results. By introducing a verification mechanism, data verification module 3 can effectively filter out misjudgments caused by nighttime imaging blur, sensor malfunctions, or short-term obstruction, improving the overall reliability of the system's judgment and providing solid data support for subsequent decisions.

[0069] The human-machine interaction module 4 couples the target monitoring reference coefficient MBJ with the data verification coefficient MBY to generate the human-machine interaction coefficient JHX, and then performs comprehensive analysis on it. JHX measures the balance between the risk level of the current driving environment and the system's confidence level. The human-machine interaction module 4 evaluates JHX in zones by setting multi-level response strategies to determine whether to prompt the driver to take over or issue a warning signal when necessary. By introducing a sensitivity adjustment mechanism and a coupling threshold model, this module enables the system to make reasonable judgments about the timing of takeover in dynamic nighttime environments, thereby optimizing human-machine collaboration efficiency and avoiding problems such as false triggering and delayed response.

[0070] Feedback module 5 is used to execute corresponding alarm or takeover operations based on the analysis results of the human-machine interaction coefficient JHX. This module issues voice, image, or tactile alarms to alert the driver to risks based on preset response levels, or automatically executes emergency takeover commands if the driver fails to respond in a timely manner. Feedback module 5 also supports tiered intervention strategies, such as providing early warnings when the risk is high but not yet critical, and directly switching to system takeover mode in critical situations, including vehicle deceleration, braking, or lane maintenance. This module realizes a closed-loop mechanism from risk analysis to control execution, enhancing the nighttime autonomous driving system's ability to respond to emergencies and its safety and reliability.

[0071] This invention provides an embedded vehicle camera system for nighttime intelligent driving. Through the organic linkage of multiple modules, it constructs a complete closed-loop architecture with capabilities for risk prediction, reliability verification, human-machine fusion decision-making, and emergency execution. Compared to existing driver assistance systems that rely solely on image recognition or single judgment logic, this system significantly improves the accuracy of target recognition, the stability of system judgment, and the timeliness of takeover response in low-light and complex traffic environments at night. It effectively overcomes the technical shortcomings of traditional solutions, such as misjudgment, slow response, or improper control at night.

[0072] Example 2: Please refer to Figure 1 The image acquisition module 1 includes a data acquisition unit 11 and a data preprocessing unit 12;

[0073] The data acquisition unit 11 is used to collect data on the surrounding environment and vehicle conditions, including ambient light intensity, camera exposure, driving speed, obstacle distance, relative position angle of obstacles, lane departure, camera clarity, and road curvature.

[0074] The data preprocessing unit 12 is used to preprocess and dimensionless the collected parameters, and reorganize them into a first data group and a second data group.

[0075] The first data set includes ambient light intensity (A), obstacle distance (B), obstacle relative position angle (C), lane departure (D), and road curvature (E).

[0076] The second data set includes camera exposure (F), camera resolution (G), and driving speed (H).

[0077] In this embodiment, the parameters collected by the data acquisition unit 11 not only cover external environmental information, but also integrate the camera's own working status parameters and vehicle operating status information, realizing dual perception of the driving environment and the status of the sensing device, effectively making up for the misjudgment problem caused by the lack of device self-test information in the traditional system.

[0078] The data preprocessing unit 12 preprocesses the aforementioned multi-source heterogeneous data, including operations such as denoising, normalization, and dimensionless transformation. Based on the data characteristics, it divides the data into a first data group and a second data group, enabling subsequent algorithms to perform differentiated processing for environmental risk factors and imaging quality / driving state factors, respectively. This structure not only improves the clarity of the data structure but also provides logically clear and well-defined data input for subsequent target detection and verification analysis modules, avoiding data redundancy and confusing information coupling.

[0079] Example 3: Please refer to Figure 1 The target detection module 2 includes a target coefficient generation unit 21 and a target coefficient analysis unit 22;

[0080] The target coefficient generation unit 21 is used to couple the first data group and the second data group by integrating ambient light intensity A, obstacle distance B, obstacle relative position angle C, lane deviation D, road curvature E, camera exposure F, camera clarity G, and driving speed H. Ambient light intensity A is calculated independently, obstacle distance B and driving speed H are coupled through an exponential function, obstacle relative position angle C, lane deviation D and road curvature E are coupled, and camera exposure F and camera clarity G are coupled. The multiple sets of calculations are integrated to finally generate the target monitoring reference coefficient MBJ.

[0081] The specific calculation formula is as follows:

[0082]

[0083] In the formula: A is the ambient light intensity, B is the distance to the obstacle, C is the relative position angle of the obstacle, D is the lane deviation, E is the road curvature, F is the camera exposure, G is the camera resolution, and H is the driving speed.

[0084] The target coefficient analysis unit 22 is used to analyze the target monitoring reference coefficient MBJ to generate a first analysis result. Based on the first analysis result, it is determined whether there are any suspicious obstacles in the area.

[0085] In this embodiment, the target detection module 2 includes a target coefficient generation unit 21 and a target coefficient analysis unit 22, which are used to accurately identify potential obstacles and assess their risk level in complex nighttime environments, thereby improving the environmental perception capability and reaction accuracy of the intelligent driving system in low-light scenarios.

[0086] The target coefficient generation unit 21 receives the first and second data sets output by the image acquisition module 1 and constructs the target monitoring reference coefficient MBJ based on a multi-dimensional parameter coupling mechanism. Specifically, the unit integrates and processes eight key parameters: ambient light intensity A, obstacle distance B, obstacle relative position angle C, lane deviation D, road curvature E, camera exposure F, camera sharpness G, and vehicle speed H. Ambient light intensity A is treated as an independent factor to reflect the impact of changes in nighttime external lighting on overall monitoring accuracy; obstacle distance B and vehicle speed H are coupled using an exponential function to reflect the nonlinear change in obstacle approach speed on driving risk at different vehicle speeds; obstacle relative position angle C, lane deviation D, and road curvature E are product-coupled to reflect the potential risk level under the combined influence of vehicle path deviation trends and road curve changes; exposure F and sharpness G jointly evaluate the camera imaging quality to quantify the system's perception accuracy of image content.

[0087] Through the coupling formula, the target coefficient generation unit 21 realizes the fusion modeling of multiple environmental variables and device parameters. The generated target monitoring reference coefficient MBJ can comprehensively reflect the potential obstacle risk level in the current monitoring area.

[0088] The target coefficient analysis unit 22 is used to analyze and process the generated MBJ, compare it with the set risk threshold or historical data trend, generate the first analysis result, and accordingly judge whether there are suspicious obstacles or abnormal conditions in the current monitoring area. This analysis process can effectively assist the system to achieve efficient and accurate obstacle recognition and early warning judgment under complex conditions such as insufficient night lighting, image degradation, or vehicle deviation from the track.

[0089] Example 4: Please refer to Figure 1 , the first analysis result is specifically as follows:

[0090] When MBJ < Y, it means that there are no potential obstacles in the current visible area;

[0091] When MBJ = Y, it means that there are potential obstacles in the current visible area and secondary analysis and calibration are required;

[0092] When MBJ > Y, it means that there are visible obstacles in the current visible area and early warning is required.

[0093] In this embodiment: By introducing an intermediate judgment state, the system can perform secondary analysis and calibration on image information with potential uncertainties, avoiding misrecognition or missed recognition caused by image noise or environmental interference, thereby improving the system's perception ability of weak and low-contrast obstacles.

[0094] Compared with the traditional binary decision method, this solution provides a three-level decision mechanism, enabling the system to adopt a hierarchical response strategy according to different risk levels, including continuous observation, secondary analysis, or immediate early warning, thus being closer to the complex and changeable visual environment in the real driving scenario.

[0095] By performing "on-demand" triggered in-depth analysis in the state of potential obstacles, it effectively avoids the system from performing high-intensity processing on all image data, saves computing resources, improves the overall operation efficiency and energy consumption performance of the system, and is especially suitable for embedded intelligent driving terminals or resource-constrained edge devices.

[0096] The MBJ coefficient, as an index that combines image features and risk parameters, can more sensitively reflect the image quality and abnormal changes in low-light environments. Combining with a multi-level threshold judgment mechanism, it can achieve early recognition and reasonable hierarchical response to night-time obstacles, thereby significantly enhancing the safety guarantee ability for night-time driving.

[0097] Example 5: Please refer to Figure 1The data verification module 3 includes a verification coefficient generation unit 31 and a verification coefficient analysis unit 32;

[0098] The verification coefficient generation unit 31 is used to couple the first data group, the second data group and the target monitoring reference coefficient MBJ. It couples the light intensity A, image clarity G and exposure F, then couples the obstacle distance B, the relative position angle of the obstacle C, the lane deviation D, the road curvature E and the camera exposure F, and then integrates the two sets of coupled data to finally generate the data verification coefficient MBY.

[0099] The specific calculation formula is as follows:

[0100]

[0101] Where: A is ambient light intensity, B is obstacle distance, C is the relative position angle of the obstacle, D is lane deviation, E is road curvature, F is camera exposure, G is camera resolution, H is driving speed, and MBJ is target monitoring reference coefficient.

[0102] The verification coefficient analysis unit 32 is used to analyze the data verification coefficient MBY, generate a second analysis result, and determine the rationality of the existence of the suspicious object based on the second analysis result.

[0103] In this embodiment, by coupling the light intensity A, image sharpness G, and camera exposure F, the effectiveness of the current image quality under actual optical conditions can be dynamically reflected, avoiding misjudgments caused by overexposure, underexposure, or image blurring, thereby improving the system's judgment stability under complex lighting conditions.

[0104] By jointly modeling obstacle distance (B), relative angle (C), lane deviation (D), road curvature (E), and driving speed (H) with MBJ to generate validation coefficients (MBY), the actual threat level of obstacles under the current driving conditions can be more comprehensively reflected, significantly improving the contextual understanding ability of obstacle recognition and reducing false alarm and false negative rates.

[0105] By calculating the verification coefficient MBY through a formulaic modeling method, the mathematical integration of environmental parameters and motion state is achieved, enabling the system to dynamically adjust the identification weights and providing more reliable basic data support for subsequent early warning decisions.

[0106] The verification coefficient analysis unit 32 generates a second analysis judgment based on the analysis results of MBY. As a means of verifying and correcting the initial judgment result of MBJ, it helps to filter out the boundary ambiguity area or low confidence target in the initial judgment, thereby further improving the system's recognition credibility and risk differentiation ability of suspicious obstacles.

[0107] The above technical solution can be applied to an autonomous driving system in urban roads, highway scenarios, night driving, and complex weather conditions, which helps to construct a perception and judgment framework with strong environmental perception, intelligent decision-making, and safety redundancy capabilities.

[0108] Embodiment Six: Please refer to Figure 1 , and the specific second analysis result is as follows:

[0109] When MBY < R, it means that the current data is not within the credible interval and no marking is performed;

[0110] When MBY = R, it means that the current data is within the low credible interval, and marking is performed and re-analysis is prepared;

[0111] When MBY > R, it means that the current data is within the high credible interval and alarm preparation is ready.

[0112] In this embodiment: By setting the "low credible interval", "high credible interval", and "uncredible interval", the system can perform sub-division processing on the data quality of the analysis result, avoiding directly using abnormally fluctuating or boundary-blurred data for decision-making, thereby reducing the false alarm and false trigger rates.

[0113] When MBY falls within the low credible interval, the system not only marks the data but also prepares to enter the re-analysis process, reflecting that the system has a self-checking and information compensation mechanism, which is beneficial to delaying decision-making to obtain more context support when the information is incomplete or of poor quality.

[0114] When MBY > R, that is, the data is within the high credible interval, the system can directly trigger the alarm preparation process, achieving a rapid response to high-confidence obstacles or risk states, effectively shortening the delay from recognition to response, and improving the real-time performance and safety of the system.

[0115] The second analysis result is based on MBY after the first analysis and data verification, realizing unified closed-loop processing at the data level and the perception level, ensuring that only after sufficient verification can it enter the alarm or intervention process, and improving the judgment logic integrity and safety robustness of the entire system. <​​​​​​​​​The takeover coefficient generation unit 41 is used to couple the target monitoring reference coefficient MBJ and the data verification coefficient MBY to finally generate the human-computer interaction coefficient JHX;

[0119] The specific calculation formula is as follows:

[0120] JHX=a1×MBJ+a2×MBY+a3×(MBJ+MBY);

[0121] In the formula: MBJ is the target monitoring reference coefficient, MBY is the data verification coefficient, and a1, a2 and a3 are weight values, and the values ​​of a1, a2 and a3 are adjusted and set by the user;

[0122] The takeover coefficient analysis unit 42 is used to perform data analysis, generate a third analysis result, and determine whether an alarm or emergency takeover is required based on the third analysis result.

[0123] The specific results of the third analysis are as follows:

[0124] When JHX < 0.4, it indicates that the current operation is stable, the risk is reduced, and the system will not intervene or issue an alarm.

[0125] When 0.4≤JHX<0.8, it indicates that the current operational risk has increased and there is a Level 1 risk. The system will issue a voice alarm and take over after 10 seconds.

[0126] When JHX ≥ 0.8, it indicates that the current operation is at extremely high risk, with a level 2 risk. The system will immediately take over and send information to the emergency contact.

[0127] In this embodiment, by weighted coupling of the target monitoring reference coefficient MBJ and the data verification coefficient MBY, a unified human-computer interaction coefficient JHX is formed. This coefficient can comprehensively reflect multiple factors such as environmental risk, image quality, obstacle characteristics, and data credibility, effectively quantifying the risk level of the current driving state and providing an efficient and accurate basis for takeover decisions.

[0128] The formula for generating the takeover coefficient incorporates weight parameters and allows users to flexibly adjust them according to actual driving needs or system scenarios, enhancing the controllability and scenario transferability of the algorithm model and facilitating adaptive optimization under different vehicle models, driving styles, or operating preferences.

[0129] This system divides the response range into three levels based on the JHX value, which correspond to the progressive response logic from "no intervention required" to "delayed takeover" and "immediate takeover". This can avoid unnecessary excessive system intervention and quickly ensure driving safety in high-risk situations, thereby improving the system's scenario adaptability and human-machine coordination.

[0130] When the system identifies a Level 2 risk, it can provide a risk warning to the driver via voice and complete the takeover preparation within a set time delay. When a Level 3 risk is identified, the system will immediately initiate automatic takeover and proactively send an alert to the preset emergency contact, which helps to build a complete closed-loop safety mechanism of pre-event warning, in-event intervention and post-event notification.

[0131] Due to the limited perception capabilities and higher risk response requirements in nighttime environments, this invention achieves rapid identification and response to sudden obstacles, image distortion, and abnormal system states at night through a refined hierarchical judgment mechanism of the takeover coefficient, significantly improving the stability and safety of nighttime intelligent driving systems.

[0132] This application also includes a method for embedding an in-vehicle camera for nighttime intelligent driving, the specific steps of which are as follows:

[0133] Step 1: Collect and preprocess data on the surrounding environment and vehicle conditions using image acquisition module 1, and organize the data into a first data group and a second data group.

[0134] Step 2: Calculate the first and second data groups using the target detection module 2 to generate the target monitoring reference coefficient MBJ, and then analyze the target monitoring reference coefficient MBJ.

[0135] Step 3: Calculate the first data group, the second data group, and the target monitoring reference coefficient through the data verification module 3 to generate the data verification coefficient MBY, and then analyze the data verification coefficient MBY.

[0136] Step 4: Calculate the target monitoring reference coefficient MBJ and data verification coefficient MBY through the human-computer interaction module 4 to generate the human-computer interaction coefficient JHX, and then analyze the human-computer interaction coefficient JHX.

[0137] Step 5: Execute takeover and alarm commands through feedback module 5.

[0138] The contents not described in detail in this specification are existing technologies known to those skilled in the art.

[0139] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An embedded system for vehicle-mounted cameras for intelligent night driving, characterized in that: It includes an image acquisition module, a target detection module, a data verification module, a human-computer interaction module, and a feedback module; The image acquisition module is used to collect and preprocess data on the surrounding environment and vehicle conditions, and organize them into a first data group and a second data group. When MBJ=Y, it means that there is a potential obstacle in the current visible area, and secondary analysis and calibration are required. The target detection module is used to calculate the first data group and the second data group to generate the target monitoring reference coefficient MBJ, and to analyze the target monitoring reference coefficient MBJ. The data verification module is used to calculate the first data group, the second data group, and the target monitoring reference coefficient to generate the data verification coefficient MBY, and to analyze the data verification coefficient MBY. The data verification module includes a verification coefficient generation unit and a verification coefficient analysis unit; The verification coefficient generation unit is used to couple the first data group, the second data group, and the target monitoring reference coefficient MBJ. It couples the light intensity A, image clarity G, and exposure F, and then couples the obstacle distance B, the relative position angle of the obstacle C, the lane deviation D, the road curvature E, and the camera exposure F. Finally, it integrates and calculates the two sets of coupled data to generate the data verification coefficient MBY. The specific calculation formula is as follows: Where: A is ambient light intensity, B is obstacle distance, C is the relative position angle of the obstacle, D is lane deviation, E is road curvature, F is camera exposure, G is camera resolution, H is driving speed, and MBJ is target monitoring reference coefficient. The verification coefficient analysis unit is used to analyze the data verification coefficient MBY, generate a second analysis result, and determine the rationality of the existence of the suspicious item based on the second analysis result. When MBY = R, it means that the current data is in the low confidence interval, and it is marked and prepared for re-analysis. The human-computer interaction module is used to calculate the target monitoring reference coefficient MBJ and the data verification coefficient MBY, thereby generating the human-computer interaction coefficient JHX, and to analyze the human-computer interaction coefficient JHX. The feedback module is used to execute takeover and alarm commands.

2. The vehicle-mounted camera embedded system for nighttime intelligent driving according to claim 1, characterized in that: The image acquisition module includes a data acquisition unit and a data preprocessing unit; The data acquisition unit is used to collect data on the surrounding environment and vehicle conditions, including ambient light intensity, camera exposure, driving speed, obstacle distance, relative position angle of obstacles, lane deviation, camera clarity, and road curvature. The data preprocessing unit is used to preprocess and dimensionless the collected parameters, and reorganize them into a first data group and a second data group. The first data set includes ambient light intensity (A), obstacle distance (B), obstacle relative position angle (C), lane departure (D), and road curvature (E). The second data set includes camera exposure (F), camera resolution (G), and driving speed (H).

3. The vehicle-mounted camera embedded system for nighttime intelligent driving according to claim 2, characterized in that: The target detection module includes a target coefficient generation unit and a target coefficient analysis unit; The target coefficient generation unit is used to couple the first data group and the second data group. By integrating the environmental light intensity A, the obstacle distance B, the relative position angle C of the obstacle, the lane deviation D, the road curvature E, the camera exposure F, the camera clarity G, and the driving speed H, where the environmental light intensity A is calculated independently, the obstacle distance B and the driving speed H are coupled through an exponential function, the relative position angle C of the obstacle, the lane deviation D, and the road curvature E are coupled, the camera exposure F and the camera clarity G are coupled, and multiple groups of calculations are integrated to finally generate the target monitoring reference coefficient MBJ; The specific calculation formula is as follows: In the formula: A is the environmental light intensity, B is the obstacle distance, C is the relative position angle of the obstacle, D is the lane deviation, E is the road curvature, F is the camera exposure, G is the camera clarity, and H is the driving speed; The target coefficient analysis unit is used to analyze the target monitoring reference coefficient MBJ, thereby generating a first analysis result. According to the first analysis result, it is judged whether there are suspicious obstacles in the area.

4. The vehicle-mounted camera embedded system for nighttime intelligent driving according to claim 3, characterized in that: The specific content of the first analysis result is as follows: When MBJ < Y, it means that there are no potential obstacles in the current visible area; When MBJ = Y, it means that there are potential obstacles in the current visible area and secondary analysis and calibration are required; When MBJ > Y, it means that there are visible obstacles in the current visible area and early warning is required.

5. The vehicle-mounted camera embedded system for nighttime intelligent driving according to claim 4, characterized in that: The specific content of the second analysis result is as follows: When MBY < R, it means that the current data is not in the credible interval and no marking is performed; When MBY = R, it means that the current data is in the low credible interval, marking is performed and re-analysis is prepared; When MBY > R, it means that the current data is in the high credible interval and alarm preparation is performed.

6. The vehicle-mounted camera embedded system for nighttime intelligent driving according to claim 5, characterized in that: The human-computer interaction module includes a takeover coefficient generation unit and a takeover coefficient analysis unit; The takeover coefficient generation unit is used to couple the target monitoring reference coefficient MBJ and the data verification coefficient MBY, and finally generate the human-computer interaction coefficient JHX; The specific calculation formula is as follows: JHX = a1 × MBJ + a2 × MBY + a3 × (MBJ + MBY); In the formula: MBJ is the target monitoring reference coefficient, MBY is the data verification coefficient, a1, a2, and a3 are weight values, and the values of a1, a2, and a3 are adjusted and set by the user; The takeover coefficient analysis unit is used to perform data analysis and generate a third analysis result. According to the third analysis result, it is judged whether alarm and emergency takeover are required.

7. The vehicle-mounted camera embedded system for nighttime intelligent driving according to claim 6, characterized in that: The specific content of the third analysis result is as follows: When JHX < 0.4, it means that the current operation is stable, the risk is reduced, and the system does not intervene or alarm; When 0.4 ≤ JHX < 0.8, it means that the current operation risk is increased, there is a primary risk, the system gives a voice prompt alarm, and takes over after 10 seconds; When JHX ≥ 0.8, it means that the current operation risk is extremely high, there is a secondary risk, the system immediately executes takeover and sends information to the emergency contact.

8. A method for embedding an in-vehicle camera for nighttime intelligent driving, characterized in that: The specific steps are as follows: S1. The image acquisition module collects and preprocesses data on the surrounding environment and vehicle conditions, and organizes them into a first data group and a second data group. S2. The target detection module calculates the first data group and the second data group to generate the target monitoring reference coefficient MBJ, and analyzes the target monitoring reference coefficient MBJ. When MBJ = Y, it means that there is a potential obstacle in the current visible area, and secondary analysis and calibration are required. S3. The data verification module calculates the first data group, the second data group, and the target monitoring reference coefficient to generate the data verification coefficient MBY, and then analyzes the data verification coefficient MBY. The data validation module includes a validation coefficient generation unit and a validation coefficient analysis unit; The verification coefficient generation unit is used to couple the first data group, the second data group, and the target monitoring reference coefficient MBJ. It couples the light intensity A, image clarity G, and exposure F, and then couples the obstacle distance B, the relative position angle of the obstacle C, the lane deviation D, the road curvature E, and the camera exposure F. Finally, it integrates and calculates the two sets of coupled data to generate the data verification coefficient MBY. The specific calculation formula is as follows: Where: A is ambient light intensity, B is obstacle distance, C is the relative position angle of the obstacle, D is lane deviation, E is road curvature, F is camera exposure, G is camera resolution, H is driving speed, and MBJ is target monitoring reference coefficient. S4. Calculate the target monitoring reference coefficient MBJ and data verification coefficient MBY through the human-computer interaction module to generate the human-computer interaction coefficient JHX, and analyze the human-computer interaction coefficient JHX. S5. Execute takeover and alarm commands through the feedback module.

Images