Tunnel portal dynamic light and signal control system based on visual algorithm

Abstract

This invention discloses a dynamic lighting and signal control system for tunnel entrances based on visual algorithms, belonging to the field of intelligent manufacturing control technology, to solve the problem of poor lighting control when the field of vision changes abruptly at tunnel entrances. This invention achieves continuous monitoring and intelligent control of the driver's visual adaptation state at tunnel entrances by constructing a dynamic perception mechanism based on visual feature understanding and multi-strategy linkage. First, multi-dimensional visual features are extracted through image acquisition and preprocessing to generate a visual risk score. Combined with a sliding window to identify abrupt changes in the field of vision and traffic anomalies, risk response commands are output to drive illumination adjustment and signal linkage control. Then, by integrating vehicle behavior continuity, parking probability, and blind spot occlusion factors, speed limit, guidance, or warning signals are dynamically issued. Based on execution feedback, the visual model is fine-tuned and the strategy is updated. Finally, combined with a control scenario library, adaptive optimization of lighting and signals across time periods and multiple environments is achieved, improving traffic safety and control intelligence.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing control technology, and more specifically, to a dynamic lighting and signal control system for tunnel entrances based on visual algorithms. Background Technology

[0002] Currently, tunnel traffic systems are an important part of urban trunk roads and highway transportation infrastructure. Their operational safety and traffic efficiency directly affect the overall quality of the road. In most tunnel entrance areas, due to the significant difference in brightness between external natural light and tunnel interior lighting, especially on sunny days, in backlight, and in the early morning and evening, drivers are very likely to experience visual adaptation difficulties when entering the tunnel, resulting in momentary blurred vision or even a "black hole effect," which has become a high-incidence cause of traffic accidents.

[0003] The shortcomings of existing technology are as follows: When a vehicle enters the tunnel entrance section at high speed from a bright outdoor environment, the traditional system only makes judgments based on the global illumination value obtained by the brightness sensor. It fails to comprehensively analyze visual influencing factors such as vehicle density, traffic speed, incident light angle and reflection interference. This results in the entrance section light intensity not being adjusted in time, or the adjustment curve not matching the actual driving visual state. Especially in complex lighting environments such as backlighting, strong glass reflection, and the boundary between light and shadow, it is difficult to issue auxiliary signals or improve the lighting level of the entrance area in time. This directly triggers a chain reaction from driver visual maladaptation to slow operation and then to increased risk, which in turn affects the safe driving status of vehicles at the tunnel entrance. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the following solution is proposed to solve the problem of poor lighting control during sudden changes in the field of vision at the tunnel entrance in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] The tunnel entrance dynamic lighting and signal control system based on vision algorithms includes a tunnel entrance image acquisition module, a response command generation module, an illumination adjustment module, a control command generation module, an update input module, and a joint control module. The modules are connected by signals.

[0007] The tunnel entrance image acquisition module acquires a continuous sequence of image frames of the tunnel entrance area, and performs defogging, contrast enhancement and lane normalization processing to extract the visual features of the image frames.

[0008] The response instruction generation module tensors and encodes the extracted visual features, calculates the driving visual adaptation risk score in the built-in visual adaptation model, and combines the sliding time window to detect short-term visual abrupt change areas and vehicle speed decrease sections, and outputs a risk response instruction candidate set.

[0009] The illuminance adjustment module constructs a lighting control state map based on the risk response command candidate set, extracts the matching relationship between traffic flow characteristics and illumination gradient, and maps and generates the entrance section lighting illuminance adjustment parameter set.

[0010] The control command generation module synchronously constructs signal linkage triggering factors, and combines the continuity of the entry behavior, the probability of the vehicle stopping and the degree of visual blind spot occlusion to determine whether to generate speed limit, guidance or warning signal control commands.

[0011] The update input module sends the lighting illuminance adjustment parameter group and signal control commands to the field control equipment for execution. At the same time, it collects the execution results and image change sequences, and extracts the lighting response deviation and behavior trajectory offset as the update input of the visual adaptation model.

[0012] The joint control module, based on behavioral feedback results, uses a fine-tuning mechanism to update the brightness response mapping relationship and linkage strategy rules in the visual adaptation model, and adaptively optimizes the joint control of light and signal in different time periods and environments.

[0013] Furthermore, a continuous sequence of image frames of the tunnel entrance area was acquired, and dehazing, contrast enhancement, and lane normalization were performed to extract the visual features of the image frames, including:

[0014] Image frames are dehazed using dark channel prior and histogram equalization algorithms;

[0015] A contrast-limited adaptive histogram equalization method is used to enhance the contrast of image frames.

[0016] Based on image perspective correction and lane template matching algorithms, lane normalization processing is performed on image frames to convert lane structures from different perspectives into a unified spatial reference system.

[0017] After preprocessing, visual features such as vehicle distribution, traffic speed, lane occupancy, and light incidence angle are extracted from the image frames.

[0018] Furthermore, the extracted visual features are tensor-encoded and used to calculate a driving visual adaptation risk score within the built-in visual adaptation model, including:

[0019] A structured input matrix is ​​constructed based on visual features, and position embedding and temporal encoding are introduced to generate a visual feature tensor with spatiotemporal continuity.

[0020] Visual feature tensors are input into a multi-layer convolutional attention network structure for feature compression, and high-response feature vectors of the driving behavior attention region are extracted.

[0021] Based on a preset sensitivity function for brightness abrupt changes and a threshold for dynamic speed changes, a visual adaptation risk scoring function is constructed, and the driving visual adaptation risk score corresponding to the current image frame is calculated through feature vectors and a risk mapping model.

[0022] The driving visual adaptation risk score reflects the probability of instantaneous visual impairment caused by the combination of changes in brightness, traffic density and entry angle under the current traffic conditions.

[0023] Furthermore, based on a preset sensitivity function for abrupt changes in brightness and a dynamic speed change threshold, a visual adaptation risk scoring function is constructed. This function calculates the driving visual adaptation risk score corresponding to the current image frame using feature vectors and a risk mapping model, including:

[0024] Based on the brightness gradient change of the corresponding area of ​​the vehicle in the image frame, a brightness abrupt change sensitivity function is constructed. The brightness abrupt change sensitivity function is used to measure the spatial gradient intensity and temporal gradient change rate of brightness in the forward region of the vehicle, and the boundary response value of the brightness variation block is extracted by local region differential convolution.

[0025] By combining the vehicle speed change sequence, a dynamic speed change threshold function is defined and used to determine the deceleration trend, speed fluctuation amplitude and instantaneous deceleration acceleration of the vehicle between adjacent image frames, and to identify the coupling signal between changes in driving behavior and visual scene switching.

[0026] A visual adaptation risk scoring function is constructed, which combines the intensity of brightness abrupt change, speed change index and incident light angle into a joint input feature vector. The vector product mapping is performed by the risk mapping model and the preset scoring weight matrix to calculate the driving visual adaptation risk score of the current image frame.

[0027] Furthermore, by combining the detection of short-term visual abrupt changes and vehicle speed reduction sections using a sliding time window, a candidate set of risk response instructions is output, including:

[0028] The driving visual adaptation risk scores in consecutive image frames are input into a fixed-length sliding time window in chronological order. The gradient change value and peak duration of driving visual adaptation within the window are calculated to identify image areas with sudden brightness changes or uneven lighting.

[0029] The vehicle speed change sequence corresponding to each image frame within the window is extracted synchronously. Combined with the dynamic speed change threshold function, it is determined whether there is deceleration behavior or speed fluctuation characteristics exceeding the set threshold, and marked as potential traffic interference sections.

[0030] The brightness abrupt change area is time-synchronized with the vehicle speed decrease section to construct risk event trigger tags, and the trigger tags are filtered according to the risk score threshold to output a candidate set of risk response instructions.

[0031] Furthermore, the illuminance adjustment module includes:

[0032] The risk level, trigger location and suggested response type in the risk response instruction candidate set are mapped to the preset lighting response state space to construct a ternary state map containing vehicle distribution density, vehicle speed trend and illumination gradient change rate.

[0033] In the ternary state map, traffic flow features corresponding to the trigger position are extracted, including the number of vehicles per unit length, average vehicle speed and speed variance, as well as the light intensity gradient and brightness change direction of the corresponding region of the image frame.

[0034] Based on the matching relationship between traffic flow characteristics and illumination gradient, the illuminance adjustment rule library is called to match the corresponding lighting response level and generate the entrance section light illuminance adjustment parameter group.

[0035] The entrance section lighting illuminance adjustment parameter group is used to drive the on-site lighting fixtures to perform segmented illuminance adjustment, including target illuminance value, illuminance gradient step size, gradient duration, and effective range label.

[0036] Furthermore, the control command generation module includes:

[0037] Extract the travel trajectory of the target vehicle from consecutive image frames, calculate the continuity index of the vehicle's entry behavior into the tunnel, and identify whether there is sudden deceleration, abnormal lane change or trajectory interruption before entering the tunnel.

[0038] Based on the vehicle speed sequence within the sliding time window and combined with historical deceleration behavior data, a short-term stopping probability model is constructed to generate a corresponding vehicle stopping probability score for each vehicle in the current traffic flow.

[0039] Occlusion analysis is performed on the visual field area of ​​each lane in the image frame. The visual blind spot occlusion degree score corresponding to the current frame is calculated by the spatial projection relationship between the overlapping area of ​​the vehicle boundary, the size of the blind spot and the incident light angle.

[0040] The system integrates the continuity index of entry behavior, the vehicle stopping probability score, and the visual blind spot occlusion score to construct a signal linkage trigger factor for decision-making. It outputs whether to generate speed limit, guidance, or warning signal control commands and selects the corresponding signal type and range of action based on the decision result.

[0041] Furthermore, the execution results and image change sequences are simultaneously collected, and the illumination response bias and behavioral trajectory offset are extracted as update inputs for the visual adaptation model, including:

[0042] The execution results of the lighting illuminance adjustment parameter group and signal control command are collected and compared with the corresponding image frames before execution to generate an image change sequence after the lighting response.

[0043] Based on the average brightness, gradient distribution, and lane boundary clarity indices of the tunnel entrance area in the image change sequence, the lighting response deviation between the current light output and the preset target illuminance is calculated.

[0044] Extract the travel trajectory of the target vehicle before and after the execution control, and calculate the vehicle behavior trajectory deviation index based on the time alignment offset between trajectory point sets, trajectory shape change rate and lane center deviation distance;

[0045] The lighting response deviation and the behavior trajectory offset index are combined into a joint feedback vector, which is then input into the visual adaptation model as incremental update data.

[0046] Furthermore, based on the behavioral feedback results, a fine-tuning mechanism is used to update the brightness response mapping relationship and linkage strategy rules in the visual adaptation model, including:

[0047] The joint feedback vector composed of lighting response deviation and behavior trajectory deviation index is compared with the historical response state of the model to determine the dynamic convergence trend between the current control effect and the prediction deviation.

[0048] When insufficient lighting response or vehicle deviation exceeds the tolerance threshold occurs within multiple consecutive time windows, the local fine-tuning mechanism of the visual adaptation model is triggered, and the mapping parameters of the affected area are updated.

[0049] The gradient descent method is used to refine the illuminance adjustment weights in the luminance response mapping relationship, thereby reducing the error between the actual lighting output and the desired visual state.

[0050] At the same time, the trigger threshold, scope of action, and signal type selection conditions in the signal linkage strategy rules are updated to construct a new round of control strategy groups that adapt to the current traffic behavior pattern.

[0051] Furthermore, adaptive optimization is performed on the joint control of lighting and signals under different time periods and environments, including:

[0052] Based on the currently detected time period information and environmental status, the optimal control template is matched in the control scenario library, and a dynamic adjustment strategy is generated by combining the latest fine-tuned model parameters.

[0053] By invoking the illuminance level boundary, signal type switching threshold, and linkage delay rules in the matching control template, adaptive optimization of the joint control of light and signal under different scenarios can be achieved.

[0054] The technical effects and advantages of the vision algorithm-based dynamic lighting and signal control system for tunnel entrances in this invention are as follows:

[0055] This invention achieves continuous monitoring and intelligent control of driver visual adaptation status under complex traffic and lighting conditions at tunnel entrances by constructing a dynamic perception mechanism based on visual feature understanding and multi-strategy linkage. Based on continuous image acquisition and high-quality preprocessing, multi-dimensional visual features such as vehicle distribution, speed trends, lane occupancy, and illumination incident angle are extracted. A driver visual adaptation risk score is generated through tensor coding and convolutional attention mechanisms. Combined with a sliding window to dynamically detect abrupt changes in the field of view and abnormal traffic behavior, a candidate set of risk response commands is output. Based on this, the system constructs a lighting control state map, mines the matching relationship between traffic flow characteristics and illumination gradients, and generates a set of segmented illuminance adjustment parameters with gradual gradient changes, improving the response accuracy and transition smoothness of lighting control.

[0056] Meanwhile, by integrating dynamic factors such as vehicle trajectory continuity, parking probability, and blind spot occlusion, a signal linkage triggering mechanism is constructed to realize the real-time release of speed limit, guidance, and warning signals. Combining post-control images and trajectory feedback, lighting response deviation and behavioral trajectory offset indicators are extracted to generate a joint feedback vector and drive the local fine-tuning of the visual adaptation model, completing the online optimization and update of brightness mapping relationship and linkage strategy rules. Finally, through the control scene library and multi-environment label matching mechanism, an adaptive joint control strategy is generated across time periods and under multiple conditions, significantly improving the traffic safety, lighting accuracy, and signal control intelligence level of tunnel entrances in dynamic traffic flow and complex lighting environments. Attached Figure Description

[0057] Figure 1 This is a schematic diagram of the structure of the tunnel entrance dynamic lighting and signal control system based on vision algorithms of the present invention. Detailed Implementation

[0058] 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.

[0059] In order to achieve the above objectives, Figure 1 A schematic diagram of the structure of the tunnel entrance dynamic lighting and signal control system based on vision algorithm of the present invention is given. Specifically, it includes a tunnel entrance image acquisition module, a response command generation module, an illuminance adjustment module, a control command generation module, an update input module, and a joint control module. The modules are connected to each other through signals.

[0060] The tunnel entrance image acquisition module acquires a continuous sequence of image frames of the tunnel entrance area, and performs defogging, contrast enhancement and lane normalization processing to extract the visual features of the image frames.

[0061] The response instruction generation module tensors and encodes the extracted visual features, calculates the driving visual adaptation risk score in the built-in visual adaptation model, and combines the sliding time window to detect short-term visual abrupt change areas and vehicle speed decrease sections, and outputs a risk response instruction candidate set.

[0062] The illuminance adjustment module constructs a lighting control state map based on the risk response command candidate set, extracts the matching relationship between traffic flow characteristics and illumination gradient, and maps and generates the entrance section lighting illuminance adjustment parameter set.

[0063] The control command generation module synchronously constructs signal linkage triggering factors, and combines the continuity of the entry behavior, the probability of the vehicle stopping and the degree of visual blind spot occlusion to determine whether to generate speed limit, guidance or warning signal control commands.

[0064] The update input module sends the lighting illuminance adjustment parameter group and signal control commands to the field control equipment for execution. At the same time, it collects the execution results and image change sequences, and extracts the lighting response deviation and behavior trajectory offset as the update input of the visual adaptation model.

[0065] The joint control module, based on behavioral feedback results, uses a fine-tuning mechanism to update the brightness response mapping relationship and linkage strategy rules in the visual adaptation model, and adaptively optimizes the joint control of light and signal in different time periods and environments.

[0066] The tunnel entrance image acquisition module acquires a continuous sequence of image frames of the tunnel entrance area, performs dehazing, contrast enhancement, and lane normalization processing, and extracts the visual features of the image frames, including the following:

[0067] The tunnel entrance image acquisition equipment is deployed on fixed supports above or to both sides of the tunnel entrance area. It can acquire a continuous sequence of image frames covering the entire traffic lane. The image acquisition frequency is set according to the traffic flow density, with a typical frame rate of 25 frames per second. The image frames are RGB color images with a resolution of no less than 1920×1080 pixels. After the image frames are acquired, image preprocessing operations are performed sequentially, including dehazing, contrast enhancement, and lane normalization, as detailed below:

[0068] An image dehazing algorithm based on dark channel prior is adopted. First, the minimum value of each pixel in each image frame in the three RGB color channels is calculated to generate a dark channel image. Then, a local minimum filter is used to calculate the local minimum region of the dark channel image to estimate the atmospheric light value. Then, the reflectivity of the original image is restored by establishing an image restoration model to output a clear image with approximately true brightness. In order to further improve the processing effect, a histogram equalization step is introduced to stretch the image brightness distribution and expand the details in the dark area, so that the vehicle outline and boundary in the image frame are clearer under low visibility or shadow conditions, thereby enhancing the image clarity.

[0069] To address the issue of overall low brightness or uneven distribution of bright and dark areas in image frames, a Contrast-Limited Adaptive Histogram Equalization (CLAHE) algorithm is employed for image enhancement. This method divides the image into multiple local regions and performs histogram equalization on each region separately to enhance local contrast. A contrast limit threshold is set during this process to prevent over-enhancement; a typical threshold range is between 2.0 and 4.0. This processing step effectively improves the grayscale gradient of areas such as vehicle edges, lane lines, and traffic signs, thereby enhancing the saliency of image edge features and improving the visual separation between lane edges and vehicle contours, facilitating subsequent target extraction and behavior discrimination.

[0070] Lane normalization is performed to maintain image analysis consistency under varying camera installation angles, tunnel structures, and viewing angles. An image perspective correction combined with a lane template matching algorithm is used to normalize the image frames. The specific steps are as follows: First, the perspective transformation matrix is ​​extracted from fixed road marking points in the image. Then, the original image is homography-transformed using the perspective transformation matrix to eliminate tilt and distortion, ensuring that lane lines are perfectly parallel in the transformed image. Finally, the corrected image is matched against a preset standard lane template to ensure that the lane areas are in a unified spatial coordinate reference system, thus achieving processing consistency across different scenarios and improving algorithm generalization capabilities.

[0071] After completing the above image preprocessing, multidimensional visual features for subsequent risk scoring are extracted from each frame. The extracted features include, but are not limited to:

[0072] The YOLOv5 object detection network was used to identify and label vehicle targets in the image frames, and the position coordinates (center point coordinates and bounding box size) and lane number of each vehicle were extracted.

[0073] Based on multi-frame target tracking algorithms (such as DeepSORT), the vehicle displacement and time interval in consecutive image frames are calculated to obtain the vehicle velocity vector, and the velocity change between the current frame and the previous frame is recorded.

[0074] The lanes are divided into fixed grid areas, and the percentage of grids in each lane that are obscured by vehicles is counted to represent the current traffic density.

[0075] By combining the vehicle shadow direction and the gradient direction of the bright part boundary in the image frame, the projection angle of the main direction of the light source in the lane coordinate system is calculated using a projection geometry method to reflect the incident influence of the current external illumination of the tunnel on the driver's field of vision.

[0076] In summary, by acquiring continuous image frames and performing multi-dimensional image preprocessing (including dehazing, contrast enhancement, and lane normalization), clear and structurally consistent image inputs can be extracted under different weather conditions, lighting conditions, and camera angles. This provides a reliable data foundation for subsequent visual feature analysis and significantly improves image interpretability.

[0077] The response command generation module tensors the extracted visual features and calculates a driving visual adaptation risk score within the built-in visual adaptation model. It then uses a sliding time window to detect short-term visual abrupt changes and vehicle speed decreases, outputting a candidate set of risk response commands, as detailed below:

[0078] The system receives visual feature data extracted by the tunnel entrance image acquisition module. These visual features include multi-dimensional information such as vehicle distribution, traffic speed, lane occupancy, and incident light angle. First, the visual features are temporally encoded at fixed frame intervals to construct a structured input matrix. Each row of the structured input matrix corresponds to a visual feature type, and each column corresponds to a spatial location unit within an image frame after grid division. To maintain the integrity of spatial information representation, a location embedding vector is introduced for each location unit in the input matrix. This vector is encoded based on the unit's position in two-dimensional spatial coordinates within the image frame. Simultaneously, to characterize the changes in visual features over time in a multi-frame image sequence, temporal encoding is introduced, assigning each frame a time step number and expressing the temporal dimension using sine / cosine functions or learnable embedding vectors. Through the dual injection of spatial location embedding and temporal encoding, a visual feature tensor with a clear spatiotemporal sequence structure is constructed. This visual feature tensor is a five-dimensional structure, corresponding to the batch dimension, number of time frames, number of feature channels, spatial height, and spatial width, respectively, possessing strong temporal representation capabilities for traffic conditions.

[0079] The generated visual feature tensor is input into a multi-layer convolutional attention network in the visual adaptation model for feature compression and focused representation. The network structure includes multiple 3D convolutional layers to simultaneously extract temporal variation and spatial pattern features. The channel attention module is used to score and weight the responsiveness of different feature channels to improve the responsiveness of channels sensitive to driving behavior. The spatial attention module, by analyzing the response differences between spatial locations in the tensor, locates regions in the current frame image that significantly affect driving decisions (such as the edge of the field of vision, the critical area of ​​tunnel entrances, the front edge of blind spots, etc.). The extracted high-response region features are aggregated into a unified feature vector to form a high-dimensional behavioral representation vector of the driving state. Subsequently, this feature vector is input into the visual adaptation risk score, and the risk score function is calculated to generate a risk score value that reflects the driver's visual stress or maladaptation level under the current traffic conditions. The risk score value is a continuous floating-point number between 0 and 1, where a higher value indicates a greater impact on the driver's visual adaptation due to factors such as sudden changes in brightness, increased traffic density, or unreasonable tunnel entry angles in the driving environment; a lower score value indicates a stable current visual environment and good vehicle traffic conditions.

[0080] The construction of the visual adaptation risk scoring function is based on three key input features: the intensity of brightness abrupt change in the forward region of the vehicle in the image frame, the vehicle speed change index, and the spatial projection information of the incident angle of light in the image frame.

[0081] The brightness abrupt change sensitivity function is used to measure the instantaneous impact of brightness changes on the driver's vision;

[0082] The brightness abrupt change sensitivity function is constructed as follows: First, the brightness value distribution within a specified area in front of the target vehicle (e.g., 5 to 15 meters in front of the vehicle) is extracted from the image frame, and this area is divided into several sub-grid units. Next, the average brightness value of each sub-grid between adjacent frames is differentially processed to calculate the spatial and temporal gradients of brightness. The spatial gradient represents the degree of brightness abrupt change in adjacent areas within a given frame, and the temporal gradient represents the rate of brightness change in that area between adjacent frames. Subsequently, local region differential convolution operations (i.e., convolution kernels that calculate brightness gradients within a 3×3 pixel range) are used to extract areas with strong brightness jumps at their boundaries. The average gradient value of all abrupt change areas is set as the brightness abrupt change intensity index. For example, if the brightness in the area in front of the vehicle rapidly increases from 80 units to 200 units at a certain moment, with a time interval of 0.2 seconds, the average temporal gradient is (200–80) / 0.2 = 600 units / second, and this value will be used as an input to the sensitivity function.

[0083] To characterize the speed fluctuations of a vehicle caused by sudden environmental changes, a dynamic speed change threshold function is defined, and its calculation method is as follows:

[0084] The speed of the same target vehicle in consecutive image frames is recorded as a time series, and the speed change is monitored using a sliding time window. Within each time window, the maximum deceleration, the variance of speed change, and the variation value of speed direction are calculated. If the speed decrease value of the current frame is greater than 1.5 times the decrease of the historical average, or the speed fluctuation variance exceeds the preset warning threshold, it can be determined that there is abnormal deceleration behavior. The result of this indicator is output in the form of a numerical score. For example, the score range is set from 0 to 1, which represents a trend from no significant deceleration to strong deceleration or stopping.

[0085] The output process of the risk response instruction candidate set is as follows:

[0086] Combining the brightness abrupt change intensity, speed change score, and illumination incident angle, a joint input feature vector is constructed and input into the risk scoring mapping model. This model uses a set of predefined scoring weight matrices to assign different influence weights to different feature components; for example, the weight for brightness abrupt change intensity is 0.5, the weight for speed fluctuation is 0.3, and the weight for illumination incident angle is 0.2. The final driving visual adaptation risk score is calculated by summing the weighted multiplications of each component. For example, if the brightness score is 0.9, the speed score is 0.6, and the illumination angle score is 0.4, then the score is: 0.9×0.5+0.6×0.3+0.4×0.2=0.45+0.18+0.08=0.71. This risk score will be output as floating-point data to indicate the degree of influence of the visual environment corresponding to the current image frame on the driver's adaptability. The higher the score, the stronger the visual abrupt change and the more unstable the driving behavior, requiring the triggering of a linked control response of lights and signals.

[0087] The driving vision adaptation risk score calculated from consecutive image frames is input into a fixed-length sliding time window in the order of image time. The length of the time window can be set according to the actual traffic speed and system reaction time. It is generally recommended to set it to cover an image frame sequence of 1 to 3 seconds. For example, at a sampling frequency of 25 frames per second, the window length can be set to 75 frames.

[0088] Within each sliding window, the system calculates the gradient change value and local peak duration of the scoring curve. The gradient change value refers to the rate of change of the score between consecutive frames, i.e., the magnitude of score increase or decrease, used to identify the point of abrupt score changes. The peak duration represents the number of frames during which the score remains in the high-risk range (e.g., greater than 0.7), used to determine the time span of illumination interference. When a sudden increase in score is detected within a window (e.g., rising from 0.3 to 0.85 in a short period of time and lasting for more than 10 frames), the system considers there to be a rapid switching behavior of sudden changes in field of view brightness or uneven illumination in the current image frame and locates it as a potential area of ​​visual shock. Synchronous with visual scoring, the system also extracts the speed change data of each vehicle from the image frames to construct a vehicle speed change sequence. Each item in the sequence is the difference in the passing speed of the target vehicle between consecutive frames. The system calls the dynamic speed change threshold function defined in the aforementioned implementation to evaluate the speed change sequence and determine whether there is deceleration behavior exceeding the normal range. For example, if the speed of the same target vehicle drops from 40 km / h to 10 km / h in a short period of time within a sliding window, or if the speed fluctuation exceeds 2 m / s for three consecutive frames, the system determines that the vehicle is currently in a potentially abnormal deceleration state. The time period that meets this condition is marked as a potential traffic interference section, which will be used for comparison with the visual change area.

[0089] Subsequently, the system performs time synchronization registration on the brightness abrupt change area and the vehicle speed decrease section. Specifically, it compares the timestamps (i.e., image frame numbers) of the two events to confirm whether they occur within the same sliding window period and whether there is overlap or adjacent occurrence. If the two types of events occur simultaneously or approximately simultaneously on the timeline (e.g., within 5 frames), a joint event label, called a risk event trigger label, is constructed. The system further filters the trigger labels based on a risk score threshold. The default score threshold is set to 0.7, meaning that only when the score is greater than this risk score threshold is the event considered to have sufficient visual impact. Eligible trigger labels are output as a risk response instruction candidate set. Each candidate includes the event location, duration, recommended lighting enhancement level, and whether it is recommended to simultaneously issue signal instructions (such as speed limits or guidance signals). This risk response instruction candidate set will be used for decision matching and response execution in the subsequent illuminance adjustment module and joint control module.

[0090] In summary, by tensorizing visual features and constructing a visual adaptation risk scoring model, and combining it with a sliding time window to detect risk mutation trends, we can quickly identify sudden changes in visual field and traffic anomalies. This effectively supports the system in predicting potential risk areas in advance and outputting response candidate strategies, thereby enhancing the system's foresight and early warning capabilities.

[0091] The illuminance adjustment module constructs a lighting control state map based on the risk response command candidate set, extracts the matching relationship between traffic flow features and illumination gradients, and maps and generates the entrance section lighting illuminance adjustment parameter set, as follows:

[0092] Based on the risk response command candidate set output by the preceding module, the risk level (score), trigger location (spatial coordinates), and suggested response type (such as lighting enhancement, signal guidance) of each command are analyzed, and this information is used as a ternary mapping vector to map to the preset lighting response state space. The lighting response state space is a set of rules generated by summarizing historical control data and human experience, used to characterize the correspondence between traffic flow status and illumination changes and the required lighting level. In the lighting response state space, the ternary core features of the state map are defined as: vehicle distribution density, vehicle speed trend, and illumination gradient change rate. These three describe the congestion of the passage space, the vehicle behavior trend, and the brightness fluctuation of the tunnel section, respectively.

[0093] During the state map construction process, actual traffic flow feature parameters are extracted from the image frame regions corresponding to the trigger locations. Specifically, vehicle distribution density refers to the number of vehicles per unit spatial length. The system divides the tunnel entrance section into equal-length grids (e.g., 10 meters per segment) and counts the number of areas obscured by vehicles in each segment as a vehicle density index. Vehicle speed trend is represented by the average and changing trend of vehicle speeds within a sliding time window. The average passing speed and speed variance of vehicles in consecutive frames within the target lane are calculated to determine whether there is a deceleration clustering or congestion trend. Illumination gradient change rate represents the magnitude of change in regional brightness distribution with position in an image frame. It is calculated using the rate of change of brightness difference in the spatial dimension and combined with the average grayscale difference between the current frame and the previous frame to evaluate the degree and directionality of illumination abrupt changes.

[0094] Based on the three types of traffic flow characteristics mentioned above, the most suitable lighting response level is matched in the illuminance adjustment rule base. The illuminance adjustment rule base is a pre-set multi-level response strategy set, containing the lighting adjustment strategies required under different combinations of vehicle density, vehicle speed, and sudden changes in illumination. For example, when vehicle density is high, vehicle speed decreases significantly, and the illumination gradient changes drastically, the corresponding lighting response level is Enhancement Mode 3, i.e., rapid brightening, high illuminance, and long gradient; while when vehicle density is low, vehicle speed is stable, but the illumination change is slight, only Gradual Brightening Mode 1 may be needed, i.e., medium brightness and short transition. The system generates an entrance section lighting illuminance adjustment parameter set based on the matching results. This lighting illuminance adjustment parameter set includes:

[0095] Target illuminance value: The expected lighting intensity to be achieved in the current section, in lux (Lux). It is dynamically set according to tunnel design standards and risk levels. For example, the standard value is 400 Lux, which can be increased to 600 Lux during high-risk periods.

[0096] Illuminance gradient step size: This refers to the increment of each change in light illuminance adjustment, avoiding sudden changes that could cause new visual shocks. For example, it can be set to increase by 20 Lux per frame.

[0097] Gradient duration: controls the time required for a smooth transition from the current illuminance to the target illuminance, measured in seconds, and is typically dynamically adjusted within the range of 1 to 5 seconds;

[0098] Scope label: This indicates which lighting segments need to participate in this adjustment. For example, three lighting units within a 0 to 30 meter range at the tunnel entrance need to execute the commands of this parameter group simultaneously.

[0099] The illuminance adjustment parameter set is sent to the on-site lighting controller through the system control interface, driving the corresponding area lighting to perform lighting adjustments according to the set illuminance value, change rhythm and dimming range, forming an intelligent lighting control strategy that is synchronized with traffic flow dynamics and visual changes.

[0100] The illuminance adjustment module constructs a lighting control status map and generates segmented illuminance adjustment parameter groups based on the relationship between traffic flow and light gradient. This enables dynamic and graded control of the light intensity at the tunnel entrance section, effectively mitigating the visual impact at the boundary between light and dark, achieving a smoother brightness transition, and improving driver adaptability and traffic safety when entering the tunnel.

[0101] The control command generation module synchronously constructs signal linkage triggering factors. Combining the continuity of entry behavior, the probability of vehicle stopping, and the degree of visual blind spot obstruction, it determines whether to generate speed limit, guidance, or warning signal control commands, as detailed below:

[0102] Based on the vehicle target position coordinates extracted from consecutive image frames, cross-frame target tracking is performed on each vehicle to form a vehicle travel trajectory. This trajectory is a sequence of vehicle center point coordinates arranged chronologically, and can be expanded to include information such as vehicle speed, lane number, and bounding box. From this, an entry behavior continuity index is calculated. This index is used to evaluate the motion smoothness of the target vehicle during its entry into the tunnel from the tunnel approach section, and mainly includes the recognition of the following three behavior patterns:

[0103] Sudden deceleration behavior: If the vehicle speed drops more than a set threshold within 3 frames (e.g., from 50km / h to 20km / h) within 20 meters before entering the tunnel, it is judged as discontinuous deceleration.

[0104] Abnormal lane change behavior: If a vehicle deviates laterally from its trajectory within a short period of time, and the lane change behavior does not correspond to the target lane rules (such as crossing the line or deviating from the center line by more than 2 meters), it will be marked as an abnormal lane change.

[0105] Track interruption: If the target vehicle has frame gaps in continuous images (e.g., more than 3 frames are not detected) or the positioning jump exceeds the spatial step size allowed by normal speed, it is considered a track interruption.

[0106] The above three types of behavior are uniformly mapped to a continuity score for entering the cave. The score ranges from 0 to 1. The lower the value, the worse the continuity and the more likely it is to trigger the signal control requirement.

[0107] Based on vehicle speed sequences and deceleration behavior patterns from historical traffic samples, a short-term stopping probability model is constructed. The model's inputs include the mean speed, variance, and deceleration distribution of each vehicle within a sliding time window. Historical data can be collected through pre-deployment or generated by the simulation system. The model employs probability estimation methods based on K-nearest neighbor discrimination or Bayesian statistics to output a probability score for each vehicle's current behavior, indicating a possible stopping or near-zero speed state within the next second. For example, if the instantaneous speed change rate of a vehicle exceeds three meters per second squared (i.e., the speed increase or decrease within one second is greater than three meters per second), and two temporary near-zero speed behaviors occur within the sliding window, its stopping probability score may be assigned a value of 0.75. The system sets a stopping probability score threshold (e.g., 0.6); vehicles exceeding this threshold are marked as high-risk vehicles.

[0108] To identify the occlusion risk of blind spots in the current frame, the system performs occlusion analysis on the image region of each lane and calculates a blind spot occlusion severity score. The blind spot occlusion severity score is based on a combination of the following factors:

[0109] By detecting the area ratio of overlapping regions of multiple vehicle bounding boxes, it is determined whether occlusion overlap has occurred. Based on the vehicle arrangement density and camera angle, the area of ​​the region that the camera cannot effectively capture is dynamically estimated. Combining the angle between the incident angle of the light and the lane direction, the probability of the shadow band obstructing the camera's field of view due to oblique sunlight is assessed. The smaller the incident angle and the longer the projected shadow, the higher the occlusion score.

[0110] Based on the above three scoring factors, the final visual blind spot occlusion score is calculated according to the set weights (for example, boundary overlap 0.4, blind spot size 0.3, illumination occlusion 0.3, weighted to 0.4×0.5+0.3×0.3+0.3×0.2=0.37). The higher the score, the greater the impact of the current visual blind spot.

[0111] Finally, the system inputs the entry behavior continuity score, vehicle stopping probability score, and blind spot occlusion score into the signal control judgment logic to construct a signal linkage trigger factor. This signal linkage trigger factor is a ternary structure vector representing three types of risk intensity under the current traffic conditions, and it has response weights and multi-layer judgment rules. That is, when any two of the three scores exceed the set risk threshold (e.g., both are higher than 0.7), the system outputs a signal control trigger flag and selects the corresponding signal response method according to the recommended signal type mapping table. For example, if there is low continuity and high occlusion, a speed limit warning is issued; if the stopping probability is high and the continuity is medium, a guidance signal is issued; if all three are high, a warning signal is issued. At the same time, based on the vehicle position and the range of the occlusion area, the effective area label of the signal control command is set (e.g., the left lane of the entrance section, the 10 to 30 meter section of the right lane, etc.), and the structured signal control instructions are output completely.

[0112] The control command generation module constructs signal linkage triggering factors and generates precise control commands by comprehensively evaluating the continuity of vehicle behavior, the probability of stopping, and the degree of visual blind spot obstruction. It can trigger speed limit, guidance, or warning signals before the vehicle is about to enter a risky state, improving the pertinence of signal issuance and the timeliness of linkage, and reducing the risk of accidents.

[0113] The update input module sends the lighting illuminance adjustment parameter group and signal control commands to the field control equipment for execution. Simultaneously, it collects the execution results and image change sequences, extracting the lighting response deviation and behavioral trajectory offset as update inputs for the visual adaptation model, as detailed below:

[0114] The illuminance adjustment parameter set and signal control command generated by the illuminance adjustment module and the control command generation module are sent to the field control equipment. Specifically, the parameter set is transmitted to the luminaire control driver and signal control terminal through wired or wireless communication links, and they are controlled to perform illuminance adjustment, gradual step size adjustment, range start and stop, as well as the display of signal prompt graphics, color switching and guidance sign control, etc.

[0115] After the control command is executed, the image feedback acquisition program is started synchronously to record a continuous sequence of image frames after the command is executed. These image frames are then matched one by one with the image frames before the control execution in chronological order to generate an image change sequence after the lighting response. The matching method is precise alignment of timestamps, and an image similarity matching algorithm is used to correct inter-frame drift when necessary. Based on this image change sequence, the system extracts three key indicators for the tunnel entrance area to calculate the lighting response deviation:

[0116] Calculate the average brightness, which is the average grayscale value of all pixels in the specified area of ​​the tunnel entrance in the current image frame, reflecting the overall lighting level; calculate the brightness gradient distribution, which is to evaluate whether the lighting transition is smooth by calculating the first derivative of the image grayscale value in space, and prevent jagged light or local overbrightness; calculate the lane boundary sharpness index: based on the edge detection algorithm, evaluate the edge strength and continuity of the lane boundary lines. This index can indirectly determine whether the lighting is helpful for lane structure identification.

[0117] The above indicators are compared with control target parameters such as target illuminance value and expected gradient change rate to calculate the lighting response deviation of the current frame. For example, if the target brightness is set to 600 Lux, but the average brightness of the acquired image is only 500 Lux, the response deviation is -100 Lux; if the gradient change rate deviates from the expectation by more than 20%, it will also be recorded as a deviation. Simultaneously, the travel trajectory of each target vehicle is extracted from the image frames before and after control execution, and the trajectory point sequence of the vehicle in the image coordinate system is obtained through a multi-frame tracking algorithm. The time alignment offset (i.e., the difference in the number of frames of trajectory delay or advance), trajectory shape change rate (i.e., the root mean square error of changes in trajectory curvature, polygonal angle, etc.), and lane center deviation distance (i.e., the average vertical distance of the vehicle trajectory relative to the ideal lane centerline) are calculated within the same time period before and after control. These indicators together constitute the vehicle behavior trajectory deviation index, used to measure whether the lighting and signal adjustments have an actual guiding or interfering effect on driving behavior.

[0118] The lighting response bias and behavior trajectory deviation indices are integrated into a set of joint feedback vectors. This vector represents the environmental visual changes and behavior response bias after each control response in a standardized manner, and is used as incremental update data input to the model training interface for the visual adaptation model.

[0119] The update input module compares and analyzes the image changes and vehicle behavior trajectories before and after the execution of control commands, extracts lighting response deviation and behavior trajectory offset indices, constructs a joint feedback vector, provides quantitative support for model updates, realizes closed-loop feedback adjustment, and improves the learning efficiency and update accuracy of the visual adaptation model.

[0120] The joint control module, based on behavioral feedback results, employs a fine-tuning mechanism to update the brightness response mapping relationship and linkage strategy rules in the visual adaptation model. It adaptively optimizes the joint control of light and signals under different time periods and environments. Specific details include:

[0121] At the end of each control cycle, the lighting response deviation and the behavior trajectory offset index are combined into a joint feedback vector. The former describes the difference between the target illuminance value and the actual brightness output, while the latter describes the spatiotemporal offset between the vehicle's travel path and the ideal lane path. The system compares this joint feedback vector with the historical model response state, specifically including:

[0122] Match historical execution data under the same environmental label or control scenario to analyze the convergence or divergence trend between the current predicted lighting and vehicle behavior state and historical results;

[0123] If the system detects any of the following abnormal situations within multiple consecutive time windows (e.g., three consecutive sliding windows, each with 30 frames): First, insufficient lighting response, i.e., the actual average brightness is consistently lower than the target illuminance value by a certain range (e.g., the deviation exceeds 50 Lux); Second, vehicle deviation, i.e., the trajectory center of the target vehicle deviates from the ideal lane center by a distance exceeding the set tolerance (e.g., greater than 0.8 meters), then the local fine-tuning mechanism of the visual adaptation model is triggered.

[0124] After the fine-tuning mechanism is activated, the system first locates the set of model sub-parameters corresponding to the affected area. For example, if the current triggering area is the tunnel entrance section from 0 to 20 meters and the risk score is high, the system will extract the brightness response mapping parameters that match the label and risk level of the area, including the illuminance output weight, brightness increment step size factor, and gradient time adjustment factor associated with the area. Subsequently, the system uses gradient descent to fine-tune and optimize the set of sub-parameters: specifically, the system uses the error between the desired visual state and the actual response state as the loss target, and updates the illuminance adjustment weight parameters by iteratively reducing this error. For example, if the target illuminance is 600 Lux and the actual illuminance is 520 Lux, the system will search in the parameter weight matrix for the optimal path to increase the output to close to 600 Lux and appropriately increase the base illuminance parameter and adjustment rate of the area.

[0125] Furthermore, the "signal linkage strategy rules" can be structurally updated simultaneously, including the following updates:

[0126] Trigger threshold updates, such as the trigger threshold for vehicle stopping probability scores or blind spot occlusion scores, are appropriately increased or decreased based on feedback results; the effective range is updated, that is, the effective lane or distance range of the signal is adjusted based on the actual feedback of vehicle deviation range; the signal type selection conditions are updated, that is, when the current template repeatedly triggers a certain type of signal but fails to effectively change the vehicle behavior, the system will consider switching to a stronger warning signal or extending the signal holding time.

[0127] Based on the above fine-tuning results, a new set of control strategies was constructed and added to the control strategy history library. This fine-tuning process is a local, lightweight adjustment that does not change the overall model structure, but only optimizes the parameter space, ensuring online running efficiency and guaranteeing system stability.

[0128] Adaptive optimization of the joint control of lighting and signals under different time periods and environments, specifically including:

[0129] In the initial deployment phase, various typical control scenarios were extracted based on long-term operational data and expert strategies, and a control scenario library was constructed. Each scenario in this library is identified by a set of environmental tags, including but not limited to: time tags (e.g., morning rush hour, nighttime, dusk), weather tags (e.g., sunny with strong light, rainy or foggy weather), traffic tags (e.g., dense traffic, intermittent traffic), and illuminance status tags (e.g., the boundary between light and dark). Each control scenario is associated with a set of control templates, which pre-set recommended illuminance ranges, signal linkage strategies, response delay adjustment parameters, and parameter correction coefficients.

[0130] By collecting real-time data from sensors and cameras, the system detects the current time period (e.g., using the system clock and local daylight schedule to determine if it is dusk), lighting conditions (e.g., using the combination of average and gradient values ​​of image brightness to assess if it is strong backlight), traffic flow information (e.g., using vehicle density per unit area and speed variance to determine if it is high-density flow), and weather conditions (e.g., using external environmental data interfaces), thus forming a set of current environmental status labels.

[0131] During the scene matching process, the system compares the similarity of the current tag set with the tag features of each template in the control scene library. It can use weight judgment, tag voting or priority matching strategies for matching. For example, if the current tag meets the three tags of night, low speed and strong occlusion, the system selects the night low speed high occlusion template with the closest tag weight as the optimal control template.

[0132] The matched control template contains several key parameters, mainly including:

[0133] Illuminance level boundaries: the minimum and maximum illuminance value range for each risk level (e.g., 400–550 Lux for medium risk and 600–750 Lux for high risk).

[0134] Signal type switching threshold: A joint scoring threshold used to determine whether to switch from a speed limit signal to a warning signal (e.g., behavior continuity score < 0.3 and blind spot score > 0.7).

[0135] Linkage delay rule: used to determine the time offset between light changes and signal switching to prevent synchronization shocks (such as light gradation 2 seconds in advance, signal start 1 second late).

[0136] By combining the above parameters with the current fine-tuned visual adaptation model, a set of dynamic adjustment strategies is generated. These strategies are continuously refreshed, invoked, and executed within the control cycle. Specifically, the execution includes: adjusting the light intensity of the luminaires within the matching illuminance level boundary; the controller dynamically determining whether to issue speed limit, guidance, or warning commands based on the signal switching threshold; and coordinating the sequence of light and signal control according to the linkage delay rules.

[0137] The joint control module fine-tunes model parameters based on behavioral feedback results and combines them with the control scenario library to achieve adaptive scheduling of lighting and signal strategies. This effectively improves the system's generalization ability and control stability under different time periods, traffic densities, and lighting conditions, ensuring that the strategy response does not become rigid and the control does not fail during long-term system operation.

[0138] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0139] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0140] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0141] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0142] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. 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. A dynamic lighting and signal control system for tunnel entrances based on vision algorithms, characterized in that: It includes a tunnel entrance image acquisition module, a response command generation module, an illumination adjustment module, a control command generation module, an update input module, and a joint control module, with each module connected by signals; The tunnel entrance image acquisition module acquires a continuous sequence of image frames of the tunnel entrance area, and performs defogging, contrast enhancement and lane normalization processing to extract the visual features of the image frames. The response instruction generation module tensors and encodes the extracted visual features, calculates the driving visual adaptation risk score in the built-in visual adaptation model, and combines the sliding time window to detect short-term visual abrupt change areas and vehicle speed decrease sections, and outputs a risk response instruction candidate set. The illuminance adjustment module constructs a lighting control state map based on the risk response command candidate set, extracts the matching relationship between traffic flow characteristics and illumination gradient, and maps and generates the entrance section lighting illuminance adjustment parameter set. The control command generation module synchronously constructs signal linkage triggering factors, and combines the continuity of the entry behavior, the probability of the vehicle stopping and the degree of visual blind spot occlusion to determine whether to generate speed limit, guidance or warning signal control commands. The update input module sends the lighting illuminance adjustment parameter group and signal control commands to the field control equipment for execution. At the same time, it collects the execution results and image change sequences, and extracts the lighting response deviation and behavior trajectory offset as the update input of the visual adaptation model. The joint control module, based on behavioral feedback results, uses a fine-tuning mechanism to update the brightness response mapping relationship and linkage strategy rules in the visual adaptation model, and adaptively optimizes the joint control of light and signal in different time periods and environments.

2. The tunnel entrance dynamic lighting and signal control system based on vision algorithm according to claim 1, characterized in that: A continuous sequence of image frames of the tunnel entrance area was acquired, and dehazing, contrast enhancement, and lane normalization were performed. Visual features of the image frames were extracted, including: Image frames are dehazed using dark channel prior and histogram equalization algorithms; A contrast-limited adaptive histogram equalization method is used to enhance the contrast of image frames. Based on image perspective correction and lane template matching algorithms, lane normalization processing is performed on image frames to convert lane structures from different perspectives into a unified spatial reference system. After preprocessing, visual features such as vehicle distribution, traffic speed, lane occupancy, and light incidence angle are extracted from the image frames.

3. The tunnel entrance dynamic lighting and signal control system based on vision algorithm according to claim 2, characterized in that: The extracted visual features are tensor-encoded and used to calculate a driving visual adaptation risk score within the built-in visual adaptation model, including: A structured input matrix is ​​constructed based on visual features, and position embedding and temporal encoding are introduced to generate a visual feature tensor with spatiotemporal continuity. Visual feature tensors are input into the multi-layer convolutional attention network structure of the visual adaptation model for feature compression, and high-response feature vectors of the driving behavior attention region are extracted. Based on the preset sensitivity function for brightness abrupt change and the threshold for dynamic speed change, a visual adaptation risk scoring function is constructed through a visual adaptation model, and the driving visual adaptation risk score corresponding to the current image frame is calculated through feature vectors and a risk mapping model. The driving visual adaptation risk score reflects the probability of instantaneous visual impairment caused by the combination of changes in brightness, traffic density and entry angle under the current traffic conditions.

4. The tunnel entrance dynamic lighting and signal control system based on vision algorithm according to claim 3, characterized in that: Based on a preset sensitivity function for abrupt changes in brightness and a threshold for dynamic speed changes, a visual adaptation risk scoring function is constructed. This function calculates the driving visual adaptation risk score corresponding to the current image frame using feature vectors and a risk mapping model, including: Based on the brightness gradient change of the corresponding area of ​​the vehicle in the image frame, a brightness abrupt change sensitivity function is constructed. The brightness abrupt change sensitivity function is used to measure the spatial gradient intensity and temporal gradient change rate of brightness in the forward region of the vehicle, and the boundary response value of the brightness variation block is extracted by local region differential convolution. By combining the vehicle speed change sequence, a dynamic speed change threshold function is defined and used to determine the deceleration trend, speed fluctuation amplitude and instantaneous deceleration acceleration of the vehicle between adjacent image frames, and to identify the coupling signal between changes in driving behavior and visual scene switching. A visual adaptation risk scoring function is constructed, which combines the intensity of brightness abrupt change, speed change index and incident light angle into a joint input feature vector. The vector product mapping is performed by the risk mapping model and the preset scoring weight matrix to calculate the driving visual adaptation risk score of the current image frame.

5. The tunnel entrance dynamic lighting and signal control system based on vision algorithm according to claim 4, characterized in that: By combining the detection of short-term visual abrupt changes and vehicle speed reduction sections using a sliding time window, a candidate set of risk response instructions is output, including: The driving visual adaptation risk scores in consecutive image frames are input into a fixed-length sliding time window in chronological order. The gradient change value and peak duration of driving visual adaptation within the window are calculated to identify image areas with sudden brightness changes or uneven lighting. The vehicle speed change sequence corresponding to each image frame within the window is extracted synchronously. Combined with the dynamic speed change threshold function, it is determined whether there is deceleration behavior or speed fluctuation characteristics exceeding the set threshold, and marked as potential traffic interference sections. The brightness abrupt change area is time-synchronized with the vehicle speed decrease section to construct risk event trigger tags, and the trigger tags are filtered according to the risk score threshold to output a candidate set of risk response instructions.

6. The tunnel entrance dynamic lighting and signal control system based on vision algorithm according to claim 5, characterized in that: The illuminance adjustment module includes: The risk level, trigger location and suggested response type in the risk response instruction candidate set are mapped to the preset lighting response state space to construct a ternary state map containing vehicle distribution density, vehicle speed trend and illumination gradient change rate. In the ternary state map, traffic flow features corresponding to the trigger position are extracted, including the number of vehicles per unit length, average vehicle speed and speed variance, as well as the illumination intensity gradient and brightness change direction of the corresponding region of the image frame. Based on the matching relationship between traffic flow characteristics and illumination gradient, the illuminance adjustment rule library is called to match the corresponding lighting response level and generate the entrance section light illuminance adjustment parameter group. The entrance section lighting illuminance adjustment parameter group is used to drive the on-site lighting fixtures to perform segmented illuminance adjustment, including target illuminance value, illuminance gradient step size, gradient duration, and effective range label.

7. The tunnel entrance dynamic lighting and signal control system based on vision algorithm according to claim 6, characterized in that: The control command generation module includes: Extract the travel trajectory of the target vehicle from consecutive image frames, calculate the continuity index of the vehicle's entry behavior into the tunnel, and identify whether there is sudden deceleration, abnormal lane change or trajectory interruption before entering the tunnel. Based on the vehicle speed sequence within the sliding time window and combined with historical deceleration behavior data, a short-term stopping probability model is constructed to generate a corresponding vehicle stopping probability score for each vehicle in the current traffic flow. Occlusion analysis is performed on the visual field area of ​​each lane in the image frame. The visual blind spot occlusion degree score corresponding to the current frame is calculated by the spatial projection relationship between the overlapping area of ​​the vehicle boundary, the size of the blind spot and the incident light angle. The system integrates the continuity index of entry behavior, the vehicle stopping probability score, and the visual blind spot occlusion score to construct a signal linkage trigger factor for decision-making. It outputs whether to generate speed limit, guidance, or warning signal control commands and selects the corresponding signal type and range of action based on the decision result.

8. The tunnel entrance dynamic lighting and signal control system based on vision algorithm according to claim 7, characterized in that: Simultaneously, execution results and image change sequences are collected, and illumination response bias and behavioral trajectory offset are extracted as update inputs for the visual adaptation model, including: The execution results of the lighting illuminance adjustment parameter group and signal control command are collected and compared with the corresponding image frames before execution to generate an image change sequence after the lighting response. Based on the average brightness, gradient distribution, and lane boundary clarity indices of the tunnel entrance area in the image change sequence, the lighting response deviation between the current light output and the preset target illuminance is calculated. Extract the travel trajectory of the target vehicle before and after the execution control, and calculate the vehicle behavior trajectory deviation index based on the time alignment offset between trajectory point sets, trajectory shape change rate and lane center deviation distance; The lighting response deviation and the behavior trajectory offset index are combined into a joint feedback vector, which is then input into the visual adaptation model as incremental update data.

9. The tunnel entrance dynamic lighting and signal control system based on vision algorithm according to claim 8, characterized in that: Based on behavioral feedback results, a fine-tuning mechanism is used to update the brightness response mapping relationship and linkage strategy rules in the visual adaptation model, including: The joint feedback vector composed of lighting response deviation and behavior trajectory deviation index is compared with the historical response state of the model to determine the dynamic convergence trend between the current control effect and the prediction deviation. When insufficient lighting response or vehicle deviation exceeds the tolerance threshold occurs within multiple consecutive time windows, the local fine-tuning mechanism of the visual adaptation model is triggered, and the mapping parameters of the affected area are updated. The gradient descent method is used to refine the illuminance adjustment weights in the luminance response mapping relationship, thereby reducing the error between the actual lighting output and the desired visual state. At the same time, the trigger threshold, scope of action, and signal type selection conditions in the signal linkage strategy rules are updated to construct a new round of control strategy groups that adapt to the current traffic behavior pattern.

10. The tunnel entrance dynamic lighting and signal control system based on vision algorithm according to claim 9, characterized in that: Adaptive optimization of joint control of lighting and signals under different time periods and environments, including: Based on the currently detected time period information and environmental status, the optimal control template is matched in the control scenario library, and a dynamic adjustment strategy is generated by combining the latest fine-tuned model parameters. By invoking the illuminance level boundary, signal type switching threshold, and linkage delay rules in the matching control template, adaptive optimization of the joint control of light and signal in different scenarios can be achieved.

Images