An integrated method for traffic visibility safety distance detection and safety warning in low visibility conditions

Abstract

This invention proposes an integrated method for detecting and warning traffic visibility safety distances in low-visibility conditions, relating to the field of road traffic safety monitoring and early warning technology. The method includes deploying cameras along the road, an edge computing unit, and multiple traffic safety warning lights arranged at preset intervals. The edge computing unit controls the warning lights to display text of specific colors. The cameras capture images containing the warning light groups. The edge computing unit uses artificial intelligence recognition and optical character recognition to determine the last warning light whose text is correctly identified consecutively, starting from the nearest warning light. It calculates the current identification safety distance based on the physical location of the warning light, obtains the traffic visibility safety distance through a preset mapping relationship, and controls the warning light groups to display matching warning information according to the safety distance level. This invention can simulate the driver's real visual perception, achieving integrated local real-time detection and graded early warning of safety distances in low-visibility conditions, effectively compensating for the deficiencies of meteorological visibility indicators.

Description

Technical Field

[0001] This invention relates to the field of road traffic safety monitoring and early warning technology, and in particular to an integrated method for detecting traffic visibility safety distance and providing safety warnings in low visibility conditions. Background Technology

[0002] Meteorological conditions are a key factor affecting road traffic safety. In recent years, local road traffic authorities have invested heavily in building traffic safety early warning systems for severe weather conditions, with visibility monitoring being a core component. Currently, visibility detection along roads mainly uses meteorological visibility meters (transmission or scattering type), which calculate visibility values ​​by emitting a specific light source and measuring the attenuation coefficient of the light intensity. However, this method has the following shortcomings: (1) The meteorological visibility index is closely related to, but not the same as, the "driving safety visibility distance" that is truly of concern in traffic safety management. The meteorological visibility meter eliminates the influence of ambient light in principle, but in actual driving, ambient light (such as day and night) has a significant impact on the driver's visual recognition ability.

[0003] (2) The construction cost of professional visibility detectors is high, making it impossible to deploy them densely along roads, resulting in limited monitoring coverage. In particular, for patchy fog with localized and sudden characteristics, the existing sparse monitoring points are unable to provide effective early warning.

[0004] (3) Visibility monitoring and safety warning are often handled by different departments (meteorological department and road management department), and data integration is lagging behind, making it impossible to achieve closed-loop control of local real-time monitoring and immediate warning. Therefore, this invention proposes an integrated method for traffic visibility safety distance detection and safety warning under low visibility conditions to solve the problems existing in the prior art. Summary of the Invention

[0005] To address the aforementioned problems, the present invention aims to propose an integrated method for traffic visual safety distance detection and safety warning under low visibility conditions. The present invention uses a camera, an edge computing unit, and a warning light group to simulate the driver's visual recognition ability, thereby achieving integrated closed-loop control of safety distance detection and graded early warning, thus solving the problems existing in the prior art.

[0006] To achieve the objectives of this invention, the invention is implemented through the following technical solution: an integrated method for detecting and warning of traffic visibility safety distance in low visibility conditions, comprising the following steps: Step 1, Equipment Deployment: Along the road driving direction, set up a camera and an edge computing unit at the monitoring starting point. Then, starting at a first preset distance from the camera, continuously deploy M controllable traffic safety warning lights at a second preset interval to form a warning light group, where M is an integer greater than or equal to 2. Then, make the edge computing unit communicate with the control units of the camera and the warning light group respectively. Step 2, Active Control Display: The edge computing unit generates a first display control command and sends it to the control unit of the warning light group to control each traffic safety warning light to display a specific text in a preset specific color; Step 3, Image Acquisition: The edge computing unit then sends a capture command to the camera to control the camera to acquire the current road image including the traffic safety warning lights in the warning light group; Step 4: Image Recognition and Distance Calculation: The edge computing unit performs artificial intelligence recognition on the acquired current road image to identify the position of each traffic safety warning light and the text displayed therein. Then, the identified text is compared with the text set in the first display control command to determine the last traffic safety warning light whose text is correctly identified, starting from the traffic safety warning light closest to the camera. Then, based on the known physical installation position of the last traffic safety warning light, the distance between it and the camera is calculated as the current identification safe distance. Step 5, Safe Visibility Distance Mapping: Based on the current safe recognition distance, the current safe traffic visibility distance is determined through a preset mapping relationship. The mapping relationship represents the correspondence between the camera's recognition distance and the driver's actual safe visibility distance. Step Six: Tiered Warning: The edge computing unit generates a corresponding second display control command based on the warning level determined by the current traffic visibility safety distance, and sends it to the control unit of the warning light group to control the warning light group to display warning information that matches the warning level.

[0007] A further improvement is made in the following: in step one, the first preset distance is ≤25 meters, the second preset spacing is 20 to 30 meters, and the value of M makes the total monitoring distance ≥200 meters.

[0008] A further improvement is made in the following way: In step four, the specific method for determining the last traffic safety warning light whose text is correctly identified is as follows: Optical character recognition is performed on the text displayed by each traffic safety warning light in the image from bottom to top, where bottom to top corresponds to from near to far. Then, the sequence number of the first traffic safety warning light that fails to be identified or whose identification result is inconsistent with the sent text is recorded, and the previous traffic safety warning light is determined as the last traffic safety warning light that is correctly identified. If the identification fails from the traffic safety warning light closest to the camera, it is determined that the current traffic visibility safety distance is less than the first preset distance.

[0009] A further improvement is that when no traffic safety warning light can be identified in three consecutive captured images, the edge computing unit determines that the current traffic visibility safety distance is extremely low and directly triggers the highest level warning.

[0010] A further improvement is that step four includes performing statistical outlier removal processing on the obtained distance original data sequence, removing outlier data points that exceed the confidence interval, and repeating the process until no new outliers are added, using the statistical feature value of the remaining valid data as the current identification safe distance.

[0011] A further improvement is that in step five, the preset mapping relationship is obtained through pre-calibration, wherein the calibration conditions include sunny daytime conditions or standard laboratory lighting conditions.

[0012] A further improvement is made in the following steps: In steps one and four, each traffic safety warning light in the warning light group is assigned a unique identifier by the edge computing unit. The identifiers are numbered 0, 1, 2, ..., M-1 from near to far. The traffic safety warning light that is identified is then automatically associated with the corresponding identifier based on its position in the image, thereby obtaining its corresponding physical distance.

[0013] Further improvements include the following steps: Step 7: Data storage: The edge computing unit saves the road images captured under low visibility conditions, the corresponding current traffic visibility safety distance, and timestamp information to local storage or uploads them to a remote data center.

[0014] A further improvement is that the edge computing unit can automatically adjust the confidence threshold for optical character recognition in step four based on ambient light intensity or the built-in clock, or actively adjust the display brightness of the traffic safety warning light in step two.

[0015] The beneficial effects of this invention are as follows: (1) The present invention uses an integrated device installed along the road to simulate the driver’s ability to identify road reference objects in a real driving environment and directly measure the traffic safety visibility distance. Compared with the traditional meteorological visibility index, it is more in line with the actual driving needs.

[0016] (2) The equipment of the present invention is inexpensive and can be deployed at high density along the road. It can effectively capture local low visibility events such as fog and achieve localized accurate monitoring.

[0017] (3) The warning light group in this invention serves as both an identification reference and a carrier for issuing early warning information, enabling the edge computing unit to immediately control the same light group to change the displayed content when it detects a low safe distance, thereby achieving real-time closed-loop early warning without waiting for instructions from the central platform.

[0018] (4) This invention actively controls the content displayed by the warning light, and combines artificial intelligence recognition and optical character recognition to distinguish the recognition capabilities at different distances, automatically calculate the safe distance value, and use statistical methods to remove abnormal data, thereby improving the reliability of detection. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the device deployment of the present invention.

[0020] Figure 2 This is a schematic diagram illustrating the principle behind how smog reduces visibility.

[0021] Figure 3 This is a schematic diagram illustrating the image effects corresponding to different depths of field.

[0022] Figure 4 This is a schematic diagram of the image recognition and distance calculation process of the present invention.

[0023] Figure 5 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0024] To enhance understanding of the present invention, the present invention will be further described in detail below with reference to embodiments. These embodiments are only used to explain the present invention and do not constitute a limitation on the scope of protection of the present invention.

[0025] Visibility refers to the maximum horizontal distance at which a person with normal vision can see and identify the outline of a target (black, of moderate size) against a background of sky under current weather conditions. At night, it is the maximum horizontal distance at which a light source of a certain intensity can be seen and identified. In foggy conditions, atmospheric conditions between the target and the observer directly affect visibility. Figure 2 As shown. Mathematically, it can be represented as Where t(x) represents the attenuation phase, A represents the atmospheric light, J(x) represents the target brightness, and d(x) represents the target depth of field. The atmospheric scattering coefficient is represented by I(x;t). The intensity of the image obtained by the camera is represented by I(x;t).

[0026] Because the fog reduces the brightness of the target, the target object becomes less clear, such as... Figure 3 As shown, the recognition of specific text of a specific color depends on the observed image intensity I(x;t), which is related to the brightness J(x,t) of the target itself, as well as the atmospheric light A (i.e., the weather conditions at the time) and the atmospheric scattering coefficient (fog or haze).

[0027] Use respectively , , , These represent the complexity, color, size, and brightness of the text and icons. Since these four factors directly affect the recognition rate, and consequently the viewing distance, we define the distance from the i-th warning light to the camera, given these four factors, as fit (regress) to the following relationship: Road safety visibility distance is the distance at which a natural person can accurately identify a target under certain conditions. Specifically, road safety distance refers to the distance at which an object of a certain size can be identified from the background under certain target brightness and natural lighting conditions. Therefore, the visibility distance recognized by a camera needs to be converted into the distance at which a natural person can accurately identify a target in the same scene. in, The mapping relationship between the camera's recognized line of sight and the road safety line of sight.

[0028] In practical applications, different warning and alert strategies are set according to different road safety visibility distances. These strategies are then used... This indicates that different levels of control strategies are used. ,Right now: The warning and alert strategy that should be adopted based on the road safety visibility distance identified through video images can be expressed as follows: Example 1 according to Figures 1-5 As shown, this embodiment provides an integrated method for detecting and warning of traffic visibility safety distance in low visibility conditions, including the following steps: Step 1: Equipment Deployment: On a straight section of a highway (two-way four lanes, design speed 120km / h), install the equipment along the right-hand emergency stopping lane guardrail in the direction of traffic. At the monitoring starting point, install a high-definition network camera at a height of 2.5 meters, with the lens facing the outer edge of the downstream lane. Starting at a first preset distance X=20 meters from the camera, install a traffic safety warning light every second preset interval Y=25 meters along the guardrail, for a total of M=12 lights. The total monitoring distance is 20 meters + 11 × 25 meters = 295 meters, greater than 200 meters. Each warning light is an LED dot-matrix controllable luminous sign, capable of displaying text or symbols in different colors such as red, yellow, and green. All warning lights are connected to a warning light group controller via an RS485 bus. The edge computing unit (using an industrial-grade embedded GPU board) is connected to the camera via a network cable and simultaneously connected to the warning light group controller via a serial port. The edge computing unit assigns a unique identifier (ID) to each warning light, with the numbers 0, 1, 2, ..., 11 from closest to furthest.

[0029] Step 2, Active Display Control: Calibration is performed during a clear day. The edge computing unit generates the first display control command, instructing the warning light controller to display all warning lights in red with the word "Safe". Each warning light displays the same red "Safe" text. The edge computing unit confirms successful command transmission via serial port.

[0030] Step 3, Image Acquisition: The edge computing unit sends a capture command to the camera every 100 milliseconds. After receiving the command, the camera immediately captures a JPEG image containing all 12 warning lights and transmits it back to the edge computing unit via the network.

[0031] Step 4: Image Recognition and Distance Calculation: The edge computing unit uses the built-in AI object detection model (trained based on YOLOv5, capable of recognizing the location of warning lights) to process the image. First, it detects the bounding boxes of all warning lights in the image and sorts them according to their center ordinates from smallest to largest (smaller ordinates correspond to lights at the top of the image, i.e., distant lights; larger ordinates correspond to lights at the bottom of the image, i.e., closer lights). Since the camera is installed higher than the warning lights, the order from bottom to top corresponds to the ID order from nearest to farthest. Next, each detected warning light area is preprocessed (grayscale conversion, binarization), and then the optical character recognition (OCR) engine (Tesseract OCR) is called to recognize the text content. The recognized text is compared with the display content ("security") already issued by the edge computing unit. Starting from ID0 (most recent), the recognition is checked sequentially to determine if it is correct. For example, in a certain snapshot: ID0 to ID7 are identified as "Safe", ID8 is identified as "SafeQ" (incomplete), and ID9 cannot be identified. Therefore, the last warning light ID that is correctly identified consecutively is 7. Given that the physical distance corresponding to ID0 is the first preset distance of 20 meters, and the warning light spacing is 25 meters, then the physical distance of ID7 = 20 + 7 × 25 = 195 meters. This value is used as the current safe identification distance di for this sample.

[0032] If identification fails at ID0 (e.g., due to heavy fog making the text on the nearest light unrecognizable), the edge computing unit directly determines that the current traffic visibility safety distance is less than 20 meters and switches to the highest level of warning.

[0033] Furthermore, due to the possibility of temporary obstructions on the highway (such as birds or fallen leaves), to eliminate occasional interference, the edge computing unit continuously collected 10 samples to obtain a set of original distance sequences: 195, 192, 196, 43, 194, 195, 197, 193, 195, 194. Among them, 43 meters was obviously abnormal (possibly due to fallen leaves obstructing the view of the wrong light). The edge computing unit used the 3σ criterion to remove anomalies: the mean and standard deviation were calculated, the upper boundary was set as mean + 3 × standard deviation, and the lower boundary was set as mean - 3 × standard deviation. Points exceeding the boundary (43) were removed, and then the mean of the remaining 9 data points was recalculated to obtain 195.7 meters. After no new anomalies were found, 195.7 meters was taken as the final effective current identification safety distance d. i .

[0034] Step 5: Safe Visual Distance Mapping: The edge computing unit pre-stores a mapping table f, which is obtained manually under clear daytime conditions. The calibration process is as follows: A driver is assigned to drive on the same road segment. When the driver can clearly identify the word "Safe" on a certain warning light, the actual distance to that light is recorded; simultaneously, the distance at which the camera correctly identifies the text on the light is also recorded. After multiple experiments, a linear relationship is fitted: d r = 0.95 ×d i That is, the camera's recognition distance is slightly greater than the human eye's recognition distance (due to the camera lens's resolution advantage). This time, d i =195.7 meters, then the current safe visibility distance for traffic is d. r = 0.95 × 195.7 ≈ 186 meters.

[0035] Step Six: Tiered Early Warning: The edge computing unit has three early warning levels: Level 3 (highest): d r If the distance is less than 50 meters, the warning lights will flash red and indicate "Stop". Level 2: 50 meters ≤ d r <100 meters, the warning lights will display a solid yellow "slow fog" warning; Level 1: 100 meters ≤ d r If the distance is less than 200 meters, the warning lights will display a solid yellow "Caution" warning. Level 0: d r ≥ 200 meters, no warning will be triggered.

[0036] This time d r =186 meters, belonging to Level 1. The edge computing unit generates a second display control command, requiring all lights in the warning light group to display the solid yellow text "Caution". This command is sent to the warning light group controller via serial port. After execution, all warning lights display "Caution", providing a safety reminder to passing vehicles.

[0037] Step 7: Data storage: The edge computing unit saves the original image captured this time, the recognition result, the current traffic visibility safety distance value of 186 meters, and the timestamp to the local TF card, and uploads it to the remote traffic management data center through the 4G module for subsequent analysis.

[0038] Example 2 The difference between this embodiment and Embodiment 1 is that this embodiment is applied to a nighttime scenario. In a nighttime scenario, the ambient light intensity decreases, and the driver's sensitivity to the brightness of the warning lights changes. In this embodiment, the edge computing unit has a built-in light sensor (or reads the camera's photosensitivity parameters). When the ambient light intensity is detected to be below 10 lux, the edge computing unit automatically performs two adjustments: (1) In the active control display step, a command is sent to increase the display brightness of the warning light from the default 80% to 100%; (2) In the optical character recognition step, the confidence threshold of OCR is reduced from 0.85 to 0.75 (because the contrast of the image decreases at night, but the human eye may still be able to see the brighter text).

[0039] The remaining steps are the same as in Example 1. Actual measurements show that in nighttime conditions, the camera's recognition distance d... i The actual visible distance d of the human eye r When the mapping relationship changes, the edge computing unit automatically switches to the pre-calibrated nighttime mapping table, thereby ensuring the accuracy of the early warning strategy.

[0040] Furthermore, during the equipment installation phase, the construction workers installed the first warning light 20 meters away from the camera, the second 45 meters away (25-meter intervals), and so on. After installation, the physical location (precise distance in meters from the camera) of each warning light was entered into the configuration table of the edge computing unit via a handheld terminal and bound to an ID (0-11). The edge computing unit automatically matches the IDs by recognizing the order of the warning lights in the image using AI, without manual intervention. When a warning light is replaced due to malfunction, only the physical location of the new light needs to be re-entered.

[0041] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the present invention without departing from its framework and scope of application, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. An integrated method for detecting and warning of traffic visibility safety distance in low visibility conditions, characterized in that: Includes the following steps: Step 1, Equipment Deployment: Along the road driving direction, set up a camera and an edge computing unit at the monitoring starting point. Then, starting at a first preset distance from the camera, continuously deploy M controllable traffic safety warning lights at a second preset interval to form a warning light group, where M is an integer greater than or equal to 2. Then, make the edge computing unit communicate with the control units of the camera and the warning light group respectively. Step 2, Active Control Display: The edge computing unit generates a first display control command and sends it to the control unit of the warning light group to control each traffic safety warning light to display a specific text in a preset specific color; Step 3, Image Acquisition: The edge computing unit then sends a capture command to the camera to control the camera to acquire the current road image including the traffic safety warning lights in the warning light group; Step 4: Image Recognition and Distance Calculation: The edge computing unit performs artificial intelligence recognition on the acquired current road image to identify the position of each traffic safety warning light and the text displayed therein. Then, the identified text is compared with the text set in the first display control command to determine the last traffic safety warning light whose text is correctly identified, starting from the traffic safety warning light closest to the camera. Then, based on the known physical installation position of the last traffic safety warning light, the distance between it and the camera is calculated as the current identification safe distance. Step 5, Safe Visibility Distance Mapping: Based on the current safe recognition distance, the current safe traffic visibility distance is determined through a preset mapping relationship. The mapping relationship represents the correspondence between the camera's recognition distance and the driver's actual safe visibility distance. Step Six: Tiered Warning: The edge computing unit generates a corresponding second display control command based on the warning level determined by the current traffic visibility safety distance, and sends it to the control unit of the warning light group to control the warning light group to display warning information that matches the warning level.

2. The integrated method for detecting and warning traffic visibility safety distance in low visibility conditions according to claim 1, characterized in that: In step one, the first preset distance is ≤25 meters, the second preset spacing is 20 to 30 meters, and the value of M makes the total monitoring distance ≥200 meters.

3. The integrated method for detecting and warning traffic visibility safety distance in low visibility conditions according to claim 1, characterized in that: In step four, the specific method for determining the last traffic safety warning light whose text is correctly identified is as follows: Optical character recognition is performed on the text displayed by each traffic safety warning light in the image from bottom to top, where bottom to top corresponds to from near to far. Then, the sequence number of the first traffic safety warning light that fails to be identified or whose identification result is inconsistent with the sent text is recorded, and the previous traffic safety warning light is determined as the last traffic safety warning light that is correctly identified. If the identification fails from the traffic safety warning light closest to the camera, it is determined that the current traffic visibility safety distance is less than the first preset distance.

4. The integrated method for detecting and warning traffic visibility safety distance in low visibility conditions according to claim 3, characterized in that: When no traffic safety warning light can be identified in three consecutive captured images, the edge computing unit determines that the current traffic visibility safety distance is extremely low and directly triggers the highest level warning.

5. The integrated method for detecting and warning traffic visibility safety distance in low visibility conditions according to claim 1, characterized in that: Step four also includes statistical outlier removal processing on the obtained distance raw data sequence, removing outlier data points that exceed the confidence interval, and repeating the process until no new outliers are added, using the statistical feature value of the remaining valid data as the current identification safe distance.

6. The integrated method for detecting and warning traffic visibility safety distance in low visibility conditions according to claim 1, characterized in that: In step five, the preset mapping relationship is obtained through pre-calibration, where the calibration conditions include sunny daytime conditions or standard laboratory lighting conditions.

7. The integrated method for detecting and warning traffic visibility safety distance in low visibility conditions according to claim 1, characterized in that: In steps one and four, each traffic safety warning light in the warning light group is assigned a unique identifier by the edge computing unit. The identifiers are numbered 0, 1, 2, ..., M-1 from near to far. The traffic safety warning light that is identified is then automatically associated with the corresponding identifier based on its position in the image, thereby obtaining its corresponding physical distance.

8. The integrated method for detecting and warning traffic visibility safety distance in low visibility conditions according to claim 1, characterized in that: It also includes the following steps: Step 7: Data storage: The edge computing unit saves the road images captured under low visibility conditions, the corresponding current traffic visibility safety distance, and timestamp information to local storage or uploads them to a remote data center.

9. The integrated method for detecting traffic visibility safety distance and providing safety warnings in low visibility conditions according to claim 1, characterized in that: The edge computing unit also automatically adjusts the confidence threshold for optical character recognition in step four based on ambient light intensity or a built-in clock, or actively adjusts the display brightness of traffic safety warning lights in step two.

Images