A night road bridge collapse early warning method and system based on visual perception

Abstract

The present application belongs to the field of automobile active safety and intelligent transportation technology, and relates to a night road and bridge collapse early warning method and system based on visual perception. The method comprises the following steps: step one, establishing a normal disappearance rule benchmark model of a preceding vehicle tail light and a normal optical characteristic model of a roadside vertical sign; step two, performing preliminary calculation of a collapse confidence, giving a first collapse confidence A1, and simultaneously analyzing the roadside vertical sign, if the optical characteristics of the vertical sign detected do not conform to the normal optical characteristic model rule, then giving a second collapse confidence A2; step three, comparing the bridge collapse comprehensive confidence A with a preset threshold value to determine whether to issue a warning. The present application can quickly detect bridge collapse hazards at night, provide instant warning for vehicles in motion, and avoid the delay and non-real-time problems of traditional fixed sensors or manual inspection.

Description

Technical Field

[0001] This invention belongs to the field of automotive active safety and intelligent transportation technology, and relates to a method and system for early warning of road and bridge collapse at night based on visual perception. Background Technology

[0002] With the increasing development of road traffic networks, bridges, as key nodes, are of paramount importance in terms of structural safety. Especially at night, when visibility is poor, drivers may not be able to detect potential hazards ahead in time if a bridge collapses or suffers other major structural damage, which can easily lead to serious traffic accidents. Currently, monitoring the health of bridges relies mainly on fixed sensors (such as strain gauges and displacement meters) or periodic manual inspections. These methods are costly, lack real-time performance, and cannot provide immediate warnings to vehicles in motion.

[0003] Existing advanced driver assistance systems (ADAS) in vehicles typically include forward collision warning (FCW) and lane departure warning (LDW) functions. However, their sensors (such as cameras and radar) primarily focus on dynamic obstacles and lane markings, lacking effective detection methods for static, large-scale scenarios such as the failure of an entire bridge structure. Especially at night, when environmental information is scarce, traditional visual algorithms face significant challenges. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and propose a method and system for early warning of road and bridge collapse at night based on visual perception.

[0005] On the one hand, the present invention provides a method for early warning of nighttime road and bridge collapse based on visual perception, comprising the following steps: Step 1: As Figure 1 As shown, a baseline model of the normal disappearance law of the taillights of the preceding vehicle and a normal optical characteristic model of the roadside elevation markings are established. Step 2: Real-time detection and tracking of vehicles ahead. If the disappearance behavior of the taillights of the vehicle ahead does not conform to the normal disappearance pattern of the taillights of the vehicle ahead, it is determined that an abnormal event has occurred. Then, it is determined whether the abnormal event is caused by an accident of the vehicle ahead. If it is determined to be an accident of the vehicle ahead, the system will not trigger a bridge collapse warning. Otherwise, a preliminary calculation of the collapse confidence is performed, and a first collapse confidence of A1 is assigned. At the same time, the roadside facade markings are analyzed. If the optical characteristics of the facade markings are detected to be inconsistent with the normal optical characteristic model, a second collapse confidence of A2 is assigned. Step 3: Perform a weighted fusion of the first collapse confidence level A1 and the second collapse confidence level A2 to obtain the final bridge collapse comprehensive confidence level A. Compare the bridge collapse comprehensive confidence level A with the preset threshold to determine whether to issue an early warning.

[0006] Furthermore, in step one, the baseline model for the normal disappearance pattern of the taillights of the preceding vehicle includes the location where the taillight brightness disappears, and the formula is: (1) Where P is the vanishing position, Vr is the relative velocity, t is the time, r is the radius of curvature, θ is the viewing angle, k is the road curvature, and d1 is the distance.

[0007] Furthermore, the reflectance model in step one is formulated as follows: (2) Where I(d,m) is the reflected brightness at a distance of d2 under mode m (low beam=0, high beam=1), I_0 is the incident light intensity, cos(ϕ) is the cosine of the incident angle, and η_m is the mode efficiency factor, which is used to characterize the relative efficiency of the light source's contribution to the reflection of the target under different lighting modes (m=0 represents low beam, m=1 represents high beam); this coefficient is obtained through actual road testing and calibration.

[0008] Furthermore, in step two, the first collapse confidence level A1 and the second collapse confidence level A2 are expressed as follows: (3) A 2 =1 or A2=0 (4) Where Pmeasured is the detected position where the taillights of the preceding vehicle disappear, and Ppredicted is the predicted position where the taillights of the preceding vehicle disappear.

[0009] Furthermore, in step two, determining whether the abnormal event was caused by the preceding vehicle specifically means: if the analysis results show obvious lateral displacement, sharp turning, or trajectory characteristics consistent with a collision, then the system determines that the abnormality was caused by the preceding vehicle's own accident.

[0010] Furthermore, in step three, the formula for the comprehensive confidence level A of bridge collapse is as follows: A = k1*A1 + k2*A2 (5) k1+k2=1 (6) Where k1 and k2 are the average weighting coefficients.

[0011] Furthermore, in step three, if A ≥ the threshold, it is determined that the probability of the bridge ahead collapsing is extremely high, the highest level of collision warning is issued, and automatic emergency braking is initiated; if A ≤ the threshold, it is considered that the risk has not reached the warning level.

[0012] On the other hand, a system based on any of the above-mentioned nighttime road and bridge collapse early warning methods includes: Data acquisition module: Real-time acquisition of the position of the vehicle in front, the status of the vehicle's taillights, and roadside facade markings; Comparison module: Compare the disappearance behavior of the taillights of the preceding vehicle with the baseline model of the normal disappearance law of the taillights of the preceding vehicle, and compare the optical characteristics of the elevation markings with the normal optical characteristics model of the roadside elevation markings; Calculation module: Calculates the overall confidence level A using the first collapse confidence level A1 and the second collapse confidence level A2; Judgment module: Compares the overall confidence level A with the preset threshold to determine whether to issue a warning.

[0013] Compared with the prior art, the present invention has the following beneficial effects: (1) Strong real-time performance: Based on the vehicle's own visual perception system (such as a camera), the present invention monitors the taillight disappearance behavior of the vehicle in front and the optical characteristics of the roadside facade markings in real time. It can quickly detect the potential danger of bridge collapse at night and provide an immediate warning for vehicles in motion, avoiding the delay and non-real-time problems of traditional fixed sensors or manual inspections.

[0014] (2) Low cost: There is no need to deploy expensive fixed bridge monitoring equipment or additional hardware. The warning function can be realized by using the existing advanced driver assistance system (ADAS) visual sensors and computing modules, which reduces the overall implementation cost and facilitates large-scale promotion and application.

[0015] (3) High adaptability: In response to low visibility at night, the novel approach utilizes the disappearance pattern of the taillights of the vehicle in front and the optical reflection characteristics of roadside facade markings under low / high beam lights as detection basis, overcoming the limitations of traditional visual algorithms in dark conditions and improving the reliability and accuracy of bridge safety monitoring at night.

[0016] (4) Low false alarm rate: By using the preceding vehicle accident feature identification mechanism (such as lateral displacement and sharp turning analysis), the bridge collapse can be effectively distinguished from the preceding vehicle single-vehicle accident, avoiding false alarms; at the same time, the comprehensive confidence A is calculated by weighted fusion of dual confidence (A1 and A2), which improves the accuracy of early warning decision-making.

[0017] (5) Enhanced safety: Once the confidence level exceeds the threshold, multimodal warning (sound, light, touch) and automatic emergency braking are triggered to minimize the probability of accidents and enhance the active safety performance of automobiles. This is applicable to intelligent transportation systems. Attached Figure Description

[0018] The accompanying drawings are incorporated in and form part of this specification, and together with the description serve to explain the principles of the invention.

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart of the collapse early warning method of the present invention; Figure 2 This is a flowchart illustrating the collapse early warning method of the present invention. Detailed Implementation

[0021] Exemplary embodiments will now be described in detail. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples consistent with some aspects of the invention as detailed in the appended claims.

[0022] To enable those skilled in the art to better understand the technical solutions of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Example

[0023] On the one hand, this invention provides a method for early warning of nighttime road and bridge collapses based on visual perception, such as... Figure 1 As shown, it includes the following steps: Step 1: Establish a baseline model of the normal disappearance pattern of the taillights of the preceding vehicle and a normal optical characteristic model of the roadside elevation markings; Step 2: Real-time detection and tracking of vehicles ahead. If the disappearance behavior of the taillights of the vehicle ahead does not conform to the normal disappearance pattern of the taillights of the vehicle ahead, it is determined that an abnormal event has occurred. Then, it is determined whether the abnormal event is caused by an accident of the vehicle ahead. If it is determined to be an accident of the vehicle ahead, the system will not trigger a bridge collapse warning. Otherwise, a preliminary calculation of the collapse confidence is performed, and a first collapse confidence of A1 is assigned. At the same time, the roadside facade markings are analyzed. If the optical characteristics of the facade markings are detected to be inconsistent with the normal optical characteristic model, a second collapse confidence of A2 is assigned. Step 3: Perform a weighted fusion of the first collapse confidence level A1 and the second collapse confidence level A2 to obtain the final bridge collapse comprehensive confidence level A. Compare the bridge collapse comprehensive confidence level A with the preset threshold to determine whether to issue an early warning.

[0024] Specifically, such as Figure 2As shown, in step one, system initialization and baseline model establishment are performed. The system begins operation after the vehicle starts. The feature extraction and modeling module establishes two key baseline models based on pre-input route alignment data (which can come from high-precision maps or real-time sensing) and the vehicle's real-time status (such as speed and yaw rate): Model for the normal disappearance pattern of the taillights of the vehicle in front: Using geometric optics and kinematics models, the model calculates the expected disappearance position and trajectory of the taillights of the vehicle in front under the road conditions ahead (such as the top of a hill or a curve), as well as the expected curves of their brightness and area decreasing smoothly and continuously with distance.

[0025] A model for the normal optical characteristics of roadside facade markings: For bridge sections, the model learns and stores the spatial distribution characteristics of the reflected light intensity of facade markings such as guardrails under normal conditions, as well as the attenuation law with increasing distance, under the illumination of low beam and high beam headlights from vehicles. The core feature is that the reflection intensity of the same marking under high beam should be significantly higher than that under low beam.

[0026] Step Two: 2.1 Real-time Perception and Preliminary Identification of Abnormal Events. The image processing module analyzes the video stream input from the image acquisition module in real time, continuously outputting information on the position of the preceding vehicle, the taillight status, and roadside facade markings. The decision fusion module compares the real-time perceived taillight disappearance behavior of the preceding vehicle (including the disappearance point location, disappearance trajectory, and brightness change curve) with the normal disappearance pattern feature model established in Step One.

[0027] If the model is met: the system determines it to be in normal operating condition and continues monitoring.

[0028] If the model is not met (e.g., the taillights disappear prematurely in a "falling" manner on a straight road section): the decision fusion module determines that a primary abnormal event has occurred and triggers the next step of in-depth analysis.

[0029] 2.2 Identification of preceding vehicle accident characteristics. To avoid misjudging a single-vehicle accident as a bridge collapse, the decision fusion module will backtrack and analyze the trajectory of the preceding vehicle's taillights just before they disappear.

[0030] If the analysis results show obvious lateral displacement, sharp turning, or trajectory characteristics consistent with a collision, the system determines that the anomaly was caused by the accident of the vehicle in front, the process ends, and no bridge collapse warning is triggered.

[0031] If the aforementioned accidental lateral displacement characteristics are not detected, the system determines the event as "suspected bridge collapse" and quantifies this result as a preliminary collapse confidence level A1. Simultaneously, the process proceeds to the next step of cross-validation.

[0032] 2.3 Anomaly Analysis of Roadside Facade Markings. Under the background of a suspected bridge collapse, the system initiated a specific analysis of roadside facade markings. The feature extraction and modeling module extracted the real-time optical features of the facade markings on the bridge section within the current field of view and performed a refined comparison with the normal optical characteristic model established in step one. The focus was on analyzing whether, under high beam mode, large areas that should have been brightly reflective exhibited a sudden loss of reflected light or a "cliff-like" drop in brightness. Comparing low beam and high beam images, it was determined whether the facade markings in the suspected collapse area had lost their normal brightness response differences (i.e., no reflection in the area regardless of the lighting).

[0033] If the real-time optical properties are significantly inconsistent with the model, another preliminary collapse confidence level A2 is generated.

[0034] Step 3: 3.1 Confidence Fusion and Decision. The decision fusion module receives the confidence scores A1 and A2 from Steps 3 and 4, and fuses them using preset weighting coefficients k1 and k2 (where k1 + k2 = 1) to calculate the final comprehensive confidence score A for bridge collapse (A = k1*A1 + k2*A2). The weighting coefficients can be calibrated using a large amount of actual road test data to optimize system performance.

[0035] If A < preset threshold: the risk is considered not to have reached the warning level.

[0036] If A ≥ preset threshold: the bridge ahead is determined to have collapsed.

[0037] 3.2 Warning Trigger Once the warning conditions are met, the decision fusion module immediately sends a command to the warning execution module. The warning execution module first issues the highest level of collision warning to the driver through multiple means such as sound, light, and touch (e.g., a rapid alarm sound, a flashing red icon on the dashboard, and strong seat vibration). Simultaneously, the system can transmit signals to the body control system via the vehicle bus, triggering or assisting in automatic emergency braking and other operations to minimize the risk of the vehicle entering the collapsed area.

[0038] Furthermore, in step one, the baseline model for the normal disappearance pattern of the taillights of the preceding vehicle includes the location where the taillight brightness disappears, and the formula is: (1) Where P is the vanishing position, V r t is the relative speed, r is the time, θ is the radius of curvature, k is the road curvature, and d1 is the distance between the vehicle in front and the vehicle itself.

[0039] Specifically, in step one, establishing the benchmark optical feature model involves: A characteristic model of the normal disappearance pattern of the taillights of the preceding vehicle is established. Under known horizontal and vertical road alignment, based on the relative position and speed of the vehicle and the preceding vehicle and the road curvature, the expected position and trajectory of the taillights of the preceding vehicle disappearing normally in the driver's field of vision (such as due to downhill, curves, etc.) are calculated by using a geometric optics model. This feature describes the law that the brightness and area of ​​the taillights decrease smoothly and continuously with increasing distance.

[0040] Furthermore, the normal optical characteristic model in step one is represented by the following formula: (2) in, The reflected brightness at a distance of d2, in mode m (near beam = 0, high beam = 1), Let ϕ be the incident light intensity, and cos(ϕ) be the cosine of the incident angle. The mode efficiency factor is used to characterize the relative efficiency of the light source's contribution to the reflection of the target under different lighting modes (m=0 represents low beam, m=1 represents high beam); this coefficient is calibrated through actual road tests.

[0041] A normal optical characteristic model for roadside facade markings was established: For reflective facade markings on bridge railings, piers, etc., baseline curves were established showing the variation of reflected light intensity with distance when vehicle headlights are on (both low and high beams). This model clearly describes that, under normal conditions, the reflected light intensity of the same facade marking under high beam illumination should be significantly higher than that under low beam illumination, and the intensity exhibits a specific attenuation pattern with increasing distance.

[0042] Furthermore, in step two, the first collapse confidence level A1 is represented as: (3) A 2 =1 or A2=0 (4) Among them, P 测 To detect the location where the taillights of the preceding vehicle disappear, P is the predicted location where the taillights of the preceding vehicle disappear.

[0043] Specifically, when the taillights disappear abnormally and the situation is not a characteristic of an accident involving the vehicle in front, a preliminary collapse confidence level A1 is assigned.

[0044] When the optical properties of the facade markings do not conform to normal patterns, an additional preliminary collapse confidence level A2 is assigned.

[0045] If obvious lateral displacement, sharp steering, or a trajectory consistent with vehicle loss of control or collision is detected, the anomaly is determined to be caused by a single-vehicle accident ahead, i.e., A1=0.

[0046] Otherwise, we get formula (3). Among them, P 测 To detect the location where the taillights of the preceding vehicle disappear, P is the predicted location where the taillights of the preceding vehicle disappear.

[0047] If, in the high beam mode, there are large areas where the reflected light from the facade markings suddenly disappears or the brightness drops drastically, and in the low beam mode, the reflected brightness of the bridge facade is significantly inconsistent with the established model, A2=1; otherwise, it is 0.

[0048] Furthermore, in step two, determining whether the abnormal event was caused by the preceding vehicle specifically means: if the analysis results show obvious lateral displacement, sharp turning, or trajectory characteristics consistent with a collision, then the system determines that the abnormality was caused by the preceding vehicle's own accident.

[0049] Specifically, step 2.1: Preliminary identification of abnormal events. Real-time detection and tracking of vehicles ahead. When the disappearance position, trajectory, or brightness change of the taillights of the vehicle ahead does not conform to the normal disappearance pattern established in step one, an abnormal event is determined to have occurred. This abnormal event may originate from an accident involving the vehicle ahead or damage to the bridge ahead.

[0050] Step 2.2: Identify the characteristics of the preceding vehicle's accident and analyze the motion characteristics of the preceding vehicle's taillights just before they disappear. If obvious lateral displacement, sharp steering, or a trajectory consistent with vehicle loss of control or collision is detected, the anomaly is determined to be caused by a single-vehicle accident, and the system will not trigger a bridge collapse warning.

[0051] Step 2.3: Anomaly Analysis of Bridge Elevation Markings. If Step 3 determines that the anomaly is not caused by the accident ahead, then an in-depth analysis of the roadside elevation markings is initiated. The optical features of the bridge elevation markings in the current field of view are extracted and compared with the normal optical characteristic model established in Step 1: In the high beam mode, check for areas where there is a sudden loss of reflected light from large facade markings or a sharp drop in brightness.

[0052] Compare the brightness reduction pattern of facade markings in the same area under both low beam and high beam headlight modes to see if it matches the model. For example, in the collapsed area, regardless of whether the high beam or low beam is on, the markings cannot effectively reflect light, resulting in an abnormal "optical brightness reduction pattern".

[0053] Furthermore, in step three, the formula for the comprehensive confidence level A of bridge collapse is as follows: A = k1*A1 + k2*A2 (5) k1+k2=1 (6) Where k1 and k2 are the average weighting coefficients.

[0054] Furthermore, in step three, if A ≥ the threshold, it is determined that the probability of the bridge ahead collapsing is extremely high, the highest level of collision warning is issued, and automatic emergency braking is initiated; if A ≤ the threshold, it is considered that the risk has not reached the warning level.

[0055] Specifically, the overall confidence level A is compared with a preset threshold. If A ≥ the threshold, it is determined that the probability of the bridge ahead collapsing is extremely high, and the driver is immediately given the highest level of collision warning through sound, light, touch, and other means, and the vehicle control system can be linked to perform automatic emergency braking.

[0056] On the other hand, the present invention provides a system based on a nighttime road and bridge collapse early warning method, comprising: Data acquisition module: Real-time acquisition of the position of the vehicle in front, the status of the vehicle's taillights, and roadside facade markings; Comparison module: Compare the disappearance behavior of the taillights of the preceding vehicle with the baseline model of the normal disappearance law of the taillights of the preceding vehicle, and compare the optical characteristics of the elevation markings with the normal optical characteristics model of the roadside elevation markings; Calculation module: Calculates the overall confidence level A using the first collapse confidence level A1 and the second collapse confidence level A2; Judgment module: Compares the overall confidence level A with the preset threshold to determine whether to issue a warning.

[0057] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention.

[0058] It should be understood that the present invention is not limited to the content already described above, and various modifications and changes can be made without departing from its scope. The scope of the present invention is limited only by the appended claims.

Claims

1. A method for early warning of road and bridge collapse at night based on visual perception, characterized in that, Includes the following steps: Step 1: Establish a baseline model of the normal disappearance pattern of the taillights of the preceding vehicle and a normal optical characteristic model of the roadside elevation markings; Step 2: Real-time detection and tracking of vehicles ahead. If the disappearance behavior of the taillights of the vehicle ahead does not conform to the normal disappearance pattern of the taillights of the vehicle ahead, it is determined that an abnormal event has occurred. The system then determines whether the abnormal event was caused by an accident involving a vehicle ahead. If it is determined to be an accident involving a vehicle ahead, the system will not trigger a bridge collapse warning. Otherwise, a preliminary calculation of the collapse confidence level is performed, and a first collapse confidence level A1 is assigned. At the same time, the roadside facade markings are analyzed. If the optical characteristics of the facade markings do not conform to the rules of the normal optical characteristic model, a second collapse confidence level A2 is assigned. Step 3: Perform a weighted fusion of the first collapse confidence level A1 and the second collapse confidence level A2 to obtain the final bridge collapse comprehensive confidence level A. Compare the bridge collapse comprehensive confidence level A with the preset threshold to determine whether to issue an early warning.

2. The method for early warning of nighttime road and bridge collapse based on visual perception according to claim 1, characterized in that, In step one, the baseline model for the normal disappearance pattern of the taillights of the preceding vehicle needs to calculate the position where the taillight brightness disappears. The formula is: （1） Where Vr is the relative speed; t is the time; r is the radius of curvature; θ is the viewing angle; k is the road curvature; and d1 is the distance between the vehicle in front and the vehicle itself.

3. The method for early warning of nighttime road and bridge collapse based on visual perception according to claim 1, characterized in that, The normal optical characteristic model in step one requires calculating the reflectance at a distance of d2 and in mode m (near beam = 0, far beam = 1). The formula is: （2） in, Let ϕ be the incident light intensity, and ϕ be the incident light angle. d2 is the model efficiency factor, and d2 is the distance between the elevation mark and the vehicle.

4. The method for early warning of nighttime road and bridge collapse based on visual perception according to claim 1, characterized in that, In step two, the first collapse confidence level A1 and the second collapse confidence level A2 are expressed as follows: （3） A 2 =1 or A2=0 (4) Among them, P 测 To detect the location where the taillights of the preceding vehicle disappear, P is the predicted location where the taillights of the preceding vehicle disappear.

5. A method for early warning of nighttime road and bridge collapse based on visual perception according to claim 1, characterized in that, In step two, determining whether the abnormal event was caused by the vehicle in front specifically involves: if the analysis results show obvious lateral displacement, sharp turning, or trajectory characteristics consistent with a collision, then the system determines that the abnormality was caused by the accident of the vehicle in front itself.

6. A method for early warning of nighttime road and bridge collapse based on visual perception according to claim 1, characterized in that, In step three, the formula for the comprehensive confidence level A of bridge collapse is as follows: A = k1*A1 + k2*A2 (5) k1+k2=1 (6) Where k1 and k2 are the average weighting coefficients.

7. A method for early warning of nighttime road and bridge collapse based on visual perception according to claim 1, characterized in that, In step three, if A ≥ the threshold, it is determined that the probability of the bridge ahead collapsing is extremely high, the highest level of collision warning is issued, and automatic emergency braking is initiated; if A ≤ the threshold, it is considered that the risk has not reached the warning level.

8. A system based on the nighttime road and bridge collapse early warning method according to any one of claims 1 to 7, characterized in that, include: Data acquisition module: Real-time acquisition of the position of the vehicle in front, the status of the vehicle's taillights, and roadside facade markings; Comparison module: Compare the disappearance behavior of the taillights of the preceding vehicle with the baseline model of the normal disappearance law of the taillights of the preceding vehicle, and compare the optical characteristics of the elevation markings with the normal optical characteristics model of the roadside elevation markings; Calculation module: Calculates the overall confidence level A using the first collapse confidence level A1 and the second collapse confidence level A2; Judgment module: Compares the overall confidence level A with the preset threshold to determine whether to issue a warning.

Images