A method and apparatus for detecting road spills

Abstract

This invention relates to the field of intelligent transportation technology and discloses a method and apparatus for detecting road debris. One method includes: automatically selecting an infrared light wavelength based on ambient light intensity and calculating an environmental compensation coefficient; an infrared emitting unit emitting a multi-angle beam to an infrared receiving unit on the opposite side, with the receiving unit determining the beam's reception status; constructing a fan-shaped infrared beam grid to determine the area of ​​suspected targets; determining whether the suspected target is debris, and using an area positioning model to obtain the center position and effective area of ​​the debris; and implementing tiered alarms. This invention replaces visible light image recognition with infrared beam technology, and through an adaptive wavelength selection mechanism, effectively overcomes the impact of adverse weather conditions on detection accuracy. Combined with a lane recognition mechanism, it can accurately determine the specific lane where the debris is located, providing precise guidance for traffic control and cleanup operations.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation technology, and more specifically, to a method and apparatus for detecting road debris. Background Technology

[0002] With the rapid development of my country's transportation infrastructure, the mileage of highways and urban expressways is constantly increasing, and road safety management is facing new challenges. Road debris, as a significant traffic safety hazard, causes a large number of traffic accidents every year, especially under adverse weather conditions, significantly increasing the difficulty and importance of debris detection.

[0003] A Chinese patent with publication number CN111814668A discloses a method and apparatus for detecting road debris. The method involves acquiring a current frame image, images of a predetermined number of consecutive frames preceding the current frame image, and images of a predetermined number of frames extracted at predetermined intervals preceding the current frame image; acquiring a pre-generated initial background model; generating a short-frame background model and a long-frame background model based on the predetermined number of consecutive frames preceding the current frame image and the predetermined number of frames extracted at predetermined intervals preceding the current frame image, respectively; performing difference operations between the current frame image and the initial background model, the short-frame background model, and the long-frame background model, respectively, to obtain a first image, a second image, and a third image; and determining whether debris exists in the current frame image based on the first image, the second image, and the third image.

[0004] Existing technologies suffer from low detection accuracy, high false alarm rates, and high false negative rates when detecting road debris under adverse weather conditions such as heavy rain, heavy snow, and dense fog. In particular, image-based detection methods become almost ineffective when visibility is below 50 meters, failing to meet the needs of practical road safety management. This technological deficiency leads to many traffic accidents going unprepared, seriously threatening road traffic safety. Summary of the Invention

[0005] The purpose of this invention is to provide a method and apparatus for detecting road debris in order to solve the above-mentioned problems.

[0006] This invention provides a method for detecting road surface spills, comprising the following steps:

[0007] The infrared light wavelength is automatically selected based on the ambient light intensity, and the environmental compensation coefficient is calculated.

[0008] The road is divided into preset units. In each preset unit, infrared emitting units are activated one by one at preset time intervals. The infrared emitting units emit multi-angle beams of specific infrared wavelengths to the infrared receiving units on the opposite side, and the emission power of the infrared emitting units is adjusted according to the environmental compensation coefficient. The infrared receiving units determine the reception status of each beam.

[0009] Based on the beam reception status, a fan-shaped infrared beam grid based on a unified coordinate system is constructed. Based on the beam reception status of each detection point in the fan-shaped infrared beam grid, the location of the suspected target is determined.

[0010] To determine whether a suspected target is a spilled object, an area localization model is used to obtain the center position and effective area of ​​the spilled object.

[0011] A tiered alarm system is implemented based on the center location and effective area of ​​the spilled material.

[0012] Furthermore, the construction process of the fan-shaped infrared beam grid includes:

[0013] Based on the physical layout of the infrared emitting unit and the infrared receiving unit, calculate the spatial trajectory of multiple beams emitted by each infrared emitting unit.

[0014] The multi-angle beams emitted by different infrared emitting units will intersect each other in the road space, and the point where multiple beams intersect is called a detection point;

[0015] A unified coordinate system is used to describe the fan-shaped infrared beam grid. The coordinates (x, y) of the detection point represent the absolute physical position in the road space. The x-coordinate value is the lateral distance of the detection point from the left boundary of the road, and the y-coordinate value is the longitudinal distance of the detection point from the starting end of the road.

[0016] Furthermore, the origin of the coordinate system is set on the left boundary of the starting end of the road. The x-axis is perpendicular to the road travel direction and points to the right boundary of the road. The coordinate range is from 0 to the road width. The y-axis is along the road travel direction and points to the road forward direction. The coordinate range is from 0 to the length of the detected road segment. For curved roads, the coordinate system is curved along the road centerline, keeping the x-axis always perpendicular to the tangent of the road centerline and the y-axis always along the tangent of the road centerline.

[0017] Furthermore, the detection points traversed by the beam whose reception state is blocked are considered suspected detection points, and multiple adjacent suspected detection points are combined through connectivity analysis to form the area where the suspected target is located.

[0018] Furthermore, methods for determining whether a suspected target is a spilled object include:

[0019] Determine whether the suspected target is obscured by a vehicle;

[0020] Determine whether a suspected target not obscured by a vehicle is due to environmental interference;

[0021] For suspected targets that are not environmental disturbances, it is determined whether the duration of the obstruction is greater than the dynamic duration threshold. Suspected targets that are greater than the dynamic duration threshold are considered as debris. The dynamic duration threshold is the product of the preset base duration threshold and the environmental compensation coefficient.

[0022] Furthermore, the area localization model adopts a hybrid architecture combining multilayer perceptron and convolutional neural network, inputting the coordinates of multiple suspected detection points within the spill at the current moment, the number of passing beams, and the number of blocked beams into the area localization model.

[0023] Furthermore, the tiered alarm mechanism includes Level 1 alarms, Level 2 alarms, and Level 3 alarms;

[0024] Specifically, a Level 1 alarm is triggered when the risk score is greater than the Level 1 risk threshold but less than the Level 2 risk threshold; a Level 2 alarm is triggered when the risk score is greater than or equal to the Level 2 risk threshold but less than the Level 3 risk threshold; and a Level 3 alarm is triggered when the risk score is greater than or equal to the Level 3 risk threshold.

[0025] The risk score is the product of the lane weight coefficient and the effective area. The lane weight coefficient is determined based on the lane determination result of the spilled material. If the spilled material is located in the main lane, the lane weight coefficient is the main lane weight; if the spilled material is located in the overtaking lane, the lane weight coefficient is the overtaking lane weight; if the spilled material is located in the emergency lane, the lane weight coefficient is the emergency lane weight; if the spilled material is located on the right shoulder, the lane weight coefficient is the right shoulder weight; if the spilled material is located on the left shoulder, the lane weight coefficient is the left shoulder weight. The lane determination result of the spilled material is determined based on the center position of the spilled material.

[0026] Furthermore, when the environmental compensation coefficient is less than the preset power threshold, the transmission power is the base power; when the environmental compensation coefficient is greater than the preset power threshold, the transmission power is the additional power, which is greater than the base power.

[0027] The reception status includes pass and block. When the original signal strength received by the infrared receiving unit is greater than the preset signal strength threshold, the reception status is pass; when the original signal strength is less than or equal to the preset signal strength threshold, the reception status is block.

[0028] Furthermore, under standard weather conditions, the standard signal intensity received by each infrared receiving unit is recorded, and the ratio of the currently measured original signal intensity to the standard signal intensity is calculated to obtain the environmental compensation coefficient.

[0029] This invention provides an apparatus for detecting road debris, which stores computer-readable instructions that, when read, can execute a method for detecting road debris as described above. The apparatus includes:

[0030] The environment module automatically selects the infrared light wavelength based on the ambient light intensity and calculates the environmental compensation coefficient.

[0031] The module divides the road into preset units. Within each preset unit, infrared emitting units are activated one by one at preset time intervals. The infrared emitting units emit multi-angle beams of specific infrared wavelengths to the infrared receiving units on the opposite side, and the emission power of the infrared emitting units is adjusted according to the environmental compensation coefficient. The infrared receiving units determine the reception status of each beam.

[0032] The network module constructs a fan-shaped infrared beam grid based on a unified coordinate system according to the beam reception status. Based on the beam reception status of each detection point in the fan-shaped infrared beam grid, the location of the suspected target is determined.

[0033] The determination module determines whether a suspected target is a spilled object, and uses an area localization model to obtain the center position and effective area of ​​the spilled object.

[0034] The alarm module implements tiered alarms based on the center location and effective area of ​​the spilled material.

[0035] The beneficial effects of this invention are as follows: This invention replaces visible light image recognition with infrared beam technology, and effectively overcomes the impact of severe weather on detection accuracy through an adaptive wavelength selection mechanism. Even in extreme weather conditions with visibility below 10 meters, it maintains a high detection accuracy rate. By constructing a fan-shaped infrared beam grid and a unified road coordinate system, combined with an area positioning model, the center position and effective area are obtained, enabling the location of spilled materials. Combined with a lane recognition mechanism, the specific lane where the spilled material is located can be accurately determined, providing precise guidance for traffic control and cleanup operations.

[0036] This invention features adaptive environmental compensation, automatically adjusting detection parameters based on real-time environmental conditions. The environmental compensation coefficient can be calculated and applied in real time, ensuring stable detection performance under various weather conditions and eliminating the attenuation effect of environmental factors on signal strength. By analyzing the temporal-spatial characteristics of occlusion patterns, it effectively distinguishes between vehicle occlusion, environmental noise, and actual debris, reducing false alarm rates. The multi-layer height detection mechanism can identify different types of debris, improving the targeting of detection.

[0037] This invention implements a multi-level alarm mechanism, automatically determining the alarm level based on the center location and effective area of ​​the spilled material. Large spills on the main lane trigger a level three alarm and initiate an emergency response, while small suspected spills are only logged. This optimizes the allocation of safety management resources and avoids overreaction.

[0038] The infrared transmitting and receiving unit of this invention has a simple structure, high reliability, and lower maintenance costs than traditional camera systems. The system uses Power Bus communication, simplifying wiring and facilitating installation, making it particularly suitable for upgrading existing roads. Attached Figure Description

[0039] Figure 1 This is a flowchart illustrating a method for detecting road surface spills according to the present invention;

[0040] Figure 2 This is an example diagram of the infrared emitting unit and infrared receiving unit of a method for detecting road debris according to the present invention;

[0041] Figure 3 This is a flowchart illustrating a method for detecting road debris according to the present invention to obtain the center position and effective area;

[0042] Figure 4 This is a module example diagram of a device for detecting road spills according to the present invention. Detailed Implementation

[0043] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, features described in some examples may be combined in other examples.

[0044] A method and apparatus for detecting road spills include the following embodiments:

[0045] Example 1:

[0046] A method for detecting road spills, such as Figure 1 As shown, this method is applied to highways, urban expressways, and other traffic roads, especially in sections under adverse weather conditions such as heavy rain, fog, and snow. It comprises infrared emitting units, infrared receiving units, a zone controller, and a communication unit, forming a complete detection network. The infrared emitting and receiving units are installed at preset intervals, with a default interval of 12.5 cm, ensuring continuous beam coverage and detection accuracy. Each infrared emitting unit can emit multiple beams, with the maximum angle between all beams being the maximum emission angle, a default value of 30 degrees, to achieve fan-shaped coverage detection. The infrared receiving units are equipped with photodiodes and filters to improve anti-interference capabilities and enhance reception accuracy. The zone controller is responsible for controlling the infrared emitting and receiving units within a 1 km radius, achieving unified management and data processing. The communication unit uses a Power Bus communication method to ensure the stability and real-time performance of data transmission.

[0047] Includes the following steps:

[0048] Step 101: Automatically select the infrared wavelength based on ambient light intensity and calculate the environmental compensation coefficient. To ensure the stability of subsequent infrared beam transmission and reception, environmental adaptability configuration is first performed. An illumination sensor installed on the infrared receiving unit monitors the ambient light intensity in real time and automatically switches the infrared wavelength according to the intensity. When the detected light intensity is below the light intensity threshold, it is determined to be a nighttime environment and an 850 nm wavelength infrared light is automatically selected; when the light intensity is equal to or higher than the threshold, it is determined to be a daytime environment and a 940 nm wavelength infrared light is automatically selected. The light intensity threshold is typically set to 100 lux, equivalent to the natural light level at dusk or dawn. This wavelength selection mechanism is based on the differences in the penetration and anti-interference capabilities of infrared light under different lighting conditions. An 850 nm wavelength is selected when nighttime light is weak to obtain a better detection distance, while a 940 nm wavelength is selected when daytime light is strong to reduce sunlight interference. The initial setting of the light intensity threshold is derived from extensive experimental data analysis and can be fine-tuned during installation according to the specific road environment.

[0049] The selection of 850 nm and 940 nm wavelengths was based on atmospheric transmission characteristics and solar spectral analysis. The 850 nm wavelength lies within the first atmospheric window of the near-infrared spectrum, where atmospheric absorption and scattering are minimal, allowing for longer signal transmission distances, but it is susceptible to interference from near-infrared components of sunlight. The 940 nm wavelength, located at the edge of the water vapor absorption band, offers slightly shorter transmission distances, but suffers from lower solar radiation intensity at this wavelength, resulting in stronger resistance to solar interference. The selection of these two wavelengths also considered the quantum efficiency of silicon photodiodes at these wavelengths, both achieving detection efficiencies exceeding 80%.

[0050] Next, adaptive environmental compensation is implemented. By comparing the actual measured signal strength with the reference signal strength under ideal conditions, an environmental compensation coefficient is calculated. This coefficient is then used to dynamically adjust the signal strength threshold, effectively compensating for environmental impacts. Specifically, during installation and commissioning, the standard signal strength received by each infrared receiving unit is recorded as a reference value under standard weather conditions and stored in the area controller's memory. During actual operation, a calibration measurement is performed every 5 minutes, calculating the ratio of the currently measured signal strength to the stored reference value to obtain the environmental compensation coefficient. The calculation frequency of the environmental compensation coefficient is automatically adjusted according to the rate of weather change; in cases of rapid weather changes, such as sudden heavy rain, it can be increased to once per minute. The purpose of this environmental compensation mechanism is to eliminate the attenuation effect of environmental factors on signal strength, ensuring stable detection performance under different environmental conditions.

[0051] The measurements were taken under standard weather conditions: clear skies, a temperature of 20°C, relative humidity of 50%, and a wind speed of less than 3 m / s. Each infrared receiving unit recorded its corresponding reference value. The actual measured signal strength was obtained through real-time measurement under the same infrared transmission power settings. The measurement process excluded the influence of obstacles such as debris, reflecting only the attenuation effect of the environment on the signal.

[0052] Step 102: Divide the road into preset units, activate infrared emitting units one by one in each preset unit according to preset time intervals, and emit multi-angle beams of specific infrared light wavelengths to infrared receiving units on the opposite side, and adjust the emission power of infrared emitting units according to environmental compensation coefficients; infrared receiving units determine the reception status of each beam.

[0053] Specifically, such as Figure 2 As shown, it includes:

[0054] Based on the obtained infrared wavelength configuration and environmental compensation coefficient, a precise beam emission strategy is implemented. The road is divided into several preset units, each with a default length of 8 meters. Within each preset unit, infrared beam emission and reception are synchronized. Specifically, the infrared emitting unit emits multi-angle beams of specific infrared wavelengths to the infrared receiving unit on the opposite side. The beams are transmitted and received simultaneously, with each beam corresponding to a specific infrared receiving unit, where adjacent beams correspond to adjacent infrared receiving units. In practical applications, the middle portion of the beam emitted by the infrared emitting unit is primarily used to achieve the best detection effect and signal quality. The 8-meter unit length design is determined based on geometric optics principles. At the maximum emission angle, when the beam coverage of a single infrared emitting unit is 8 meters, the minimum distance between the emitting and receiving units is approximately 15 meters, similar to the standard width of a two-way four-lane highway. This design ensures that the beams emitted by the infrared emitting units within each preset unit do not interfere with each other in terms of time-division multiplexing and spatial distribution, effectively avoiding signal crosstalk between adjacent units, so that each infrared receiving unit receives only the beam emitted by a single infrared emitting unit at any given time.

[0055] A time-division multiplexing transmission strategy based on numbering is adopted. The infrared transmitting units in each preset unit are sequentially numbered along the road's travel direction, starting from 1 and incrementing. At any given time, all infrared transmitting units with the same number in all preset units are simultaneously activated, while other numbered units remain off. For example, at time t1, infrared transmitting unit numbered 1 in all preset units is activated; at time t2, infrared transmitting unit numbered 2 in all preset units is activated, and so on. Each numbered infrared transmitting unit operates sequentially according to a transmission time interval of 5–1000 milliseconds. Specifically, each infrared transmitting unit in a unit starts operating according to the transmission time interval, then turns off, and then the next infrared transmitting unit starts operating. The maximum angle between the multiple beams emitted by each infrared transmitting unit is the maximum transmission angle, with a default value of 30 degrees. This design ensures that the beams can effectively cover the detection area while avoiding mutual interference. The transmission time interval is determined based on signal processing theory and hardware response time. The rise time of the infrared emitting diode is 1-2 milliseconds, and the stable emission time requires 2 milliseconds. The response time of the photodiode is 0.5 milliseconds, and signal processing and data transmission require 1 millisecond. Therefore, a single complete transmit-receive-processing cycle takes approximately 4.5 milliseconds. Setting the minimum transmission interval to 5 milliseconds provides a 0.5 millisecond safety margin to ensure stable signal quality. Simultaneously, the 5-millisecond interval keeps the complete scan time of a single preset unit within a reasonable range; for example, the scan time for a unit covering 20 infrared emitting units is 100 milliseconds, meeting real-time detection requirements.

[0056] In practical applications, to obtain the best detection effect and signal quality, the central portion of the beam emitted by the infrared emitting unit is mainly used. The central portion beam is defined as the beam located within a certain proportion of the central region among all beams emitted by each infrared emitting unit. The specific selection method is as follows: First, determine the central portion proportion coefficient. This coefficient represents the proportion of the selected central beam to all beams, with a value greater than 0 and not exceeding 1. The default value is 0.6, meaning 60% of the beams are selected. Then, symmetrically eliminate the same number of edge beams from both ends of all beams. The remaining central beams are the desired central portion beams. For example, when an infrared emitting unit has 9 beams and the central portion proportion coefficient is 0.6, 60% of the beams need to be selected, i.e., 5 beams. Therefore, eliminate 2 beams from each end, ultimately selecting the 3rd to 7th beams as the central portion beams. These central portion beams have the best transmission characteristics and the lowest edge distortion because they are far from the edge region of the infrared emitting unit, experiencing less optical distortion and power attenuation. The range of the central portion beams can be dynamically changed by adjusting the central portion proportion coefficient value according to actual environmental conditions and detection requirements. This sequential, cyclical emission method ensures accurate identification of the obstruction status of each beam, effectively avoiding signal interference problems that may occur when multiple infrared emitting units operate simultaneously. This value is determined based on optical transmission theory and experimental optimization. The beam distribution of the infrared emitting unit exhibits a Gaussian pattern, with the central beam possessing the highest power density and optimal transmission characteristics, while the edge beams are significantly affected by optical distortion and power attenuation. Through experimental testing of detection accuracy at different scaling factors, it was found that a scaling factor of 0.6 maximizes the effective detection area while ensuring detection accuracy.

[0057] The infrared emitting units are controlled centrally by a regional controller, which sends control commands to each unit via a power bus. Each unit has a built-in microprocessor that receives these commands and executes the corresponding transmission operations. The maximum transmission power is dynamically adjusted based on an environmental compensation coefficient. When the coefficient is less than the power threshold, the transmission power is the base power, ranging from 200 to 1000 milliwatts. When the coefficient is greater than the threshold, the power is automatically increased to an additional 1.5 times the base power to enhance signal penetration. The default power threshold is 3.0. This power boosting mechanism is linked to the environmental compensation coefficient to ensure stable detection performance even under severe weather conditions with significant signal attenuation. The minimum base power of 200 milliwatts is set based on transmission requirements under standard atmospheric conditions. Under standard conditions of 10 km visibility and 60% relative humidity, 200 milliwatts ensures a signal strength more than three times the detection threshold within a 50-meter range. The additional power setting takes into account signal attenuation in severe weather. According to meteorological studies, infrared signal attenuation during heavy rain can be 2 to 3 times that under standard conditions. 300 milliwatts of power can maintain reliable detection under moderately severe weather.

[0058] The infrared receiving unit is responsible for recording the reception status of each beam. The reception status is determined by comparing the original signal strength received by the infrared receiving unit with a signal strength threshold. When the original signal strength is higher than the signal strength threshold, the beam is considered unobstructed, and the reception status is recorded as "passed." When the original signal strength is lower than or equal to the signal strength threshold, the reception status is recorded as "blocked." The initial setting of the signal strength threshold is based on extensive experimental data analysis, fully considering the signal attenuation characteristics under different weather conditions. The default value of the signal strength threshold is a preset multiple of the standard signal strength, and the default value of the preset multiple is 0.5.

[0059] The signal strength threshold is automatically adjusted based on an environmental compensation coefficient, employing a dynamic threshold adjustment algorithm to adapt to environmental changes. This algorithm utilizes the principle of inverse correction using the environmental compensation coefficient; that is, when environmental factors cause signal attenuation, the judgment threshold is lowered accordingly to maintain detection sensitivity. The adjusted signal strength threshold equals the default signal strength threshold divided by the environmental compensation coefficient. This adjustment method ensures that even in adverse weather conditions with signal attenuation, the beam reception status can still be determined with the same actual detection capability. To prevent false alarms due to over-adjustment of the threshold, upper and lower limits are set for threshold adjustment. The adjusted threshold must not be lower than 20% of the default signal strength threshold, nor higher than 150% of the default signal strength threshold. When the environmental compensation coefficient exceeds a reasonable range—less than 0.67 or greater than 5—an environmental anomaly alarm will be triggered, reminding technicians to check the hardware status or recalibrate the reference value. This dynamic threshold adjustment mechanism ensures accurate determination of the reception status under different weather conditions, avoiding missed detections in adverse weather conditions and preventing false alarms caused by slight environmental interference.

[0060] The infrared receiving unit employs a high-sensitivity photodiode array, coupled with a narrowband filter to enhance anti-interference capabilities. Each infrared receiving unit monitors a specific beam emitted by its corresponding infrared emitting unit and transmits the receiving status to the area controller in real time. The distance between each infrared emitting unit and each infrared receiving unit is the unit spacing, with a default value of 12.5 cm, ensuring precise beam alignment and stable transmission. The area controller aggregates all receiving status information to form complete road monitoring data, providing a foundation for subsequent debris detection. This binary receiving status recording method simplifies subsequent data processing. The binary receiving status includes either a passable or blocked state, where a blocked state indicates an obstacle in the beam path, and a passable state indicates an unobstructed beam path.

[0061] Step 103: Based on the beam reception status, construct a fan-shaped infrared beam grid based on a unified coordinate system, and determine the suspected target area based on the beam reception status of each detection point in the infrared beam grid.

[0062] Specifically, it includes:

[0063] Based on the collected beam reception status data, a fan-shaped infrared beam grid is constructed to represent the reception status distribution at various locations on the road. Since the infrared emitting unit is located on one side of the road, it emits multi-angle beams to the infrared receiving unit on the other side, forming a fan-shaped radiation grid rather than a square grid.

[0064] The process of constructing the fan-shaped infrared beam grid includes:

[0065] Based on the physical layout of the infrared emitting and receiving units, calculate the spatial trajectories of multiple beams emitted by each infrared emitting unit. Assume the infrared emitting unit is located on one side of the road and emits beams towards multiple infrared receiving units on the opposite side. Each infrared emitting unit emits M beams, and the angular interval between the beams is calculated by dividing the maximum emission angle by the number of beams minus one.

[0066] The spatial trajectory of the light beam is calculated based on the coordinate positions of the infrared emitting and receiving units. For the j-th infrared emitting unit, the coordinates are... The coordinates of the corresponding infrared receiving unit for the i-th emitted beam are: The spatial trajectory of the light beam is represented by the equation of a straight line determined by the emission and reception points, and the direction vector of this line in space is... The coordinates of any point on the spatial trajectory of the beam can be calculated using parametric equations:

[0067]

[0068] in, Represents the trajectory of the i-th beam, with parameters The value range is from 0 to 1. The corresponding infrared emitting unit, The corresponding infrared receiving unit.

[0069] The beam angle of the i-th beam is obtained by calculating the angle between the spatial trajectory of the i-th beam and the y-axis.

[0070] Because the multi-angle beams emitted by different infrared emitting units intersect each other in the road space, irregular detection areas are formed. For any point in the road space, it is necessary to determine which beams pass through that point. Points where multiple beams intersect are called detection points. The criterion for determining whether a beam has passed through a point is: the point follows the spatial trajectory of the beam.

[0071] A unified coordinate system is used to describe the fan-shaped infrared beam grid, specifically as follows: Figure 2As shown. The origin of the coordinate system is set on the left boundary of the road's starting point. The x-axis represents the road's lateral position, usually perpendicular to the road's direction of travel, with the positive direction pointing towards the right boundary of the road, and its coordinate range is 0 to the road width. The y-axis represents the road's longitudinal position, usually along the road's direction of travel, with the positive direction pointing towards the road's forward direction, and its coordinate range is 0 to the length of the detected road segment. For curved roads, the coordinate system curves along the road's centerline, keeping the x-axis always perpendicular to the tangent of the road's centerline, and the y-axis always along the tangent of the road's centerline. Therefore, the detection point (x, y) represents the absolute physical position in the road space; the x-coordinate value is equal to the lateral distance of the point from the left boundary of the road, and the y-coordinate value is equal to the longitudinal distance of the point from the road's starting point. All infrared emitting and receiving units within the entire detection area use this unified coordinate system, ensuring that the multi-angle fan-shaped beams of multiple infrared emitting units can be cross-analyzed and accurately positioned under the same coordinate system.

[0072] For each detection point (x, y) in the fan-shaped infrared beam grid, the reception state of all beams passing through that detection point is calculated. In a unified coordinate system, the same detection point (x, y) may be simultaneously covered by beams from multiple different infrared emitting units. By traversing all infrared emitting units, identifying all beams passing through the detection point, and comprehensively analyzing the reception states of these beams, a comprehensive detection result for that point is obtained. This multi-beam coverage mechanism significantly improves the reliability and accuracy of the detection.

[0073] In a fan-shaped infrared beam grid, a suspected target refers to a spatial region where obstacles may exist, identified based on the reception status of detection points. Specifically, when a certain number of beams passing through a detection point (x, y) show a blocking state, that detection point is marked as a suspected detection point. Multiple adjacent suspected detection points are combined through connectivity analysis to form the region containing the suspected target. Suspected targets may include various factors that cause beam blocking, such as real road debris, passing vehicles, animals, equipment malfunctions, or environmental interference, and their true nature needs to be further distinguished in subsequent steps using intelligent judgment algorithms.

[0074] Furthermore, a multi-layer height detection mechanism can be implemented. By installing several layers of infrared emitting and receiving units at different heights, the intersection of infrared beams at different height levels can be detected, allowing the inference of the height and volume of the spilled object. For example, three layers of infrared emitting and receiving units at different heights are installed on both sides of the road, located at 15 cm, 30 cm, and 45 cm above the ground, respectively. Each layer of infrared emitting and receiving units is installed at intervals to ensure that each beam has a corresponding receiving unit. By comparing the obstruction at different height levels, small spilled objects, such as bolts and stones, and large spilled objects, such as tires and cargo, can be distinguished. For example, if only the lowest layer detects obstruction, it is determined to be a small spilled object with a height of less than 15 cm; if the lowest two layers detect obstruction, it is determined to be a medium-sized spilled object with a height between 15 and 30 cm; if all three layers detect obstruction, it is determined to be a large spilled object with a height exceeding 45 cm. This multi-level detection mechanism improves the ability to identify different types of spilled objects and provides a basis for subsequent alarm classification.

[0075] Step 104: Determine whether the suspected target is a spilled object. For confirmed spilled objects, use the area positioning model to obtain the center position and effective area.

[0076] specific Figure 3 As shown, it includes:

[0077] Based on the area where the suspected target is located, a two-stage processing method is adopted: first, the interference filtering algorithm is used to determine whether the suspected target is a dumped object, and then the area positioning model is used to calculate the center position and effective area of ​​the confirmed dumped object.

[0078] Within the fan-shaped infrared beam grid, interference patterns of suspected targets are first identified, and the temporal-spatial characteristics of the occlusion patterns are analyzed to distinguish between real projectiles and various interference factors.

[0079] To determine whether a suspected target is obstructed by a vehicle, vehicle obstruction is identified by analyzing the temporal-spatial characteristics of the obstruction pattern. Within a fan-shaped infrared beam grid, vehicle obstruction manifests as a specific fan-shaped obstruction sequence.

[0080] An obstruction pattern refers to the distribution characteristics of the reception status of multiple infrared receiving units over a specific time period, varying with time and spatial location. A vehicle obstruction pattern consists of the obstructed receiving unit, its corresponding reception status, and the beam angle. The reception status represents the reception state of a specific infrared receiving unit at a given time, and is categorized into two states: pass-through or blockage.

[0081] The occlusion features produced by vehicles within a fan-shaped infrared beam grid exhibit typical regularities:

[0082] The occlusion moves laterally along the road, manifesting as beams of light at adjacent angles being blocked in sequence, forming a continuous angular occlusion sequence.

[0083] The obstruction width is related to the actual width of the vehicle and the distance between the vehicle and the infrared emitting unit. The closer the vehicle is to the infrared emitting unit, the larger the observed obstruction angle range; the farther the distance, the smaller the observed obstruction angle range.

[0084] The duration of occlusion is related to vehicle speed and beam span. The duration of vehicle occlusion is determined by both the effective coverage distance of the beam at the vehicle's location and the vehicle's speed; the faster the speed, the shorter the occlusion duration.

[0085] Vehicle obstruction usually affects detection units at all height levels simultaneously, because the vehicle height is much greater than the highest detection level by 45 centimeters.

[0086] The specific criteria for determining a suspected target to be a vehicle obstructing the view include:

[0087] The angular velocity of the obstruction is calculated, and the moving speed is estimated by the angle and time difference of the continuously obstructed light beams. The condition for continuous movement is met when the angular velocity is within a reasonable vehicle speed range (corresponding to speeds of 10-160 km / h) and the obstruction propagates continuously along the angular sequence. This range is determined based on Chinese road traffic regulations and actual surveys. Urban road speed limits are typically 30-80 km / h, and highway speed limits are 60-120 km / h. Considering speeding and emergency vehicles, the upper limit is set at 160 km / h. The lower limit of 10 km / h corresponds to the slow movement of congested or disabled vehicles.

[0088] The duration of vehicle obstruction should be within a reasonable range, i.e., greater than the default minimum obstruction time threshold and less than the maximum obstruction time threshold. The default minimum obstruction time threshold is 0.1 seconds, corresponding to vehicles passing at high speeds; the default maximum obstruction time threshold is 3 seconds, corresponding to large vehicles passing slowly. 0.1 seconds corresponds to the time it takes for a vehicle to pass an 8-meter detection unit at a speed of 160 km / h, and 3 seconds corresponds to the time it takes for a large vehicle to pass at a speed of 10 km / h. This time range covers the passing characteristics of all normally moving vehicles.

[0089] Vehicle obstruction must simultaneously affect all installed height layer detection units to form a complete obstruction column in the vertical direction.

[0090] Vehicle occlusion feature templates for a fan-shaped infrared beam grid are pre-stored in the area controller. These templates take into account the differences in occlusion patterns of different types of vehicles at different distances and angles from the infrared emitting unit. Due to the non-uniformity of the fan-shaped infrared beam grid, the occlusion pattern of the same vehicle will differ at different locations: the near-end region has high beam density and high angular resolution, which can more accurately describe the vehicle outline; the far-end region has low beam density and low angular resolution, and the detailed features of vehicle occlusion are relatively blurred. When the matching degree between the detected occlusion pattern and these feature templates exceeds the vehicle recognition matching degree threshold, the template matching condition is met.

[0091] If a suspected target meets all of the above conditions, the occlusion is determined to be caused by a vehicle; otherwise, it is determined to be a non-vehicle occlusion, and the debris analysis continues. The default value for the vehicle identification matching threshold is 85%.

[0092] To determine whether a suspected target not obscured by a vehicle is an environmental interference, various types of environmental interference are identified by analyzing the multidimensional characteristics of the obscuration pattern, making full use of the characteristics of the fan-shaped infrared beam grid and the environmental compensation mechanism established in the previous steps.

[0093] Based on environmental compensation coefficient and light intensity monitoring data, signal interference caused by severe weather can be identified.

[0094] When the environmental compensation coefficient fluctuates drastically within a short period, such as 1-3 minutes, and the fluctuation amplitude exceeds the environmental fluctuation threshold, it indicates the presence of rain and fog interference. Rain and fog interference manifests as a large-scale signal intensity attenuation within the fan-shaped infrared beam grid, but it does not produce a clear spatial boundary; the occlusion pattern exhibits a diffuse distribution. The judgment criteria include: the occlusion area covers more than 80% of the detection range; the signal intensity attenuation shows a gradual distribution rather than an abrupt change in space; all height layers are affected simultaneously and to similar degrees; and the occlusion intensity is highly correlated with the current environmental compensation coefficient, with a correlation coefficient greater than 0.8. The default value for the environmental fluctuation threshold is 1.5, indicating that the environmental compensation coefficient changes by more than 50% within a short period.

[0095] In heavy snow, snowflakes randomly obstruct the path of the infrared beam, manifesting as a high-frequency random flickering pattern within the fan-shaped infrared beam grid. The criteria for determination include: the obstruction pattern is randomly distributed with no obvious spatial clustering; the obstruction duration is extremely short, less than 0.2 seconds, and occurs frequently; the obstruction intensity is weak, typically affecting only a portion of the beam rather than completely blocking it.

[0096] Combining light intensity monitoring and an adaptive infrared wavelength selection mechanism, various light-related interferences are identified. When the light intensity sensor detects strong sunlight, exceeding the strong light threshold, and signal saturation occurs in beams at specific angles, it is determined to be direct sunlight interference. In a fan-shaped infrared beam grid, direct sunlight primarily affects beams in specific directions, manifesting as highly directional signal anomalies. Judgment criteria include: the abnormal signal is concentrated within a specific angular range, typically towards the sun; the signal intensity of the affected beam exceeds the upper limit of the normal range rather than the lower limit; the interference moves with the sun's position, exhibiting a clear temporal regularity; the interference is significantly reduced when switching to a 940 nm wavelength; and in multi-layer detection, the upper layer is typically more affected than the lower layer. The default value for the strong light threshold is 1000 lux.

[0097] Water accumulation on the road surface, reflections from metal objects, or glass can generate strong localized signals within a fan-shaped infrared beam grid. Criteria for determining this include: abnormally high signal strength rather than abnormally low strength; the interference area typically has a regular geometric shape, such as a rectangle or circle; the interference location corresponds to a known reflective light source, such as road markings or guardrails; the interference intensity varies with the angle of illumination; and it only affects a specific height layer, usually the lowest layer.

[0098] Based on cyclic transmission technology and a power bus communication mechanism, interference caused by equipment failure can be identified. When the transmission power of an infrared transmitting unit is abnormal or completely fails, a systematic signal loss will occur in the corresponding receiving area. The judgment conditions include: all beam signals within a specific sector are abnormal simultaneously; the abnormal mode perfectly matches the coverage area of ​​the infrared transmitting unit; the fault mode persists and does not change over time; the power bus communication shows that the corresponding infrared transmitting unit is communicating abnormally; and other infrared transmitting units are working normally, forming a clear contrast.

[0099] Localized detection anomalies are caused by damage to the photodiode or contamination of the filter in the infrared receiving unit. Judgment criteria include: a single or a few adjacent infrared receiving units continuously displaying an abnormal state; the anomaly not changing with environmental conditions; all beams corresponding to the abnormal infrared receiving unit displaying the same abnormal pattern; and normal detection being restored by switching to a backup infrared receiving unit.

[0100] Data transmission anomalies occur when the power bus is subjected to electromagnetic interference. Judgment criteria include: packet loss rate exceeding the communication anomaly threshold; obvious encoding errors in received data; periodic interference, typically related to the operating cycle of nearby electrical equipment; and temporary recovery by resending commands. The default value for the communication anomaly threshold is 5%.

[0101] Utilizing the multi-layer detection characteristics and high temporal resolution of a fan-shaped infrared beam grid, biological interference caused by small animals and birds can be distinguished. When birds fly over the detection area, they create rapid-moving obstructions within the fan-shaped infrared beam grid. Judgment criteria include: extremely short obstruction duration (0.1–1 second); obstruction path exhibiting a clear flight trajectory, typically a straight line or arc; primarily affecting upper detection units at a height of 30–45 cm, with minimal or no impact on lower units; small effective area but high movement speed; and high obstruction intensity but clear edges.

[0102] Occlusion caused by small animals such as cats and dogs crossing the road. Judgment criteria include: moderate occlusion duration, 2-10 seconds, depending on the animal size and movement speed; primarily affecting the lowest layer detection unit, 15 cm in height, and possibly affecting the middle layer; the occlusion path exhibits an irregular movement pattern, which may include behaviors such as pauses and turns; and the movement speed is within the typical speed range of small animals, 0.5-5 m / s.

[0103] Periodic shading caused by strong winds swaying roadside vegetation or hanging objects. Criteria for identification include: the shading pattern exhibits a clear periodicity, typically with a period of 2–8 seconds; the shading location is relatively fixed, usually in the roadside area; the shading intensity varies with wind speed; it primarily affects the edge beams, with the central area rarely affected; and when environmental monitoring indicates strong winds, the sensitivity to vegetation disturbance should be increased.

[0104] The impact of atmospheric density gradient on infrared signals under extreme temperature conditions. Criteria include: a slow, gradual attenuation of signal strength; interference levels correlated with temperature differences, with greater temperature differences having a more pronounced impact; primarily affecting detection accuracy in distant regions; and interference changing slowly over time, typically consistent with diurnal temperature variations.

[0105] When the occlusion pattern of a suspected target meets the characteristic conditions of any of the above-mentioned types of environmental interference, it is classified as the corresponding type of environmental interference and excluded from the list of candidates for projectiles.

[0106] For suspected targets that are not environmental interference, it is determined whether the duration of occlusion exceeds a dynamic duration threshold. Suspected targets exceeding the dynamic duration threshold are classified as projectiles. The dynamic duration threshold is the product of a preset base duration threshold and an environmental compensation coefficient. The key characteristic that distinguishes projectiles from transient interference is the persistence and stability of their occlusion. Real projectiles do not produce transient, irregular occlusion patterns like birds or snowflakes.

[0107] The specific methods for determining the duration of occlusion include:

[0108] When the number of blocked beams in the area where the suspected target is located first exceeds the minimum blocked beam number threshold, a blocking duration timer is started. The minimum blocked beam number threshold is dynamically calculated based on the beam density in the area where the suspected target is located, and is generally set to 30% of the total number of beams in that area to ensure that sporadic signal interference is filtered out. At the same time, the start timestamp of the blocking event and the initial blocking characteristic parameters are recorded, including basic information such as the spatial distribution of the blocked beams, the coordinates of the center position, and the blocking intensity.

[0109] Within each detection cycle of sequential transmission, the persistence of the suspected target's obstruction state is verified. Continuity verification employs a sliding time window method, with the window length set as the continuity verification window, defaulting to 200 milliseconds and encompassing approximately 40 detection cycles. Within this time window, the continuity ratio of the obstruction state is calculated, i.e., the proportion of detection cycles in which the number of blocked beams exceeds the minimum blocked beam number threshold out of the total detection cycles. When the continuity ratio is greater than or equal to the continuity ratio threshold, the obstruction state is considered continuous and valid; when the continuity ratio is less than the continuity ratio threshold, the obstruction is considered interrupted, and the duration timer is reset. The default value for the continuity ratio threshold is 80%, allowing for temporary signal recovery in 20% of detection cycles to accommodate random noise in signal transmission.

[0110] Continuous monitoring of the spatial stability of areas containing suspected targets prevents moving targets from being misidentified as static projectiles. Spatial stability is assessed by calculating the change in the position of the obstruction center over consecutive detection cycles. Specifically, within each detection cycle, the centroid coordinates of the obstruction area are calculated based on the spatial distribution of the blocking beam, and then the distance offset between the current and initial centroid coordinates is calculated. When the distance offset is less than or equal to the position stability threshold, the obstruction position is considered stable; when the distance offset is greater than the position stability threshold, it is identified as a moving target, the duration timer is reset, and the obstruction event is reinitialized. The position stability threshold is dynamically adjusted based on the distance of the detection point from the infrared emitting unit: 0.3 meters for near-field regions (distance less than 5 meters), 1.0 meter for far-field regions (distance greater than 15 meters), and linear interpolation is used for intermediate regions.

[0111] In addition to positional stability, the stability of obstruction intensity is also evaluated to ensure that the actual projectile does not cause drastic changes in obstruction intensity. Obstruction intensity is defined as the ratio of the number of blocked beams to the total number of beams passing through the area. The standard deviation of obstruction intensity is calculated over consecutive detection periods. When the standard deviation is less than or equal to the intensity stability threshold, the obstruction intensity is considered stable; when the standard deviation is greater than the intensity stability threshold, it is considered unstable obstruction, and the corresponding detection period is not included in the cumulative duration. The default value for the intensity stability threshold is 0.15, indicating that the variation in obstruction intensity should not exceed 15%.

[0112] Only detection cycles that simultaneously meet the requirements of continuity, spatial stability, and intensity stability are included in the effective duration accumulation. The cumulative duration equals the number of effective detection cycles multiplied by the length of a single detection cycle, which is 5 milliseconds. When the cumulative duration reaches the dynamic duration threshold, the suspected target is determined by the time persistence feature; when the occlusion event ends but the cumulative duration does not reach the dynamic duration threshold, the suspected target is excluded from the projectile candidate list, and the occlusion event ends when the number of blocking beams drops below the minimum blocking beam number threshold.

[0113] The dynamic duration threshold is the product of a preset base duration threshold and an environmental compensation coefficient. The default value for the base duration threshold is 2 seconds.

[0114] When an abnormal interruption occurs during the calculation of the occlusion duration, such as equipment failure or communication interruption, a fault-tolerant mechanism is employed. If the interruption time is less than the fault-tolerant time threshold (default value is 1 second), the duration continues to accumulate after recovery. If the interruption time is greater than the fault-tolerant time threshold, the duration timer is reset and the judgment process restarts. This design ensures that reliable detection performance is maintained even in the face of technical failures.

[0115] For the projectiles that pass the determination, an area positioning model is used to calculate the center position and effective area. The area positioning model is specifically designed and trained for the characteristics of the fan-shaped infrared beam grid.

[0116] The area localization model adopts a hybrid architecture that combines a multilayer perceptron (MLP) and a convolutional neural network (CNN).

[0117] The area localization model comprises four core components: an input feature extraction module, a convolutional feature encoding module, a multilayer perceptron mapping module, and a regression output module. The input feature extraction module converts the raw detection data within the suspected target's region into a structured feature representation. The convolutional feature encoding module, based on a convolutional neural network, learns the spatial adjacency relationships and local patterns between detection points in a fan-shaped infrared beam grid. The multilayer perceptron mapping module maps the spatial features extracted by convolution to a high-dimensional feature space and performs a nonlinear transformation. The regression output module outputs key parameters such as the center coordinates of the projectile and the effective area.

[0118] For each suspected target region identified in the first stage, an input feature containing the following information is constructed:

[0119] The basic spatial features are for each suspected detection point within the suspected target area, including the coordinates of the suspected detection point, the total number of beams passing through the detection point, and the number of beams blocked. The coordinates of the suspected detection point represent the absolute position of the suspected target area in the coordinate system, the total number of beams passing through the suspected detection point reflects the beam density at that position, and the number of blocked beams reflects the occlusion intensity.

[0120] The area localization model outputs the coordinates of the center position of the sprayed material based on the input features. In, and effective area.

[0121] Model training employs multiple methods to collect training data. Real-world calibration involves placing standard projectiles of known size and location on the test road to collect corresponding detection data. Historical data annotation involves manually annotating historical detection data to establish a correspondence between input features and center positions / effective areas. Simulation data generation uses a geometric model based on a fan-shaped infrared beam grid to generate simulation training data under different conditions. Data augmentation expands the training dataset by adding noise, rotation, scaling, and other techniques.

[0122] The model training strategy employs a multi-task learning loss function, simultaneously optimizing both positional and area regressions. Dropout and weight decay are used for regularization to prevent overfitting. Sample balancing is performed for projectiles of different sizes and locations, and k-fold cross-validation is used to evaluate the model's generalization performance.

[0123] Model deployment employs quantization and pruning to compress the model, reducing computational complexity to suit the computational capabilities of the region controller. Inference time is kept below 100 milliseconds, meeting real-time detection requirements.

[0124] Step 105: Implement tiered alarms based on the center location and effective area of ​​the spilled material;

[0125] Specifically, it includes:

[0126] Based on the spillage detection results, lane determination is first performed according to the coordinates of the spillage center, and then a tiered alarm mechanism is implemented. Due to the characteristics of the fan-shaped infrared beam grid, the determination of the alarm level needs to consider the impact of the detection location on accuracy, and a weighted assessment of the hazard level of spillage in different lanes is performed.

[0127] Based on the output coordinates of the center position of the projectile And the effective area, automatically determine the lane where the spilled material is located, among which, The x-coordinate represents the center position of the spilled material. This represents the ordinate of the center position of the spilled material. During the initialization phase, a lane boundary mapping table is created to store the lateral position range of each lane.

[0128] The lane boundary mapping table is established by first obtaining the total road width and lane configuration information, including basic data such as the number of lanes, the width of each lane, and the shoulder width. Then, the boundary coordinates of each lane are calculated. The x-coordinate range of lane 1 (leftmost) is [left shoulder width, left shoulder width + lane 1 width], the x-coordinate range of lane 2 is [left shoulder width + lane 1 width, left shoulder width + lane 1 width + lane 2 width], and so on, until the rightmost lane. Lane numbering follows a unified rule: main lanes are numbered 1 to N (from left to right), left shoulder is numbered 0, and right shoulder is numbered N+1.

[0129] By comparing the coordinates of the center position of the spilled material Determine the lane where the spilled material is located by considering the boundaries of each lane. The specific determination rule is as follows: When the spilled material falls within the boundary of a lane, it is identified as being located in that lane; if If it is less than the left boundary of lane 1, it is identified as the left shoulder; if If the distance is greater than the right boundary of the last lane, it is identified as the right shoulder.

[0130] For large spills that span multiple lanes, when the effective area is large and crosses multiple lane boundaries, an area-weighted algorithm is used. This algorithm calculates the area ratio of the spill within each lane, designates the lane with the largest ratio as the primary lane, and marks all involved lanes for subsequent alarm processing.

[0131] Based on lane numbering and road configuration, lanes are divided into different types. The main lane is the normal driving lane and has the highest risk level; the overtaking lane is a dedicated overtaking lane and has the second highest risk level; the emergency lane is an emergency stopping lane and has a lower risk level; the shoulder is the edge area of ​​the road and has the lowest risk level.

[0132] Lane recognition results are formatted as lane number-lane type-hazard level, such as 2-main lane-high hazard, 0-left shoulder-low hazard, etc. Dynamic lane configuration is supported; when the number or width of road lanes changes, the lane boundary mapping table can be updated to adapt to the new road layout.

[0133] Based on the confirmed information about the spilled material, a three-tiered alarm mechanism is implemented according to the effective area of ​​the spilled material, its location in the lane, and its height.

[0134] First, a risk score is calculated for each confirmed spill, serving as the core basis for tiered alarms. The risk score is the product of the lane weight coefficient and the effective area. The lane weight coefficient is determined based on the lane determination result of the spill: if the spill is located in the main lane, the lane weight coefficient is the main lane weight; if the spill is located in the overtaking lane, the lane weight coefficient is the overtaking lane weight; if the spill is located in the emergency lane, the lane weight coefficient is the emergency lane weight; if the spill is located on the right shoulder, the lane weight coefficient is the right shoulder weight; if the spill is located on the left shoulder, the lane weight coefficient is the left shoulder weight.

[0135] The default value for the weight coefficient of the main lane is 1.0, which is the area with the highest traffic volume and the most severe impact; the default value for the weight coefficient of the overtaking lane is 0.9, which is less important than the main lane; the default value for the weight coefficient of the emergency lane is 0.5, which is mainly used for emergency parking and has less traffic under normal circumstances; the default value for the weight coefficient of the left shoulder is 0.3, and the default value for the weight coefficient of the right shoulder is 0.2. They are located at the edge of the road and have a relatively small impact on traffic.

[0136] A Level 1 alarm is triggered when the risk score is greater than the Level 1 risk threshold but less than the Level 2 risk threshold. The default value for the Level 1 risk threshold is 0.1, and the default value for the Level 2 risk threshold is 0.3.

[0137] Level 1 alerts primarily target small debris located on the shoulder or emergency lane, or debris located in the main lane but covering a small area. These types of debris have a relatively minor impact on traffic safety, but still require attention and handling.

[0138] When a Level 1 alarm is triggered, perform the following operations: Record the spilled material information in the alarm log of the area controller, including the detection time and precise center location coordinates. The system displays the effective area, lane identification results, height layer distribution, and risk score; it marks the location of the spilled material in blue on the electronic map in the monitoring center; it sends a Level 1 alarm message to the monitoring center via the power bus; it automatically generates standard handling suggestions, recommending cleanup within the next maintenance cycle, without requiring emergency handling; and it records the alarm level as Level 1 - Small Road Spilled Material.

[0139] A level 2 alarm is triggered when the risk score is greater than or equal to the level 2 risk threshold but less than the level 3 risk threshold. The default value for the level 3 risk threshold is 0.6.

[0140] Level 2 alerts are primarily for medium-sized debris spills, or small debris spills located in the main lane or overtaking lane. These types of debris spills have a significant impact on traffic safety and require timely handling.

[0141] When a level 2 alarm is triggered, perform the following actions: Highlight the location of the spilled material with a yellow marker on the electronic map in the monitoring center, and use the output precise center location coordinates. Detailed alarm information is sent to the monitoring center via the power bus, including detection time, complete location information (road number, direction of travel, lane identification result, kilometer marker), detailed characteristics of the spilled material (effective area, center coordinates, height distribution, risk score), and current environmental conditions (weather status, environmental compensation coefficient). Based on the lane identification results, automatic handling suggestions are generated: spilled material in the main lane is recommended to be handled within 2 hours, in the overtaking lane within 4 hours, and in the emergency lane within 8 hours. The alarm information, including the lane location of the spilled material and the recommended handling time, is sent to the on-duty road management personnel via SMS or mobile application. The alarm level is recorded as Level II - Medium-sized road spilled material.

[0142] A Level 3 alarm is triggered when the risk score is greater than or equal to the Level 3 risk threshold.

[0143] Level 3 alarms are primarily for large-scale debris spills, medium to large-sized debris spills located on the main lanes, or debris spills at a considerable height. Such debris poses a serious threat to traffic safety and requires immediate attention.

[0144] When a Level 3 alarm is triggered, the emergency response procedure is executed as follows: the location of the spilled material is highlighted in red on the electronic map of the monitoring center and flashed on the large screen; an audible alarm is issued to alert the monitoring center staff; the emergency response team and relevant management personnel are immediately notified through various means such as telephone, SMS, and app push notifications; an emergency response plan is automatically generated based on the lane assessment results and the characteristics of the spilled material, including suggested traffic control measures such as temporary closure of relevant lanes, vehicle dispatching plans, estimated handling time, and warning measures for upstream vehicles; for spilled materials located in the main lane with an effective area greater than 0.5 square meters, it is recommended to immediately close the lane; warning information is issued to upstream vehicles through the road information dissemination system, such as variable message signs, with content customized based on the lane assessment results; a complete event log is recorded, including the alarm trigger time, detailed characteristics of the spilled material, emergency measures taken, and handling completion time; the alarm level is recorded as Level 3 - Large Hazardous Spilled Material.

[0145] Develop specific response strategies based on lane determination results, and fully utilize lane information to optimize resource allocation:

[0146] For debris scattered on the main lane, regardless of size, it should be dealt with first, with the strictest handling time requirements. When the effective area exceeds 0.3 square meters, lane control will be automatically recommended. For debris scattered on the overtaking lane, focus should be placed on medium and large debris with an effective area greater than 0.2 square meters. Small debris can be dealt with appropriately later. For debris scattered on the emergency lane, unless the area is particularly large, exceeding 0.8 square meters, or the height is very high, affecting all three levels, it is generally classified as a level two alarm, allowing for a longer handling time window. For debris scattered on the shoulder, the main concern is whether it may slide into the driving lane, and secondary hazards should be assessed based on the location of the debris and the road slope.

[0147] It also implements an alarm tracking mechanism to continuously monitor changes in the status of spilled materials. If the spilled materials are detected to have been cleared and the obstruction removed, the alarm will be automatically deactivated and the completion time will be recorded. This closed-loop management mechanism ensures that every alarm is properly handled, improving the efficiency of road safety management.

[0148] Through the above-mentioned graded alarm mechanism, corresponding measures can be taken according to the actual degree of danger of the spilled material. This avoids overreaction to small, low-risk spills while ensuring timely handling of large, high-risk spills, thus achieving optimized allocation of road safety management resources.

[0149] The method provided in this embodiment achieves accurate detection of road debris under adverse weather conditions through five steps, and implements tiered alarms based on the detection results, providing effective support for road safety management. The entire detection process fully adapts to the characteristics of a fan-shaped beam grid with single-sided transmission and contra-sided reception. Through various compensation and correction algorithms, it ensures reliable detection of road debris under various adverse weather conditions, providing strong protection for road traffic safety.

[0150] Example 2:

[0151] See Figure 4 As shown, an apparatus for detecting road debris is provided, which stores computer-readable instructions that, when read, can execute the aforementioned method for detecting road debris. The apparatus includes:

[0152] The environment module 601 automatically selects the infrared light wavelength based on the ambient light intensity and calculates the environmental compensation coefficient.

[0153] The dividing module 602 divides the road into preset units. In each preset unit, infrared emitting units are activated one by one at preset time intervals. The infrared emitting units emit multi-angle beams of specific infrared wavelengths to the infrared receiving units on the opposite side, and adjust the emission power of the infrared emitting units according to the environmental compensation coefficient. The infrared receiving units determine the receiving status of each beam.

[0154] Network module 603 constructs a fan-shaped infrared beam grid based on a unified coordinate system according to the beam reception status, and determines the location of the suspected target based on the beam reception status of each detection point in the fan-shaped infrared beam grid.

[0155] The determination module 604 determines whether the suspected target is a spilled object, and uses an area positioning model to obtain the center position and effective area of ​​the confirmed spilled object;

[0156] The alarm module 605 implements tiered alarms based on the center location and effective area of ​​the spilled material.

[0157] The embodiments of the present invention have been described above. However, the embodiments are not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make more equivalent embodiments under the guidance of the present embodiments, and all of them are within the protection scope of the present embodiments.

Claims

1. A method for detecting road surface spills, characterized in that, Includes the following steps: The infrared light wavelength is automatically selected based on the ambient light intensity, and the environmental compensation coefficient is calculated. The road is divided into preset units. In each preset unit, an infrared emitting unit is activated one by one at a preset time interval. The infrared emitting unit emits a multi-angle beam of selected infrared wavelength to the infrared receiving unit on the opposite side, and the emission power of the infrared emitting unit is adjusted according to the environmental compensation coefficient. The infrared receiving unit determines the receiving status of each beam. Based on the beam reception status, a fan-shaped infrared beam grid based on a unified coordinate system is constructed. The location of the suspected target is determined based on the beam reception status of each detection point within the fan-shaped infrared beam grid. The construction process of the fan-shaped infrared beam grid includes: Based on the physical layout of the infrared emitting unit and the infrared receiving unit, calculate the spatial trajectory of multiple beams emitted by each infrared emitting unit. The multi-angle beams emitted by different infrared emitting units intersect each other in the road space, and the point through which multiple beams intersect is called the detection point; A unified coordinate system is used to describe the fan-shaped infrared beam grid. The coordinates (x, y) of the detection point represent the absolute physical position in the road space. The x-coordinate value is the lateral distance of the detection point from the left boundary of the road, and the y-coordinate value is the longitudinal distance of the detection point from the starting end of the road. To determine whether a suspected target is a spilled object, an area localization model is used to obtain the center position and effective area of ​​the spilled object. A tiered alarm system is implemented based on the center location and effective area of ​​the spilled material.

2. The method for detecting road spills according to claim 1, characterized in that, The origin of the coordinate system is set on the left boundary of the starting end of the road. The x-axis is perpendicular to the road travel direction and points to the right boundary of the road. The coordinate range is from 0 to the road width. The y-axis is along the road travel direction and points to the road forward direction. The coordinate range is from 0 to the length of the detected road segment. For curved roads, the coordinate system is curved along the road centerline, keeping the x-axis always perpendicular to the tangent of the road centerline and the y-axis always along the tangent of the road centerline.

3. The method for detecting road surface spills according to claim 1, characterized in that, The detection points traversed by a beam whose reception status is blocked are considered suspected detection points. Multiple adjacent suspected detection points are combined through connectivity analysis to form the area where the suspected target is located.

4. The method for detecting road surface spills according to claim 1, characterized in that, Methods for determining whether a suspected target is a spilled object include: Determine whether the suspected target is obscured by a vehicle; If a suspected target is not obscured by a vehicle, determine whether it is due to environmental interference. For suspected targets that are not environmental disturbances, it is determined whether the duration of the obstruction is greater than the dynamic duration threshold. Suspected targets that are greater than the dynamic duration threshold are considered as debris. The dynamic duration threshold is the product of the preset base duration threshold and the environmental compensation coefficient.

5. The method for detecting road spills according to claim 4, characterized in that, The area localization model adopts a hybrid architecture combining multilayer perceptron and convolutional neural network. The coordinates of multiple suspected detection points within the spill at the current moment, the number of passing beams, and the number of blocked beams are input into the area localization model.

6. The method for detecting road spills according to claim 1, characterized in that, The tiered alarm mechanism includes Level 1 alarms, Level 2 alarms, and Level 3 alarms; Specifically, a Level 1 alarm is triggered when the risk score is greater than the Level 1 risk threshold but less than the Level 2 risk threshold; a Level 2 alarm is triggered when the risk score is greater than or equal to the Level 2 risk threshold but less than the Level 3 risk threshold; and a Level 3 alarm is triggered when the risk score is greater than or equal to the Level 3 risk threshold. The risk score is the product of the lane weight coefficient and the effective area. The lane weight coefficient is determined based on the lane determination result of the spilled material. If the spilled material is located in the main lane, the lane weight coefficient is the main lane weight; if the spilled material is located in the overtaking lane, the lane weight coefficient is the overtaking lane weight; if the spilled material is located in the emergency lane, the lane weight coefficient is the emergency lane weight; if the spilled material is located on the right shoulder, the lane weight coefficient is the right shoulder weight; if the spilled material is located on the left shoulder, the lane weight coefficient is the left shoulder weight. The lane determination result of the spilled material is determined based on the center position of the spilled material.

7. The method for detecting road spills according to claim 1, characterized in that, When the environmental compensation coefficient is less than the preset power threshold, the transmission power is the base power; when the environmental compensation coefficient is greater than the preset power threshold, the transmission power is the additional power, which is greater than the base power. The reception status includes pass and block. When the original signal strength received by the infrared receiving unit is greater than the preset signal strength threshold, the reception status is pass; when the original signal strength is less than or equal to the preset signal strength threshold, the reception status is block.

8. The method for detecting road spills according to claim 1, characterized in that, Under standard weather conditions, the standard signal strength received by each infrared receiving unit is recorded, and the ratio of the currently measured original signal strength to the standard signal strength is calculated to obtain the environmental compensation coefficient.

9. A device for detecting road spills, characterized in that, The apparatus is used to store computer-readable instructions that, when read, enable the execution of a method for detecting road debris as described in any one of claims 1-8, the apparatus comprising: The environment module automatically selects the infrared light wavelength based on the ambient light intensity and calculates the environmental compensation coefficient. The module divides the road into preset units. Within each preset unit, infrared emitting units are activated one by one at preset time intervals. The infrared emitting units emit multi-angle beams of specific infrared wavelengths to the infrared receiving units on the opposite side, and the emission power of the infrared emitting units is adjusted according to the environmental compensation coefficient. The infrared receiving units determine the reception status of each beam. The network module constructs a fan-shaped infrared beam grid based on a unified coordinate system according to the beam reception status. Based on the beam reception status of each detection point in the fan-shaped infrared beam grid, the location of the suspected target is determined. The determination module determines whether a suspected target is a spilled object, and uses an area localization model to obtain the center position and effective area of ​​the spilled object. The alarm module implements tiered alarms based on the center location and effective area of ​​the spilled material.

Images