A chemical product leakage early warning system based on a multi-modal fusion algorithm

Abstract

The application provides a chemical product leakage early warning system based on a multi-modal fusion algorithm, comprising an image monitoring module, an air pressure monitoring module, a test paper monitoring module, a temperature monitoring module, a vehicle driving monitoring module, and a leakage risk estimation module; the leakage risk estimation module estimates the leakage risk level based on the feedback data of the image monitoring module, the air pressure monitoring module, the test paper monitoring module, the temperature monitoring module, and the vehicle driving monitoring module. The multi-dimensional fusion based on the system can accurately and quickly find the occurrence of chemical product leakage.

Description

TECHNICAL FIELD

[0001] The present application relates to a safety device for transporting or loading special goods or objects, in particular to a chemical product leakage early warning system based on a multi-modal fusion algorithm. BACKGROUND

[0002] Chemical product leakage is a serious accident that may occur during storage and transportation, which has multidimensional, long-term and diffusive hazards, involving human health, ecological environment, social safety and other aspects. For example: after the leakage of chemical products (such as toxic gases, corrosive liquids, flammable and explosive substances, etc.), it can invade the human body through respiratory, skin contact, and accidental ingestion, causing immediate or long-term damage. Some toxic gases can cause respiratory tract burns, suffocation in a short time, and fatal in high concentration within a few minutes. Contact with corrosive substances (such as sulfuric acid and caustic soda) on the skin or mucous membranes can cause chemical burns, and if it enters the eyes, it may cause blindness; some chemical products (such as benzene, formaldehyde, and heavy metals) have cumulative toxicity, and long-term low-dose exposure may cause cancer (such as benzene causing leukemia), liver and kidney failure, and nervous system damage (such as lead poisoning), and the symptoms may be latent for several years. In addition, if the leaked chemical products (such as organic solvents, heavy metals, and pesticides) enter rivers, lakes, or groundwater, they can destroy water quality and cause fish and aquatic plants to die. Corrosive or toxic substances can penetrate into the soil, killing soil microorganisms, destroying soil structure, and causing land desertification and crop failure. Volatile chemicals (such as pesticides) can volatilize into the atmosphere and form "acid rain" through rainfall, further damaging vegetation and forests. The animals and plants in the leakage area may die or migrate due to poisoning and habitat destruction, leading to the collapse of the local ecosystem. In addition, many chemical products (such as gasoline, ethanol, and propane) are flammable and explosive, and if they come into contact with a fire source (open flame, static electricity, high temperature), they can cause fires and explosions, expanding the scope of harm. The shock wave and high temperature generated by the explosion can destroy surrounding buildings and cause casualties; the fire may also release more toxic smoke (such as carbon monoxide and phosgene), forming a "toxic fire complex disaster". Moreover, the leakage event may cause the surrounding residents to evacuate urgently, schools to close, and businesses to shut down, leading to panic buying of supplies, rumor spreading, and other social panic, increasing the difficulty of emergency management.

[0003] Therefore, in general, if the chemical product leakage cannot be stopped immediately, the resulting personnel and economic losses are incalculable.

[0004] Therefore, how to effectively assess the leakage risk of chemical products inside the truck and how to quickly terminate and stop the leakage are problems that have been faced and need to be solved in the field. SUMMARY

[0005] The present application aims to provide a comprehensive and effective method for chemical product leakage risk during chemical product truck transportation, specifically as follows:

[0006] The present application provides a chemical product leakage early warning system based on a multi-modal fusion algorithm, which comprises an image monitoring module, a gas pressure monitoring module, a test paper monitoring module, a temperature monitoring module, a vehicle driving monitoring module, and a leakage risk estimation module.

[0007] The image monitoring module monitors in real time whether the chemical product bottles are tilted, leaked, or leaked by monitoring instruments installed inside the vehicle.

[0008] The gas pressure monitoring module monitors in real time whether the internal environment of the truck carrying the chemical product bottles changes in pressure by a pressure sensor installed inside the vehicle.

[0009] The test paper monitoring module monitors in real time whether the chemicals are volatilized and the volatilization level due to leakage by chemical test paper installed inside the vehicle.

[0010] The temperature monitoring module monitors in real time the temperature anomaly of the vehicle internal environment by temperature sensors installed inside and outside the vehicle.

[0011] The vehicle driving monitoring module monitors in real time the frequency and amplitude of the vehicle's jolt by a displacement sensor installed on the vehicle.

[0012] The leakage risk estimation module estimates the leakage risk level based on the feedback data of the image monitoring module, the gas pressure monitoring module, the test paper monitoring module, the temperature monitoring module, and the vehicle driving monitoring module.

[0013] Further, the present application provides a chemical product leakage early warning system based on a multi-modal fusion algorithm,

[0014] The monitoring instruments are multiple and installed on the top surface of the vehicle, and are arranged in multiple matrices based on the size of the vehicle body.

[0015] The image monitoring module splices the images returned by multiple monitoring instruments to eliminate image distortion caused by the monitoring instruments themselves and form a top-down image with near-planar visual effect.

[0016] Further, the present application provides a chemical product leakage early warning system based on a multi-modal fusion algorithm,

[0017] The upper surface of the chemical product bottle has an identifier.

[0018] The image monitoring module identifies the number and shape change of the identifier in the image data to determine whether the chemical product bottle is tilted or tilted.

[0019] The identifier is a pattern arranged transversely on the bottle cap of the chemical product packaging bottle;

[0020] The specific identification method is:

[0021] SY1. Call the image of each frame or specified interval frame, and extract all identifiable identifiers;

[0022] SY2. Determine whether the number of identifiers is consistent with loading,

[0023] When "consistent", SY3 is performed;

[0024] When "inconsistent", it is determined that the chemical product packaging bottle is poured, and the pouring result is returned to the front end of the device;

[0025] SY3. Determine whether the identifier is deformed,

[0026] When "no deformation occurs", return to SY1;

[0027] When "deformation occurs", SY4 is performed;

[0028] SY4. Determine the level of chemical product packaging bottle tilt based on the pattern deformation model, and return the tilt result to the front end of the device.

[0029] Further, the present application provides a chemical product leakage early warning system based on a multi-modal fusion algorithm,

[0030] The above image monitoring module also identifies the leakage trace in the image data to determine whether the chemical product packaging bottle leaks;

[0031] The specific identification method is:

[0032] SX1. Call the image of each frame or specified interval frame, and determine whether there is suspicious trace based on the trace model;

[0033] When the result is "no", the next image is continuously determined;

[0034] When the result is "yes", SX2 is performed;

[0035] SX2. Compare the suspicious trace with the trace of the previous image to determine whether it is a new trace;

[0036] When the result is "no", the next image is continuously determined;

[0037] When the result is "yes", SX3 is performed, and the device front end is warned suspiciously;

[0038] SX3. For the suspicious trace, starting from the first image where the suspicious trace is found, iterate through the next N images to determine whether the suspicious trace has expansion signs;

[0039] When the result is "no", the suspicious continues to be monitored;

[0040] When the result is "yes", a leakage alarm is returned to the front end of the device.

[0041] Further, the application provides a chemical product leakage early warning system based on a multi-modal fusion algorithm,

[0042] The pressure sensors are multiple, and are installed on the top surface and the side surface of the vehicle;

[0043] The air pressure monitoring module monitors the return data of the pressure sensor, and judges whether the chemical product is gasified or volatilized beyond a reasonable level;

[0044] The specific identification method is: based on a chemical product air pressure model, whether the current air pressure exceeds a safety value is judged.

[0045] Further, the application provides a chemical product leakage early warning system based on a multi-modal fusion algorithm,

[0046] The construction method of the chemical product air pressure model is as follows:

[0047] SS1. Collect the pressure conditions of different chemical products under different V1 / V2 and different temperature conditions;

[0048] SS2. Complete the actual data by computer learning, and form a chemical product air pressure model.

[0049] Further, the application provides a chemical product leakage early warning system based on a multi-modal fusion algorithm,

[0050] The vehicle interior is provided with multiple test paper installation boxes, and the chemical test paper is detachably arranged in the test paper installation box.

[0051] Further, the application provides a chemical product leakage early warning system based on a multi-modal fusion algorithm,

[0052] The specific method for the temperature monitoring module to monitor the temperature abnormity of the vehicle interior environment in real time is:

[0053] ST1. According to a vehicle temperature model, whether the current outdoor temperature based on the vehicle temperature is within a reasonable range is judged;

[0054] When the result is "yes", the monitoring is continuously maintained;

[0055] When the result is "no", ST2 is performed;

[0056] ST2. Whether the current indoor temperature is higher or lower than the reasonable range is judged;

[0057] When the result is "higher", ST3.1 is performed;

[0058] When the result is "lower", ST3.2 is performed;

[0059] ST3.1. Determine whether the chemical product currently carried by the vehicle will cause the ambient temperature to rise when exposed to air;

[0060] When the result is "yes", the result of possible leakage is returned to the front end of the device;

[0061] When the result is "no", the vehicle temperature anomaly is reported to the front end of the device;

[0062] ST3.2. Determine whether the chemical product currently carried by the vehicle will cause the ambient temperature to decrease when exposed to air;

[0063] When the result is "yes", the result of possible leakage is returned to the front end of the device;

[0064] When the result is "no", the vehicle temperature anomaly is reported to the front end of the device.

[0065] Further, the present application provides a chemical product leakage early warning system based on a multi-modal fusion algorithm,

[0066] The vehicle driving monitoring module further judges the road condition level of the front road according to the AI road condition data.

[0067] Further, the present application provides a chemical product leakage early warning system based on a multi-modal fusion algorithm,

[0068] The leakage risk estimation module scores the return data of the image monitoring module, the air pressure monitoring module, the test paper monitoring module, the temperature monitoring module, and the vehicle driving monitoring module;

[0069] The leakage risk coefficient is calculated according to the following formula:

[0070] R=k1×MI+k2×AP+k3×TP+k4×TM+k5×BP;

[0071] Wherein, k1, k2, k3, k4, k5 are weight coefficients;

[0072] MI is the score based on the return data of the image monitoring module;

[0073] AP is the score based on the return data of the air pressure monitoring module;

[0074] TP is the score based on the return data of the test paper monitoring module;

[0075] TM is the score based on the return data of the temperature monitoring module;

[0076] BP is a score based on backhaul data of the vehicle driving monitoring module;

[0077] R is rated based on the leakage risk level. BRIEF DESCRIPTION OF DRAWINGS

[0078] Figure 1 The embodiment provides a running flowchart of a chemical product leakage early warning system based on a multi-modal fusion algorithm. DETAILED DESCRIPTION

[0079] The present application can be implemented in various modifications and can have various embodiments, so that specific embodiments are illustrated in the drawings and described. However, this is not intended to limit the present application to a specific embodiment, but should be understood to include all modifications, equivalents and even alternatives falling within the spirit and technical scope of the present application.

[0080] As shown in Figure 1 The embodiment provides a chemical product leakage early warning system based on a multi-modal fusion algorithm, which comprises an image monitoring module, an air pressure monitoring module, a test paper monitoring module, a temperature monitoring module, a vehicle driving monitoring module, and a leakage risk estimation module.

[0081] The image monitoring module monitors whether the chemical product packaging bottle is tilted, leaks or leaks in real time through a monitoring instrument installed in the vehicle interior.

[0082] Specifically, the monitoring instrument is multiple, respectively installed on the top surface of the vehicle, based on the size of the vehicle body, a plurality of matrix settings are performed, and the multiple monitoring instruments are set to achieve a real restoration of the overhead state view of the truck interior environment, rather than a distorted visual result of near large and far small formed after shooting based on one angle.

[0083] The image monitoring module uses libraries such as OpenCV and JavaCV to first use related functions to collect video frames of images returned by the monitoring instrument, then performs geometric correction, feature extraction and matching on the video frames, determines the overlapping area between images, and finally splices the matched images into a panoramic picture and displays, thereby eliminating the image distortion caused by the monitoring instrument itself, forming an overhead image with near-plane visual effect.

[0084] The upper surface of the chemical product packaging bottle has an identifier, which is a pattern such as a horizontal line arranged on the cap of the chemical product packaging bottle.

[0085] The image monitoring module judges whether the chemical product bottled product is tilted or poured by recognizing the number and shape change of the identifier in the image data; that is, when the bottle is poured, the identifier of the bottle cannot be captured by the camera on the top, so it is judged as pouring; when the bottle is tilted, the identifier of the bottle will be deformed due to tilting, for example, a character will be shortened, etc.

[0086] The specific identification method is:

[0087] SY1. Call the image of each frame or specified interval frame, and extract all identifiable identifiers;

[0088] SY2. Determine whether the number of identifiers is consistent with loading,

[0089] When "consistent", SY3 is performed;

[0090] When "inconsistent", it is judged that the chemical product bottled product is poured, and after the pouring result is returned to the front end of the device, SY3 is performed;

[0091] SY3. Determine whether the identifier is deformed,

[0092] When "no deformation occurs", return to SY1;

[0093] When "deformation occurs", SY4 is performed;

[0094] SY4. Determine the level of chemical product bottled product tilt based on the graphical deformation model, and return the tilt result to the front end of the device.

[0095] The construction method of the graphical deformation module here is to collect the top view of the figure at different tilt angles, measure the figure shrinkage ratio at different tilt angles in the figure, and draw a linear graph. Or directly use L=line×cosɑ, line is the length of a character, and ɑ is the included angle to draw a linear graph of L / line and ɑ. Thus, when measuring the collected video, the current tilt angle can be obtained through the linear graph.

[0096] In addition, the image monitoring module also identifies the leakage traces in the image data to judge whether the leakage of the chemical product caused by severe jolting, bottle collision / aging or inherent quality problems occurs;

[0097] The specific identification method is:

[0098] SX1. Call the image of each frame or specified interval frame, and determine whether there are suspicious traces based on the trace model;

[0099] When the result is "no", continue to judge the next image;

[0100] When the result is "yes", SX2 is performed;

[0101] The construction and use method of the trace model is to collect images of various bottle body leaks, and determine whether there is a leakage phenomenon in the current image through machine recognition;

[0102] SX2. Compare the suspicious trace with the trace of the previous image to determine whether it is a new trace; the purpose of this step is to exclude the prints that cannot be cleaned in the original vehicle body and avoid misjudgment;

[0103] When the result is "no", the next image is continuously judged;

[0104] When the result is "yes", SX3 is performed, and a suspicious warning is given to the front end of the device;

[0105] SX3. For the suspicious trace, starting from the first image where the suspicious trace is found, iterate through the next N images to determine whether the suspicious trace has expansion signs, that is, after locking the suspicious trace, extract the trace based on the traditional image processing method and calculate the area of the trace;

[0106] When the result is "no", mark the suspicious and continue monitoring;

[0107] When the result is "yes", return a leakage warning to the front end of the device.

[0108] The air pressure monitoring module monitors whether the internal environment of the truck carrying the chemical product tank bottle changes in pressure in real time through the pressure sensor installed in the vehicle interior;

[0109] The pressure sensor is multiple and installed on the top and side of the vehicle;

[0110] The air pressure monitoring module monitors the return data of the pressure sensor to determine whether the chemical product gasifies or volatilizes beyond a reasonable level; it is worth noting that the returned data is cleaned, irrelevant or caused by the instrument change due to vehicle bumping. The specific cleaning method is: eliminate sudden disordered data in a stable state, that is, sudden increase or sudden decrease. Because the increase or decrease of internal pressure cannot be achieved at one time, there is a linear change process.

[0111] The specific identification method for valid data is: based on the chemical product air pressure model, determine whether the current air pressure exceeds the safety value.

[0112] Here, the construction method of the chemical product air pressure model is to collect the pressure conditions of different chemical products under different warehouse temperature conditions and different exudation amounts (i.e., V1 / V2, V1 is the exudation amount, and V2 is the warehouse volume), complete the actual data through computer learning, form the chemical product air pressure model, and determine the safe limit pressure of the exudation amount.

[0113] And since the pressure sensors are set at different positions of the vehicle body, the main leakage position can be determined by the first pressure sensor that changes, which is particularly suitable for the compartment storage state.

[0114] The test paper monitoring module monitors in real time whether the chemicals are volatilized due to leakage and the volatilization level through the chemical test paper installed in the vehicle interior; the chemical test paper is detachably arranged in the test paper installation box, and the chemical test paper can be various commercially available general chemical test papers for detecting chemicals such as pH test paper, starch-potassium iodide test paper, acidic potassium permanganate test paper / magenta test paper, ion-specific test paper, etc.

[0115] The temperature monitoring module monitors in real time the temperature abnormality of the vehicle interior environment through the temperature sensor installed in and outside the vehicle.

[0116] The specific method of the temperature monitoring module for monitoring in real time the temperature abnormality of the vehicle interior environment is as follows:

[0117] ST1. Determine whether the vehicle interior temperature is within a reasonable range based on the current vehicle exterior temperature according to the vehicle temperature model; the vehicle temperature model is to collect the interior and exterior temperatures under different exterior temperatures of the freight vehicle and form a model.

[0118] When the result is "yes", continue to monitor;

[0119] When the result is "no", perform ST2.

[0120] ST2. Determine whether the current interior temperature is higher or lower than the reasonable range.

[0121] When the result is "higher", perform ST3.1.

[0122] When the result is "lower", perform ST3.2.

[0123] ST3.1. Determine whether the chemical product currently carried by the vehicle will cause the ambient temperature to rise when exposed to air.

[0124] When the result is "yes", return the possible leakage result to the front end of the device.

[0125] When the result is "no", report the vehicle interior temperature abnormality to the front end of the device, that is, these abnormalities may be caused by the failure of the vehicle sealing or thermal insulation layer.

[0126] ST3.2. Determine whether the chemical product currently carried by the vehicle will cause the ambient temperature to decrease when exposed to air.

[0127] When the result is "yes", return the possible leakage result to the front end of the device.

[0128] When the result is "no", the vehicle inner temperature abnormality reporting device front end, that is, these abnormalities may be caused by the failure of the vehicle sealing or thermal insulation layer.

[0129] The vehicle driving monitoring module monitors the frequency and amplitude of the vehicle's jolt in real time through the displacement sensor installed on the vehicle. In addition, the module also judges the road condition level of the front road according to the AI road condition data, that is, whether there is road maintenance that will cause a lot of jolt on the front road, whether there is an accident that needs to frequently change lanes or turn around to bypass, etc.

[0130] The leakage risk estimation module estimates the leakage risk level based on the feedback data of the image monitoring module, the air pressure monitoring module, the test paper monitoring module, the temperature monitoring module, and the vehicle driving monitoring module.

[0131] The specific method is that the leakage risk estimation module scores the feedback data of the image monitoring module, the air pressure monitoring module, the test paper monitoring module, the temperature monitoring module, and the vehicle driving monitoring module.

[0132] The leakage risk coefficient R is calculated according to the following formula:

[0133] R=k1×MI+k2×AP+k3×TP+k4×TM+k5×BP;

[0134] Wherein, k1, k2, k3, k4, k5 are weight coefficients;

[0135] MI is the score based on the feedback data of the image monitoring module;

[0136] AP is the score based on the feedback data of the air pressure monitoring module;

[0137] TP is the score based on the feedback data of the test paper monitoring module;

[0138] TM is the score based on the feedback data of the temperature monitoring module;

[0139] BP is the score based on the feedback data of the vehicle driving monitoring module;

[0140] The leakage risk level is rated based on R.

[0141] Although the above has been described centering on the embodiments, this is only an example and does not limit the present application, and those skilled in the art will clearly understand that various modifications and applications not exemplified above can be made within the scope of the essential characteristics of the embodiments. For example, each of the constituent elements specifically shown in the embodiments can be modified and implemented. Moreover, various differences related to such modifications and applications should be interpreted as being included in the scope of the present application defined in the appended claims.

Claims

1. A chemical leakage early warning system based on a multi-modal fusion algorithm, characterized in that: The monitoring system comprises an image monitoring module, a gas pressure monitoring module, a test paper monitoring module, a temperature monitoring module, a vehicle driving monitoring module, and a leakage risk estimation module. The image monitoring module monitors in real time whether the chemical product bottles are tilted, seep or leaked through monitoring instruments installed in the vehicle. The gas pressure monitoring module monitors in real time whether the internal environment of the vehicle carrying the chemical product bottles is changed in pressure through pressure sensors installed in the vehicle. The test paper monitoring module monitors in real time whether the chemical products are volatilized due to leakage and the volatilization level through chemical test papers installed in the vehicle. The temperature monitoring module monitors in real time the temperature anomaly of the internal environment of the vehicle through temperature sensors installed in and outside the vehicle. The vehicle driving monitoring module monitors in real time the frequency and amplitude of the vehicle bumping through displacement sensors installed on the vehicle. The leakage risk estimation module estimates the leakage risk level based on the feedback data of the image monitoring module, the gas pressure monitoring module, the test paper monitoring module, the temperature monitoring module, and the vehicle driving monitoring module. The monitoring instruments are multiple and are installed on the top surface of the vehicle and are arranged in multiple matrices based on the size of the vehicle body. The image monitoring module splices the images fed back by the multiple monitoring instruments, eliminates the image distortion caused by the monitoring instruments themselves, and forms a top-down image with a near-planar visual effect. The upper surface of the chemical product bottle has an identifier. The image monitoring module judges whether the chemical product bottle is tilted or dumped by recognizing the number and shape change of the identifier in the image data. The identifier is a pattern arranged horizontally on the cap of the chemical product bottle. The specific recognition method is as follows: SY1. Call the image of each frame or specified interval frame, and extract all identifiable identifiers. SY2. Judge whether the number of identifiers is consistent with the loading. When the result is "consistent", SY3 is performed. When the result is "inconsistent", it is judged that the chemical product bottle is dumped, and after the dumping result is returned to the front end of the device, SY3 is performed. SY3. Judge whether the identifier is deformed. When the result is "not deformed", return to SY1. When the result is "deformed", perform SY4. SY4. Judge the level of the chemical product bottle tilt based on the pattern deformation model, and return the tilt result to the front end of the device. The image monitoring module also recognizes the leakage traces in the image data to judge whether the chemical product bottle is leaked. The specific recognition method is as follows: SX1. Call the image of each frame or specified interval frame, and judge whether there is suspicious trace based on the trace model. When the result is "no", continue to judge the next image. When the result is "yes", perform SX2. SX2. Compare the suspicious trace with the trace of the previous image to judge whether it is a new trace. When the result is "no", continue to judge the next image. When the result is "yes", perform SX3 and alert the suspicious to the front end of the device. SX3. For the suspicious trace, starting from the first image where the suspicious trace is found, iterate through the next N images to judge whether the suspicious trace has expansion signs. When the result is "no", mark the suspicious for continuous monitoring. When the result is "yes", a leakage alarm is returned to the front end of the device.

2. The chemical product leakage early warning system based on a multi-modal fusion algorithm according to claim 1, characterized in that: The pressure sensors are multiple and installed on the top and side surfaces of the vehicle. The air pressure monitoring module monitors the return data of the pressure sensors to determine whether the chemical product is gasified or volatilized beyond a reasonable level. The specific identification method is to determine whether the current air pressure exceeds the safety value based on the chemical product air pressure model.

3. The chemical product leakage early warning system based on a multi-modal fusion algorithm according to claim 2, characterized in that: The construction method of the chemical product air pressure model is as follows: SS1. Collect the pressure conditions of different chemical products under different V1 / V2, V1 being the leakage amount, V2 being the warehouse volume, and different temperature conditions. SS2. Complete the actual data through computer learning to form the chemical product air pressure model.

4. The chemical product leakage early warning system based on a multi-modal fusion algorithm according to claim 1, characterized in that: The vehicle interior is provided with multiple test paper installation boxes, and the chemical test paper is detachably arranged in the test paper installation box.

5. The chemical product leakage early warning system based on a multi-modal fusion algorithm according to claim 1, characterized in that: The specific method for the temperature monitoring module to monitor the temperature anomaly of the vehicle interior environment is as follows: ST1. According to the vehicle temperature model, determine whether the current vehicle interior temperature is within a reasonable range based on the current vehicle exterior temperature. When the result is "yes", continue to monitor. When the result is "no", perform ST2. ST2. Determine whether the current vehicle interior temperature is higher or lower than the reasonable range. When the result is "higher", perform ST3.

1. When the result is "lower", perform ST3.

2. ST3.

1. Determine whether the chemical product currently carried by the vehicle will cause the environmental temperature to rise when exposed to air. When the result is "yes", return the possible leakage result to the front end of the device. When the result is "no", report the vehicle interior temperature anomaly to the front end of the device. ST3.

2. Determine whether the chemical product currently carried by the vehicle will cause the environmental temperature to decrease when exposed to air. When the result is "yes", return the possible leakage result to the front end of the device. When the result is "no", report the vehicle interior temperature anomaly to the front end of the device.

6. The chemical product leakage early warning system based on a multi-modal fusion algorithm according to claim 1, characterized in that: The vehicle driving monitoring module also determines the road condition level of the front road according to AI road condition data.

7. The chemical product leakage early warning system based on a multi-modal fusion algorithm according to claim 1, characterized in that: The leakage risk estimation module scores the return data of the image monitoring module, the air pressure monitoring module, the test paper monitoring module, the temperature monitoring module, and the vehicle driving monitoring module. The leakage risk coefficient is calculated according to the following formula: R=k1×MI+k2×AP+k3×TP+k4×TM+k5×BP; wherein k1, k2, k3, k4, and k5 are weight coefficients; MI is the score based on the return data of the image monitoring module. AP is the score based on the return data of the air pressure monitoring module; TP is the score based on the return data of the test paper monitoring module; TM is the score based on the return data of the temperature monitoring module; BP is the score based on the return data of the vehicle driving monitoring module; R is rated based on the leakage risk level.

Images