Visual feature-based emergency detection method for liquid leakage in semiconductor liquid supply cabinet

Abstract

The application discloses a visual feature-based emergency detection method for liquid leakage in a semiconductor liquid supply cabinet, and particularly relates to the technical field of safety monitoring of semiconductor production equipment, and comprises the following steps: obtaining internal structure information of the semiconductor liquid supply cabinet, and establishing a plurality of visual detection areas according to the pipelines, pipeline joints, pump bodies, valves, filters, liquid storage containers, cabinet bottom plates and liquid discharge areas in the liquid supply cabinet; and collecting detection images of each visual detection area by an image acquisition device arranged in the liquid supply cabinet. The application realizes early detection of small leakage, dripping and liquid accumulation in the liquid supply cabinet by comprehensively analyzing visual features such as liquid drops, liquid films, liquid flows, reflective boundaries and motion trajectories; in combination with the running state and multi-level risk determination, false alarms caused by reflection, condensate water and maintenance operations are reduced, and the pump can be stopped, the valve can be closed, liquid can be discharged, alarms can be given and equipment interlocking can be performed, so that the leakage disposal time is shortened, and the safety of the liquid supply system and the accident tracing capability are improved.

Description

Technical Field

[0001] This invention relates to the field of safety monitoring technology for semiconductor manufacturing equipment, and more specifically, to an emergency detection method for leakage in semiconductor liquid supply cabinets based on visual features. Background Technology

[0002] In semiconductor manufacturing, it is common to supply acids, alkalis, organic solvents, specialty chemical liquids, ultrapure water, coolants, or other process media to process equipment via a liquid supply cabinet. The liquid supply cabinet contains liquid storage containers, supply pumps, valves, filters, pressure regulating components, pipe joints, and drainage components. There are many pipes and joints, and some of the liquids are corrosive, volatile, conductive, or flammable.

[0003] Existing methods for detecting leaks in liquid supply cabinets mainly include level detection, pressure detection, flow detection, conductive leak sensors, float-type level switches, and point-type leak detectors located at the bottom of the cabinet. These methods have the following shortcomings:

[0004] 1. Point-type leak detectors can only detect liquids near the sensor installation location. When liquids flow along the cabinet bottom plate, pipe surface, or equipment support, there are problems with detection lag or missed detection.

[0005] 2. The detection sensitivity of conductive sensors is insufficient for ultrapure water, some organic solvents, and liquids with low conductivity.

[0006] 3. Liquid level, pressure and flow rate detection can usually only reflect the overall status of the liquid supply system, and it is difficult to determine the specific location of the leak, as well as the status of minor leaks, liquid films, hanging liquid and dripping.

[0007] 4. The interior of the liquid supply cabinet has metal surfaces, transparent pipes, reflective fittings, and complex shadows, which can easily cause image interference due to highlights, reflections, fog, condensation, and equipment vibration.

[0008] 5. Existing visual inspection methods typically use single-frame brightness thresholds or simple background subtraction, which are difficult to adapt to different liquid colors, different cabinet lighting conditions, and changes in cabinet door opening and closing, maintenance operations, and equipment operating status.

[0009] 6. The existing detection system lacks effective linkage with the shutdown of the liquid supply pump, the closure of the liquid supply valve, the start of the exhaust equipment, the interlocking of process equipment, and remote alarms, making it difficult to meet the emergency response requirements for liquid leakage incidents.

[0010] Therefore, there is an urgent need for an emergency detection method for leakage in semiconductor liquid supply cabinets based on visual characteristics to solve the above problems. Summary of the Invention

[0011] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide an emergency detection method for leakage in a semiconductor liquid supply cabinet based on visual features, thereby solving the problems mentioned in the background art.

[0012] To achieve the above objectives, the present invention provides the following technical solution: an emergency detection method for leakage in a semiconductor liquid supply cabinet based on visual features, comprising the following steps:

[0013] S1. Obtain the internal structure information of the semiconductor liquid supply cabinet, and establish multiple visual inspection areas based on the internal pipes, pipe joints, pumps, valves, filters, liquid storage containers, cabinet bottom plate and drainage area of ​​the liquid supply cabinet;

[0014] S2. The image acquisition device installed inside the liquid supply cabinet acquires detection images of each visual detection area and simultaneously obtains at least one of the following operating status information: liquid supply pump status, liquid supply valve status, pipeline pressure, pipeline flow rate, liquid level of storage container, cabinet door status, and maintenance status. The image acquisition device includes at least one of the following: industrial camera, low-light camera, infrared camera, polarization camera, and multispectral camera. The image acquisition device acquires images using a fixed field of view, multiple field of view, or switchable field of view mode.

[0015] S3. Select a background image corresponding to the current operating state based on the operating state information, and perform distortion correction, brightness normalization, noise suppression, reflective marking and image registration on the detected image to obtain a preprocessed image;

[0016] S4. Compare the preprocessed image with the corresponding background image, and extract liquid candidate regions by combining liquid brightness features, color features, edge features, texture features and motion features between consecutive image frames;

[0017] S5. Perform connected component analysis and morphological processing on the liquid candidate region, and extract the area, shape, boundary, texture, brightness, motion, duration, area growth rate and spatial location features of the liquid target;

[0018] S6. Calculate the leakage risk value based on the visual characteristics, temporal continuity characteristics, and spatial correlation between the liquid target and high-risk components inside the liquid supply cabinet, and determine whether the liquid target is a suspected leakage target.

[0019] S7. Determine the leakage level based on the leakage risk value, the area growth rate of the liquid target, the dripping motion characteristics, the duration, and the consistency with other operating status information;

[0020] S8. Based on the leakage level, output a graded alarm signal or emergency linkage control signal to at least one of the following: liquid supply pump, liquid supply valve, liquid discharge valve, ventilation equipment, audible and visual alarm device, process equipment control system and remote monitoring platform.

[0021] S9. After the emergency response is executed, continue to collect the detection images of the visual detection area, determine the leakage handling results based on the changes in the area, diffusion and duration of the liquid target, and save the leakage event record.

[0022] In a preferred embodiment, the liquid supply cabinet is equipped with at least one of a visible light source, a near-infrared light source, a structured light source, or a pulsed light source. The image acquisition device acquires at least two sets of images under different lighting conditions according to a preset light source sequence, and determines the liquid reflection characteristics based on the differences between the images under different lighting conditions.

[0023] In a preferred embodiment, the background image includes a reference image of the liquid supply cabinet in at least one of the following states: normal standby state, normal liquid supply state, liquid replacement state, cleaning state, and maintenance state, and the corresponding background image is selected according to the current operating state of the liquid supply cabinet.

[0024] In a preferred embodiment, the liquid candidate region is extracted through at least two of the following methods:

[0025] The detected image is compared with the background image;

[0026] Candidate regions are extracted based on local brightness and color changes in the detected image;

[0027] Edge extraction is performed based on droplet boundaries, liquid film boundaries, or liquid flow boundaries;

[0028] Extracting motion regions based on displacement changes between consecutive image frames;

[0029] Liquid regions are extracted based on the differences in response under different polarization states or different spectral bands.

[0030] In a preferred embodiment, the morphological characteristics of the liquid target include at least one of area, perimeter, aspect ratio, roundness, convexity, slenderness ratio, boundary curvature, and boundary closure degree, and the motion characteristics of the liquid target include at least one of displacement direction, motion velocity, motion acceleration, dripping period, and motion trajectory continuity.

[0031] In a preferred embodiment, target tracking is performed on liquid candidate regions within multiple consecutive time windows. When the liquid candidate regions meet at least one of the following conditions: the liquid candidate regions remain present in multiple consecutive image frames, the area of ​​the liquid candidate regions continues to increase, the liquid candidate regions spread along the direction of gravity, the center position of the liquid candidate regions shifts along the direction of gravity, the liquid candidate regions form a continuous dripping trajectory below the pipe joint, and the liquid candidate regions are connected to adjacent liquid candidate regions, the leakage risk value is increased.

[0032] In a preferred embodiment, when calculating the leakage risk value, an interference correction parameter is determined based on at least one of the following: high-gloss area, metal reflection area, transparent component reflection area, fog area, condensate area, cabinet door opening / closing status, and personnel maintenance status. The interference correction parameter is then used to reduce the leakage risk value of the corresponding liquid candidate area. The leakage level includes at least Level 1 anomaly, Level 2 suspected leakage, Level 3 confirmed leakage, and Level 4 severe leakage. Level 1 anomaly corresponds to a low-confidence liquid candidate area in a single frame or for a short duration. Level 2 suspected leakage corresponds to a persistent liquid target or a droplet target with a continuous movement trajectory. Level 3 confirmed leakage corresponds to a liquid target area continuously increasing, continuous dripping, liquid film diffusion, or a liquid target located at a high-risk pipeline connection. Level 4 severe leakage corresponds to rapid diffusion of the liquid target, large-area liquid accumulation, continuous leakage, leakage in multiple visual detection areas simultaneously, or leakage visual characteristics occurring simultaneously with at least one of pressure anomaly, flow anomaly, and liquid level anomaly.

[0033] In a preferred embodiment, for a liquid supply cabinet in maintenance, liquid replacement, or cleaning status, the leakage risk weight of the corresponding visual inspection area is reduced, and a normal background image is re-established or updated after the maintenance status is completed. When a Level 3 confirmed leakage is detected, at least one emergency linkage action is performed, including shutting down the liquid supply pump, reducing the operating power of the liquid supply pump, closing the liquid supply valve corresponding to the leakage area, isolating the pipeline corresponding to the leakage area, activating the audible and visual alarm device, sending alarm information to the remote monitoring platform, and sending an interlock signal to the process equipment control system. When a Level 4 severe leakage is detected, at least one action is performed, including stopping the liquid supply pump, closing the liquid supply valve, isolating the relevant pipeline, opening the drain valve, starting the exhaust equipment, interlocking the process equipment, and remote emergency alarm.

[0034] In a preferred embodiment, after emergency response is initiated, the area, area growth rate, diffusion rate, dripping frequency, and duration of the liquid target are continuously monitored. When the leakage risk value is lower than a preset clearance threshold within multiple consecutive time windows, a leakage handling completion status is generated. When the leakage risk value does not decrease or continues to increase, the current alarm level is maintained or the alarm level is upgraded. The leakage event record includes at least the leakage occurrence time, leakage visual detection area, leakage image sequence, liquid target location, liquid target area, liquid target movement trajectory, leakage risk value, leakage level, operating status, alarm signal, emergency response action, and response execution result. The leakage risk value is dynamically corrected based on at least one of the following: the hazard level of the liquid being transported in the supply cabinet, liquid type, pipeline pressure, pipeline flow rate, risk level of the visual detection area, and operating status of the process equipment.

[0035] The technical effects and advantages of this invention are as follows:

[0036] This invention utilizes visual features such as droplets, liquid films, liquid flows, reflective boundaries, diffusion boundaries, motion trajectories, and temporal changes to enable early identification of minute leaks, continuous drips, liquid film diffusion, liquid accumulation, and flow leaks within semiconductor liquid supply cabinets. It is not entirely dependent on liquid conductivity and exhibits good adaptability to ultrapure water, low-conductivity liquids, and some organic solvents. By establishing background models under different operating conditions and identifying and correcting interference from metal reflections, transparent component reflections, fog, condensate, and manual maintenance, false alarms can be reduced. By dividing the visual detection area and analyzing the spatial relationship between liquid targets and pipe joints, valves, pumps, and other high-risk components, the source of leakage can be located. Furthermore, based on information such as leakage area, growth rate, drip frequency, duration, pressure, flow rate, and liquid level, graded judgments and emergency response can be implemented, shortening the time for leak detection and handling, reducing the risk of chemical liquid diffusion, equipment damage, and process contamination. Finally, by maintaining complete event records, it provides a basis for equipment maintenance, fault analysis, and safety traceability. Detailed Implementation

[0037] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0038] This invention provides an emergency detection method for leakage in a semiconductor liquid supply cabinet based on visual features, comprising the following steps:

[0039] S1. Obtain the internal structure information of the semiconductor liquid supply cabinet, and establish multiple visual inspection areas based on the internal pipes, pipe joints, pumps, valves, filters, liquid storage containers, cabinet bottom plate and drainage area of ​​the liquid supply cabinet;

[0040] S2. The image acquisition device installed inside the liquid supply cabinet acquires detection images of each visual detection area and simultaneously obtains at least one of the following operating status information: liquid supply pump status, liquid supply valve status, pipeline pressure, pipeline flow rate, liquid level of storage container, cabinet door status, and maintenance status. The image acquisition device includes at least one of the following: industrial camera, low-light camera, infrared camera, polarization camera, and multispectral camera. The image acquisition device acquires images using a fixed field of view, multiple field of view, or switchable field of view mode.

[0041] S3. Select a background image corresponding to the current operating state based on the operating state information, and perform distortion correction, brightness normalization, noise suppression, reflective marking and image registration on the detected image to obtain a preprocessed image;

[0042] S4. Compare the preprocessed image with the corresponding background image, and extract liquid candidate regions by combining liquid brightness features, color features, edge features, texture features and motion features between consecutive image frames;

[0043] S5. Perform connected component analysis and morphological processing on the liquid candidate region, and extract the area, shape, boundary, texture, brightness, motion, duration, area growth rate and spatial location features of the liquid target;

[0044] S6. Calculate the leakage risk value based on the visual characteristics, temporal continuity characteristics, and spatial correlation between the liquid target and high-risk components inside the liquid supply cabinet, and determine whether the liquid target is a suspected leakage target.

[0045] S7. Determine the leakage level based on the leakage risk value, the area growth rate of the liquid target, the dripping motion characteristics, the duration, and the consistency with other operating status information;

[0046] S8. Based on the leakage level, output a graded alarm signal or emergency linkage control signal to at least one of the following: liquid supply pump, liquid supply valve, liquid discharge valve, ventilation equipment, audible and visual alarm device, process equipment control system and remote monitoring platform.

[0047] S9. After the emergency response is executed, continue to collect the detection images of the visual detection area, determine the leakage handling results based on the changes in the area, diffusion and duration of the liquid target, and save the leakage event record.

[0048] The liquid supply cabinet is equipped with at least one of visible light source, near-infrared light source, structured light source or pulsed light source. The image acquisition device acquires at least two sets of images under different lighting conditions according to a preset light source sequence, and determines the liquid reflection characteristics based on the differences in the images under different lighting conditions.

[0049] The background image includes reference images of the liquid supply cabinet in at least one of the following states: normal standby state, normal liquid supply state, liquid replacement state, cleaning state, and maintenance state, and the corresponding background image is selected according to the current operating state of the liquid supply cabinet.

[0050] The liquid candidate region is extracted using at least two of the following methods:

[0051] The detected image is compared with the background image;

[0052] Candidate regions are extracted based on local brightness and color changes in the detected image;

[0053] Edge extraction is performed based on droplet boundaries, liquid film boundaries, or liquid flow boundaries;

[0054] Extracting motion regions based on displacement changes between consecutive image frames;

[0055] Liquid regions are extracted based on the differences in response under different polarization states or different spectral bands.

[0056] The morphological characteristics of the liquid target include at least one of area, perimeter, aspect ratio, roundness, convexity, slenderness ratio, boundary curvature, and boundary closure degree. The motion characteristics of the liquid target include at least one of displacement direction, motion velocity, motion acceleration, dripping period, and motion trajectory continuity.

[0057] Target tracking is performed on liquid candidate regions within multiple consecutive time windows. When the liquid candidate region satisfies at least one of the following conditions: the liquid candidate region remains present in multiple consecutive image frames, the area of ​​the liquid candidate region continues to increase, the liquid candidate region spreads along the direction of gravity, the center position of the liquid candidate region shifts along the direction of gravity, the liquid candidate region forms a continuous dripping trajectory below the pipe joint, and the liquid candidate region connects with adjacent liquid candidate regions, the leakage risk value is increased.

[0058] When calculating the leakage risk value, interference correction parameters are determined based on at least one of the following: high-gloss area, metal reflection area, transparent component reflection area, fog area, condensate area, cabinet door opening / closing status, and personnel maintenance status. The interference correction parameters are then used to reduce the leakage risk value of the corresponding liquid candidate area. The leakage level includes at least Level 1 anomaly, Level 2 suspected leakage, Level 3 confirmed leakage, and Level 4 severe leakage. Level 1 anomaly corresponds to a low-confidence liquid candidate area in a single frame or for a short duration. Level 2 suspected leakage corresponds to a persistent liquid target or a droplet target with a continuous movement trajectory. Level 3 confirmed leakage corresponds to a liquid target area continuously increasing, continuous dripping, liquid film diffusion, or a liquid target located at a high-risk pipeline connection. Level 4 severe leakage corresponds to rapid diffusion of the liquid target, large-area liquid accumulation, continuous leakage, leakage in multiple visual detection areas simultaneously, or leakage visual characteristics occurring simultaneously with at least one of pressure anomaly, flow anomaly, and liquid level anomaly.

[0059] For supply cabinets under maintenance, liquid replacement, or cleaning conditions, reduce the leakage risk weight of the corresponding visual inspection area, and re-establish or update the normal background image after maintenance is completed. When a Level 3 confirmed leakage is detected, execute at least one of the following emergency linkage actions: shut down the supply pump, reduce the operating power of the supply pump, close the supply valve corresponding to the leakage area, isolate the pipeline corresponding to the leakage area, activate the audible and visual alarm device, send alarm information to the remote monitoring platform, and send an interlock signal to the process equipment control system. When a Level 4 severe leakage is detected, execute at least one of the following actions: shut down the supply pump, close the supply valve, isolate the relevant pipeline, open the drain valve, start the exhaust equipment, interlock the process equipment, and remote emergency alarm.

[0060] After emergency response is initiated, the area, area growth rate, diffusion rate, dripping frequency, and duration of the liquid target are continuously monitored. When the leakage risk value is lower than the preset clearance threshold within multiple consecutive time windows, a leakage handling completion status is generated. If the leakage risk value does not decrease or continues to increase, the current alarm level is maintained or the alarm level is upgraded. The leakage event record includes at least the leakage occurrence time, leakage visual detection area, leakage image sequence, liquid target location, liquid target area, liquid target movement trajectory, leakage risk value, leakage level, operating status, alarm signal, emergency response action, and response execution result. The leakage risk value is dynamically corrected based on at least one of the following: the hazard level of the liquid being transported in the supply cabinet, liquid type, pipeline pressure, pipeline flow rate, risk level of the visual detection area, and operating status of the process equipment.

[0061] According to the present invention, an image acquisition device and an auxiliary light source are arranged inside a semiconductor liquid supply cabinet, visual detection areas are established for pipelines, pipeline joints, pump bodies, valves, filters, liquid storage containers, cabinet bottom plates and liquid discharge areas, and operation information such as the liquid supply pump, liquid supply valve, pipeline pressure, flow rate, liquid level, cabinet door and maintenance status is synchronously collected; a corresponding background image is selected according to the current operation state of the liquid supply cabinet, and the collected image is subjected to distortion correction, brightness normalization, noise suppression, reflective marking and image registration. Liquid candidate regions are extracted through background difference, brightness and color changes, edge features, texture changes, motion changes of consecutive frames, and differences in response under different illumination, polarization or spectra, and visual features such as the area, morphology, boundary, brightness, texture, motion trajectory, duration, area growth rate and spatial position of the liquid target are further acquired, and correction is performed in combination with interference factors such as highlight, metal reflection, fog, condensed water and manual maintenance, so as to calculate a liquid leakage risk value and determine the liquid leakage level; when suspected liquid leakage, confirmed liquid leakage or severe liquid leakage is determined, emergency measures such as enhanced monitoring, graded alarm, liquid supply pump shutdown, liquid supply valve closing, pipeline isolation, liquid discharge, exhaust start, process equipment interlock and remote alarm are respectively executed, and after linkage, the area, diffusion speed and duration of the liquid target are continuously detected to determine the liquid leakage disposal result, and meanwhile, the liquid leakage image, detection features, alarm information and linkage execution result are stored. By comprehensively utilizing visual features such as liquid droplets, liquid films, liquid flows, reflective boundaries, diffusion boundaries, motion trajectories and time changes, the present invention can perform early identification of micro leakage, continuous dripping, liquid film diffusion, accumulated liquid and flowing leakage in the semiconductor liquid supply cabinet, does not completely rely on liquid conductivity, and has good adaptability to ultrapure water, low-conductivity liquids and some organic solvents; by establishing background models under different operation states and identifying and correcting interferences such as metal reflection, reflection of transparent components, fog, condensed water and manual maintenance, false alarms can be reduced; by dividing the visual detection areas and analyzing the spatial relationship between the liquid target and high-risk components such as pipeline joints, valves, pump bodies and others, the liquid leakage source can be positioned in an auxiliary manner; meanwhile, graded judgment and emergency linkage are carried out based on information such as the liquid leakage area, growth rate, dripping frequency, duration, as well as pressure, flow rate and liquid level, which can shorten the time for liquid leakage discovery and disposal, reduce the risks of chemical liquid diffusion, equipment damage and process contamination, and provide a basis for equipment maintenance, fault analysis and safety traceability by storing complete event records.

[0062] Finally: the above description is only a preferred embodiment of the present invention, and is not intended to limit the present invention. Any modification, equivalent replacement, improvement and the like made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for emergency detection of leakage in a semiconductor liquid supply cabinet based on visual features, characterized in that, Includes the following steps: S1. Obtain the internal structure information of the semiconductor liquid supply cabinet, and establish multiple visual inspection areas based on the internal pipelines, pipeline joints, pumps, valves, filters, liquid storage containers, cabinet bottom plate and drainage area of ​​the liquid supply cabinet; S2. The image acquisition device installed inside the liquid supply cabinet acquires detection images of each visual detection area and simultaneously obtains at least one of the following operating status information: liquid supply pump status, liquid supply valve status, pipeline pressure, pipeline flow rate, liquid level of storage container, cabinet door status, and maintenance status. The image acquisition device includes at least one of the following: industrial camera, low-light camera, infrared camera, polarization camera, and multispectral camera. The image acquisition device acquires images using a fixed field of view, multiple field of view, or switchable field of view mode. S3. Select a background image corresponding to the current operating state based on the operating state information, and perform distortion correction, brightness normalization, noise suppression, reflective marking and image registration on the detected image to obtain a preprocessed image; S4. Compare the preprocessed image with the corresponding background image, and extract liquid candidate regions by combining liquid brightness features, color features, edge features, texture features and motion features between consecutive image frames; S5. Perform connected component analysis and morphological processing on the liquid candidate region, and extract the area, shape, boundary, texture, brightness, motion, duration, area growth rate and spatial location features of the liquid target; S6. Based on the visual characteristics, temporal continuity characteristics, and spatial correlation between the liquid target and high-risk components inside the liquid supply cabinet, calculate the leakage risk value and determine whether the liquid target is a suspected leakage target. S7. Determine the leakage level based on the leakage risk value, the area growth rate of the liquid target, the dripping motion characteristics, the duration, and the consistency with other operating status information; S8. Based on the leakage level, output a graded alarm signal or emergency linkage control signal to at least one of the following: liquid supply pump, liquid supply valve, liquid discharge valve, ventilation equipment, audible and visual alarm device, process equipment control system and remote monitoring platform. S9. After the emergency response is executed, continue to collect the detection images of the visual detection area, determine the leakage handling results based on the changes in the area, diffusion and duration of the liquid target, and save the leakage event record.

2. The emergency detection method for leakage in a semiconductor liquid supply cabinet based on visual features according to claim 1, characterized in that: The liquid supply cabinet is equipped with at least one of visible light source, near-infrared light source, structured light source or pulsed light source. The image acquisition device acquires at least two sets of images under different lighting conditions according to a preset light source sequence, and determines the liquid reflection characteristics based on the differences in images under different lighting conditions.

3. The emergency detection method for leakage in a semiconductor liquid supply cabinet based on visual features according to claim 1, characterized in that: The background image includes reference images of the liquid supply cabinet in at least one of the following states: normal standby state, normal liquid supply state, liquid replacement state, cleaning state, and maintenance state, and the corresponding background image is selected according to the current operating state of the liquid supply cabinet.

4. The emergency detection method for leakage in a semiconductor liquid supply cabinet based on visual features according to claim 1, characterized in that, The liquid candidate region is extracted using at least two of the following methods: The detected image is compared with the background image; Candidate regions are extracted based on local brightness and color changes in the detected image; Edge extraction is performed based on droplet boundaries, liquid film boundaries, or liquid flow boundaries; Extracting motion regions based on displacement changes between consecutive image frames; Liquid regions are extracted based on the differences in response under different polarization states or different spectral bands.

5. The emergency detection method for leakage in a semiconductor liquid supply cabinet based on visual features according to claim 1, characterized in that: The morphological characteristics of the liquid target include at least one of area, perimeter, aspect ratio, roundness, convexity, slenderness ratio, boundary curvature, and boundary closure degree. The motion characteristics of the liquid target include at least one of displacement direction, motion velocity, motion acceleration, dripping period, and motion trajectory continuity.

6. The emergency detection method for leakage in a semiconductor liquid supply cabinet based on visual features according to claim 1, characterized in that: Target tracking is performed on liquid candidate regions within multiple consecutive time windows. When the liquid candidate region satisfies at least one of the following conditions: the liquid candidate region remains present in multiple consecutive image frames, the area of ​​the liquid candidate region continues to increase, the liquid candidate region spreads along the direction of gravity, the center position of the liquid candidate region shifts along the direction of gravity, the liquid candidate region forms a continuous dripping trajectory below the pipe joint, and the liquid candidate region connects with adjacent liquid candidate regions, the leakage risk value is increased.

7. The emergency detection method for leakage in a semiconductor liquid supply cabinet based on visual features according to claim 1, characterized in that: When calculating the leakage risk value, interference correction parameters are determined based on at least one of the following: high-gloss area, metal reflection area, transparent component reflection area, fog area, condensate area, cabinet door opening / closing status, and personnel maintenance status. The interference correction parameters are then used to reduce the leakage risk value of the corresponding liquid candidate area. The leakage level includes at least Level 1 anomaly, Level 2 suspected leakage, Level 3 confirmed leakage, and Level 4 severe leakage. Level 1 anomaly corresponds to a low-confidence liquid candidate area in a single frame or for a short duration. Level 2 suspected leakage corresponds to a persistent liquid target or a droplet target with a continuous movement trajectory. Level 3 confirmed leakage corresponds to a liquid target area continuously increasing, continuous dripping, liquid film diffusion, or a liquid target located at a high-risk pipeline connection. Level 4 severe leakage corresponds to rapid diffusion of the liquid target, large-area liquid accumulation, continuous leakage, leakage in multiple visual detection areas simultaneously, or leakage visual characteristics occurring simultaneously with at least one of pressure anomaly, flow anomaly, and liquid level anomaly.

8. The emergency detection method for leakage in a semiconductor liquid supply cabinet based on visual features according to claim 1, characterized in that: For supply cabinets under maintenance, liquid replacement, or cleaning conditions, reduce the leakage risk weight of the corresponding visual inspection area, and re-establish or update the normal background image after maintenance is completed. When a Level 3 confirmed leakage is detected, execute at least one of the following emergency linkage actions: shut down the supply pump, reduce the operating power of the supply pump, close the supply valve corresponding to the leakage area, isolate the pipeline corresponding to the leakage area, activate the audible and visual alarm device, send alarm information to the remote monitoring platform, and send an interlock signal to the process equipment control system. When a Level 4 severe leakage is detected, execute at least one of the following actions: shut down the supply pump, close the supply valve, isolate the relevant pipeline, open the drain valve, start the exhaust equipment, interlock the process equipment, and remote emergency alarm.

9. The emergency detection method for leakage in a semiconductor liquid supply cabinet based on visual features according to claim 1, characterized in that: After emergency response is initiated, the area, area growth rate, diffusion rate, dripping frequency, and duration of the liquid target are continuously monitored. When the leakage risk value is lower than the preset clearance threshold within multiple consecutive time windows, a leakage handling completion status is generated. If the leakage risk value does not decrease or continues to increase, the current alarm level is maintained or the alarm level is upgraded. The leakage event record includes at least the leakage occurrence time, leakage visual detection area, leakage image sequence, liquid target location, liquid target area, liquid target movement trajectory, leakage risk value, leakage level, operating status, alarm signal, emergency response action, and response execution result. The leakage risk value is dynamically corrected based on at least one of the following: the hazard level of the liquid being transported in the supply cabinet, liquid type, pipeline pressure, pipeline flow rate, risk level of the visual detection area, and operating status of the process equipment.