Canned container leakage point early warning device, control method and control device

Abstract

The invention relates to the technical field of industrial early warning, in particular to a canned container leakage point early warning device, a canned container leakage point early warning control method and a canned container leakage point early warning control device. The technical problem of how to acquire a thermal image in a non-contact manner, accurately determine the boundary of a thermal anomaly region and further analyze an anomaly reason in the prior art is solved. Comprising an infrared thermal imaging unit and a controller, the controller is in communication connection with the infrared thermal imaging unit and is configured to analyze a pixel array to identify a plurality of interconnected abnormal pixels, and the plurality of abnormal pixels jointly form an abnormal thermal feature with a determined boundary; classifying the anomalous thermal feature as a potential leakage point or a transient disturbance event based on one or more morphological characteristics derived from determining the boundary; an alarm signal is generated when the anomalous thermal feature is classified as a potential leakage point.

Description

Technical Field

[0001] This invention relates to the field of industrial early warning technology, and in particular to an early warning device, control method and control device for leak points in canned containers. Background Technology

[0002] In many industrial sectors, such as chemical, oil, and natural gas, the structural integrity of large storage tanks or containers is crucial. These containers often store high-pressure, high-temperature, or toxic materials, and even a minor leak can quickly escalate into a major safety incident. Therefore, effectively monitoring the potential weaknesses or leakage risks of these containers is a core requirement in industrial production.

[0003] Existing detection methods, such as contact sensors or manual inspections, have inherent drawbacks such as monitoring "blind spots," inability to provide real-time continuous monitoring, and strong subjectivity. To overcome these problems, automated infrared monitoring has begun to be applied. However, these methods have fundamental limitations in analytical capabilities, preventing them from solving a core technical problem: how to distinguish between genuine equipment defects and harmless environmental interference.

[0004] In thermal imaging monitoring, actual defects on the equipment surface—such as damaged insulation, internal corrosion leading to wall thinning, or stress concentration from microcracks—usually manifest as a "static thermal anomaly." A "static thermal anomaly" refers to an abnormal temperature region that remains fixed in spatial location and persists over time. Its root cause lies in a change in the physical state of the storage tank itself, thus representing a direct thermal characterization of equipment structural integrity issues.

[0005] In contrast, there are numerous "dynamic or transient thermal disturbances," such as temporary temperature anomalies caused by a worker walking near a container, a hot vehicle driving by, or even changes in the angle of sunlight. These disturbances are characterized by their non-fixed location and short duration.

[0006] Existing technologies typically employ primitive thresholding techniques to detect thermal anomalies. This method processes pixels in isolation, judging only whether their temperature values ​​exceed preset thresholds, completely ignoring the spatial correlation between pixels. The direct consequence is an inability to effectively distinguish between two fundamentally different types of thermal signals (i.e., static thermal anomalies versus dynamic or transient thermal interference). More importantly, this method cannot accurately extract the boundary of a complete, continuous anomaly region, often producing scattered, discontinuous noise, making subsequent shape analysis or pattern recognition extremely difficult or unreliable. If even the accurate contour of the anomaly region cannot be obtained, then analyzing its cause (whether it's a defect or interference) becomes impossible.

[0007] Therefore, there is an urgent need in the field for a leak warning device for canned containers that can acquire thermal images non-contactly and accurately determine the boundaries of thermal anomaly areas in order to facilitate the analysis of the causes of the anomalies. Summary of the Invention

[0008] This invention provides a leak point early warning device, control method, and control device for canned containers, which solves the technical problem in the prior art of how to acquire thermal images in a non-contact manner, accurately determine the boundary of thermal anomaly areas, and then analyze the cause of the anomaly.

[0009] The above-mentioned objectives of the present invention can be achieved by the following technical solutions: This invention provides a leak warning device for canned containers, comprising an infrared thermal imaging unit and a controller. The infrared thermal imaging unit is configured to detect the temperature distribution on the surface of the canned container and convert the temperature distribution into a thermal image containing a pixel array, wherein each pixel in the pixel array has a temperature value. The controller is communicatively connected to the infrared thermal imaging unit and is configured to: analyze the pixel array to identify multiple interconnected abnormal pixels, wherein the multiple abnormal pixels together constitute an abnormal thermal feature with a defined boundary; classify the abnormal thermal feature as a potential leak point or a transient interference event based on one or more morphological characteristics derived from the defined boundary; and generate an alarm signal when the abnormal thermal feature is classified as a potential leak point.

[0010] According to one embodiment of the present invention, analyzing the pixel array to identify a plurality of interconnected anomalous pixels, the plurality of anomalous pixels collectively constituting an anomalous thermal feature with defined boundaries, includes: identifying an initial anomalous pixel from the pixel array as a seed pixel, and creating a blank data structure designated for storing the anomalous thermal feature; adding the pixel coordinates of a pixel to be processed to the data structure, and subsequently identifying a set of valid neighboring pixels that meet predetermined anomalous criteria from all neighboring pixels of the pixel to be processed, wherein the pixel to be processed represents the seed pixel at the time of the first execution; processing the identified set of valid neighboring pixels as a new input group; for each pixel in the new input group that is not yet present in the data structure, recursively repeating steps 1 and 2, wherein the process terminates when no new valid neighboring pixels can be identified, and the final state of the data structure constitutes an anomalous thermal feature with defined boundaries.

[0011] According to one embodiment of the present invention, the one or more morphological properties include at least one of the following: shape, area, aspect ratio, and boundary complexity.

[0012] According to one embodiment of the present invention, classifying the abnormal thermal feature as a potential leak point or transient interference event based on one or more morphological properties derived from the defined boundary includes: classifying the abnormal thermal feature as a transient interference event if the shape of the abnormal thermal feature matches the profile of a known interference source that has been stored.

[0013] According to one embodiment of the present invention, the known interference source profile is a human figure profile or a vehicle profile.

[0014] According to one embodiment of the present invention, classifying the abnormal thermal features as potential leak points or transient interference events based on one or more morphological properties derived from the determined boundary includes: using the one or more morphological properties as input to a pre-trained machine learning model to classify the abnormal thermal features as potential leak points or transient interference events.

[0015] According to one embodiment of the present invention, an adjustment unit is further included, which is communicatively connected to the controller and configured to adjust the position of the infrared thermal imaging unit; the infrared thermal imaging unit is disposed on the adjustment unit.

[0016] According to one embodiment of the present invention, the potential leak point is characterized by its abnormal thermal features remaining in position across multiple consecutively captured thermal images.

[0017] The present invention also provides a control method for a leak point early warning device for canned containers, comprising: analyzing the pixel array to identify multiple interconnected abnormal pixels, wherein the multiple abnormal pixels together constitute an abnormal thermal feature with a defined boundary; classifying the abnormal thermal feature as a potential leak point or a transient interference event based on one or more morphological properties derived from the defined boundary; and generating an alarm signal when the abnormal thermal feature is classified as a potential leak point.

[0018] The present invention also provides a control device, comprising: a canned container leak point early warning device applied to the above embodiments, the canned container leak point early warning device comprising: an analysis module configured to analyze the pixel array to identify multiple interconnected abnormal pixels, the multiple abnormal pixels collectively constituting an abnormal thermal feature with a defined boundary; a classification module configured to classify the abnormal thermal feature as a potential leak point or a transient interference event based on one or more morphological characteristics derived from the defined boundary; and a generation module configured to generate an alarm signal when the abnormal thermal feature is classified as a potential leak point.

[0019] The features and advantages of the leak warning device, control method, and control device for canned containers of the present invention are as follows: First, an infrared thermal imaging unit captures the complete thermal distribution on the tank surface in real time using a non-contact method, converting it into a pixel array composed of temperature values. Then, the controller performs connectivity analysis on the pixel array, identifying multiple interconnected anomalous pixels and aggregating them into an anomalous thermal feature with a defined boundary. This "defined boundary" is not a vague temperature transition zone, but a closed contour strictly defined by pixel connectivity and anomaly criteria, allowing for subsequent morphological quantification of the region (e.g., area, shape). Thus, not only is non-contact perception of anomalous events such as leaks achieved, but more importantly, it solves the technical bottleneck of vague boundaries and difficulty in accurately extracting thermal anomaly regions, laying a data foundation for automated classification and root cause analysis. Attached Figure Description

[0020] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0021] Figure 1 This is a control principle diagram of the present invention; Figure 2 This is a perspective view of the adjustment unit of the present invention; Figure 3 for Figure 2 Enlarged view of section A.

[0022] Explanation of reference numerals in the attached figures: 1. Adjustment unit; 11. Lifting mechanism; 12. Rotating bracket; 121. Fixing component; 122. Damping hinge; 123. Support component. Detailed Implementation

[0023] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. 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.

[0024] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention.

[0025] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified. Furthermore, the terms "installed," "connected," and "linked" should be interpreted broadly; for example, they may refer to a fixed connection, a detachable connection, or an integral connection; they may refer to a mechanical connection or an electrical connection; they may refer to a direct connection or an indirect connection through an intermediate medium; and they may refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0026] like Figures 1 to 3 As shown, the present invention provides a leak warning device for canned containers, comprising: an infrared thermal imaging unit configured to detect the temperature distribution on the surface of the canned container and convert the temperature distribution into a thermal image containing a pixel array, wherein each pixel in the pixel array has a temperature value; a controller communicatively connected to the infrared thermal imaging unit and configured to: analyze the pixel array to identify multiple interconnected abnormal pixels, wherein the multiple abnormal pixels together constitute an abnormal thermal feature with a defined boundary; classify the abnormal thermal feature as a potential leak point or a transient interference event based on one or more morphological characteristics derived from the defined boundary; and generate an alarm signal when the abnormal thermal feature is classified as a potential leak point.

[0027] In practice, the infrared thermal imaging unit first captures the complete thermal distribution on the tank surface in real time using a non-contact method, converting it into a pixel array composed of temperature values. Subsequently, the controller performs connectivity analysis on the pixel array, identifying multiple interconnected abnormal pixels and aggregating them into an abnormal thermal feature with a defined boundary. This "defined boundary" is not a vague temperature transition zone, but a closed contour strictly defined by pixel connectivity and anomaly criteria, allowing for subsequent morphological quantification of the region (e.g., area, shape). This not only achieves non-contact perception of abnormal events such as leaks, but more importantly, it solves the technical bottleneck of vague boundaries and difficulty in accurately extracting thermal anomaly regions, laying a data foundation for automated classification and root cause analysis.

[0028] In this embodiment, the "defined boundary" is not a rough isotherm drawn after simply setting a temperature threshold, but rather a closed region boundary constructed based on pixel spatial connectivity. For example, when liquefied gas leaks from a tiny crack, it will form a localized low-temperature zone on the tank wall due to heat absorption during vaporization. After the infrared thermal imager captures this area, the controller will identify all pixels with temperatures below the threshold that are adjacent to each other. These pixels can be connected into a whole using a region growing algorithm, generating a clear outer contour line. This contour line is the "defined boundary," and its coordinates and geometric properties can be precisely calculated, thereby supporting subsequent shape matching or machine learning classification.

[0029] In some implementations, a buzzer may also be included for connection to the controller to emit an alarm signal.

[0030] According to one embodiment of the present invention, analyzing a pixel array to identify multiple interconnected anomalous pixels, the multiple anomalous pixels collectively constituting an anomalous thermal feature with defined boundaries, includes: (i) identifying an initial anomalous pixel from the pixel array as a seed pixel and creating a blank data structure designated for storing the anomalous thermal feature; (ii) identifying a set of valid neighboring pixels that meet predetermined anomalous criteria from all neighboring pixels of the pixel to be processed by adding the pixel coordinates of a pixel to be processed to the data structure, wherein the pixel to be processed represents the seed pixel at the time of the first execution; (iii) processing the identified set of valid neighboring pixels as a new input group; (iv) for each pixel in the new input group that is not yet present in the data structure, recursively repeating steps (ii) and (iii), wherein the process terminates when no new valid neighboring pixels can be identified, and the final state of the data structure constitutes an anomalous thermal feature with defined boundaries.

[0031] In practice, the identification process for anomalous thermal features employs a recursive algorithm based on "region growing." Starting from an initial anomalous pixel (seed point), it gradually expands to all neighboring pixels that meet the anomalous criteria, ultimately constructing a complete boundary for the anomalous region. This method effectively avoids misclassifying isolated noise as leakage signals, while ensuring that even anomalous regions with complex shapes (such as crack-like leaks) can be completely captured. Because this process strictly relies on the spatial connectivity between pixels and the consistency of temperature thresholds, the generated anomalous thermal features possess spatial continuity and physical authenticity, providing reliable input for subsequent classification.

[0032] In this implementation, "region growing" starts from one or more "seed points" and continuously "absorbs" neighboring pixels that meet the criteria until it can no longer expand. For example, suppose a pixel's temperature is more than 5°C lower than its surroundings and is selected as a seed; then, its four neighboring points (up, down, left, and right) are checked. If any of these neighboring points also meets the "low-temperature anomaly" condition, it is added to the same region; then, the search continues recursively, starting with the newly added pixel, until the boundary is closed. This algorithm is particularly suitable for detecting continuous leakage traces, such as linear low-temperature bands formed by slow seepage along welds. The "data structure" here typically refers to a list or set used to dynamically store the coordinates of all pixels confirmed to belong to the anomaly region.

[0033] According to one embodiment of the present invention, one or more morphological properties include at least one of the following: shape, area, aspect ratio, and boundary complexity.

[0034] In practice, shape, area, aspect ratio, and boundary complexity can characterize the geometric properties of thermal anomaly regions from different dimensions, thus providing multidimensional criteria for distinguishing leak points from interference events. For example, real leak points are usually small in area, regular in shape (close to a circle or ellipse), and have smooth boundaries; while thermal interference caused by human shadows or vehicles often has a large area, high aspect ratio, and jagged boundaries. By quantifying these characteristics, more robust classification rules can be established, reducing the risk of misjudgment based on a single indicator.

[0035] In this embodiment, "aspect ratio" refers to the ratio of the length to the width of the rectangle circumscribed in the abnormal area, used to determine whether it is strip-shaped (such as the thermal shadow left by a person walking by); "boundary complexity" measures the degree of curvature of the contour, which can be obtained by calculating the ratio of the actual boundary length to the equivalent circumference of the circle—the more complex the boundary (such as the heat spot caused by the shadow of a leaf), the higher the ratio. For example, a circular leak point with a diameter of 2 cm has an aspect ratio close to 1, indicating low boundary complexity; while the thermal shadow cast by a pedestrian passing by the side of the tank may be as long as 1 meter and only 0.3 meters wide, with an aspect ratio of about 3.3, and the edges are highly irregular due to clothing folds, indicating high complexity. These differences can serve as key criteria for automatic interference filtering.

[0036] According to one embodiment of the present invention, classifying abnormal thermal features as potential leak points or transient interference events based on one or more morphological properties derived from a defined boundary includes: classifying the abnormal thermal feature as a transient interference event if the shape of the abnormal thermal feature matches the profile of a known interference source that has been stored.

[0037] In practice, if the shape of an abnormal thermal feature is highly similar to the outline of a pre-stored known interference source (such as a human figure or vehicle), it is directly identified as a transient interference event and no alarm is triggered. This "template matching" strategy significantly reduces the false alarm rate caused by common human or environmental factors, and is particularly suitable for scenarios where inspection personnel or transport vehicles frequently pass by the tank area. By pre-establishing a database of interference source thermal outlines, comparisons can be completed within milliseconds, achieving real-time and accurate interference suppression.

[0038] In this embodiment, "contour matching" refers to calculating the similarity between the boundary of the currently detected abnormal region and a standard contour (e.g., a thermal silhouette of a standing human figure) stored in the database. Common methods include Hu moment invariants, Fourier descriptors, or deep learning feature comparison. For example, a typical human thermal contour template (approximately 1.7 meters high, 0.5 meters wide, with prominent shoulders) can be stored. When a hot spot of a similar shape is detected, even if its temperature is abnormal, it will be marked as "non-leaking" due to a successful match. This approach is similar to template matching in face recognition, but it is applied to the field of thermal imaging specifically to exclude foreseeable non-faulty heat sources.

[0039] According to one embodiment of the present invention, the known interference source profile is a human figure profile or a vehicle profile.

[0040] In practice, the known types of interference sources are human or vehicle silhouettes, which are the two most common sources of thermal interference in industrial settings. When vehicles or personnel approach storage tanks during inspections, their own heat radiation or shadows cast by blocking sunlight can create significant anomalies in thermal images, easily mistaken for leaks. By specifically modeling the thermal characteristics of these two types of interference, up to 80% of daily false alarms can be effectively filtered, significantly improving the availability of early warnings and user trust.

[0041] In this embodiment, during actual deployment, thermal imaging profiles of multiple workers standing and walking, as well as thermal images of different types of forklifts and tank trucks, can be pre-collected to construct a small interference template library. For example, when a tank truck fully loaded with liquefied gas is parked, its engine compartment and tires will emit heat, forming specific hot spots on the tank surface; and when an inspector holds metal tools close to the tank wall, local temperature changes may also occur due to heat conduction. After standardizing the thermal profiles of these typical scenarios, the controller can quickly compare them during operation, avoiding frequent false alarms.

[0042] According to one embodiment of the present invention, classifying abnormal thermal features as potential leak points or transient interference events based on one or more morphological properties derived from a defined boundary includes: using one or more morphological properties as input to a pre-trained machine learning model to classify abnormal thermal features as potential leak points or transient interference events.

[0043] In practice, machine learning models replace traditional rule-based judgments, using morphological characteristics as input features. The pre-trained model automatically classifies leakage points and interference events. Compared to fixed thresholds or simple template matching, machine learning methods can capture more complex nonlinear relationships (such as the joint distribution of area and boundary complexity), adapt to more diverse leakage patterns and interference types, and possess stronger generalization and adaptability. Especially when facing novel interference or rare leakage patterns, the model maintains a high accuracy rate.

[0044] In this implementation, a "pre-trained machine learning model" typically refers to a classifier trained on a large amount of labeled thermal image data, such as a Support Vector Machine (SVM), Random Forest, or Lightweight Convolutional Neural Network (CNN). The training data includes tens of thousands of anomalous thermal feature samples labeled as "leakage" or "interference," each sample accompanied by its morphological parameters (e.g., area = 150 pixels squared, aspect ratio = 1.2, complexity = 1.05). After learning the mapping relationship between these parameters and categories, the model can intelligently identify newly emerging anomalous regions. For example, even if a leak point is irregularly star-shaped, it will still be correctly identified as long as its parameter combination falls within the "leakage" region determined by the model.

[0045] According to one embodiment of the present invention, it further includes an adjustment unit 1, which is communicatively connected to the controller and configured to adjust the position of the infrared thermal imaging unit; the infrared thermal imaging unit is disposed on the adjustment unit 1.

[0046] In practical implementation, an adjustment unit 1 is added to dynamically adjust the position of the infrared thermal imaging unit, enabling the device to cover a larger area of ​​the tank surface or focus on high-risk areas (such as valves and welds). This mechanically adjustable structure solves the problem of limited field of view of fixed-installation thermal imagers, and is particularly suitable for large spherical tanks or multi-tank clusters. By instructing the adjustment unit 1 to operate via the controller, periodic scanning or fixed-point monitoring can be achieved, improving detection coverage and flexibility.

[0047] In this embodiment, the adjustment unit 1 may include a lifting mechanism 11 and a rotating bracket 12; the lifting mechanism 11 may be a vertically arranged linear module and electrically connected to the controller; the rotating bracket 12 is arranged on the slide of the linear module, and the rotating bracket 12 may include a fixing member 121, a damping hinge 122 and a support member 123; the fixing member 121 is arranged on the slide, and the fixing member 121 is connected to the support member 123 through the damping hinge 122; the support member 123 has a mounting hole for mounting an infrared thermal imaging unit, and the infrared thermal imaging unit is mounted in the mounting hole.

[0048] In practice, the above-mentioned structural design allows for the adjustment of not only the position of the infrared thermal imaging unit but also its angle.

[0049] According to one embodiment of the present invention, the potential leak point is characterized by its abnormal thermal features remaining in position across multiple consecutively captured thermal images.

[0050] In practice, a potential leak point is only identified if the abnormal thermal feature remains in a constant position across multiple consecutive thermal images. This mechanism effectively filters out transient thermal interference (such as birds flying by or sunlight flashing), because a true leak is a continuous physical process whose thermal features are spatially stable. Inter-frame consistency verification significantly improves the confidence level of the alarm, preventing false triggering due to single-frame anomalies.

[0051] In this implementation, "continuous multiple frames" typically refers to 3-5 frames of thermal images acquired consecutively within a short period (e.g., within 1 second). The controller records the center coordinates of abnormal thermal features in each frame. If the coordinate deviation between frames is less than a preset threshold (e.g., ±2 pixels), it is considered "position unchanged." For example, a bird flying over a tank surface might create a low-temperature shadow in one frame, but disappear or shift significantly in the next frame, thus being excluded; while a real valve leak point would appear in the same location in every frame, continuously releasing cold air and forming a stable low-temperature spot. This spatiotemporal joint judgment is a classic noise reduction strategy in video analysis.

[0052] The present invention also provides a control method for a leak point early warning device for canned containers, comprising: S101 analyzing a pixel array to identify multiple interconnected abnormal pixels, wherein the multiple abnormal pixels together constitute an abnormal thermal feature with a defined boundary; S102 classifying the abnormal thermal feature as a potential leak point or a transient interference event based on one or more morphological properties derived from the defined boundary; and S103 generating an alarm signal when the abnormal thermal feature is classified as a potential leak point.

[0053] The present invention also provides a control device, comprising: a canned container leak point early warning device applied to the above embodiments, the canned container leak point early warning device comprising: an analysis module configured to analyze a pixel array to identify multiple interconnected abnormal pixels, the multiple abnormal pixels collectively constituting an abnormal thermal feature with a defined boundary; a classification module configured to classify the abnormal thermal feature as a potential leak point or a transient interference event based on one or more morphological characteristics derived from the defined boundary; and a generation module configured to generate an alarm signal when the abnormal thermal feature is classified as a potential leak point.

[0054] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A can container leak point early warning device characterized by, comprising an infrared thermal imaging unit and a controller; the infrared thermal imaging unit is configured to detect a temperature distribution of a surface of a canned container and convert the temperature distribution into a thermal image comprising an array of pixels, each pixel having a temperature value; the controller is communicatively coupled to the infrared thermal imaging unit and configured to: analyze the array of pixels to identify a plurality of connected abnormal pixels that collectively form an abnormal thermal feature having a determined boundary; classify the abnormal thermal feature as a potential leak point or a transient interference event based on one or more morphological characteristics derived from the determined boundary; generate an alarm signal when the abnormal thermal feature is classified as the potential leak point.

2. The can container leak point early warning device of claim 1, wherein, the analyzing the array of pixels to identify a plurality of connected abnormal pixels that collectively form an abnormal thermal feature having a determined boundary comprises: (i) identifying an initial abnormal pixel as a seed pixel from the array of pixels and creating an empty data structure designated for storing the abnormal thermal feature; (ii) identifying a set of valid neighboring pixels from all neighboring pixels of a pixel to be processed that satisfy a predetermined abnormality criterion, wherein the pixel to be processed represents the seed pixel at a first execution; (iii) processing the identified set of valid neighboring pixels as a new input set; (iv) recursively repeating steps (ii) and (iii) for each pixel in the new input set that is not already present in the data structure, wherein each pixel is treated as a new pixel to be processed, wherein the recursion terminates when no new valid neighboring pixel can be identified, and the final state of the data structure constitutes the abnormal thermal feature having a determined boundary.

3. The can container leak point early warning device of claim 1, wherein, the one or more morphological characteristics include at least one of shape, area, aspect ratio, and boundary complexity.

4. The can container leak point early warning device of claim 3, wherein, the classifying the abnormal thermal feature as a potential leak point or a transient interference event based on one or more morphological characteristics derived from the determined boundary comprises: classifying the abnormal thermal feature as a transient interference event if its shape matches a stored profile of a known interference source.

5. The can container leak point early warning device of claim 4, wherein, the known interference source profile is a human profile or a vehicle profile.

6. The can container leak point early warning device of claim 3, wherein, the classifying the abnormal thermal feature as a potential leak point or a transient interference event based on one or more morphological characteristics derived from the determined boundary comprises: classifying the abnormal thermal feature as a potential leak point or a transient interference event using the one or more morphological characteristics as inputs to a pre-trained machine learning model.

7. The can container leak point early warning device of claim 1, wherein, further comprising an adjustment unit communicatively coupled to the controller and configured to adjust a position of the infrared thermal imaging unit; the infrared thermal imaging unit is disposed on the adjustment unit.

8. The can container leak point early warning device of claim 1, wherein, the potential leak point is characterized by its abnormal thermal feature being stationary in position across a plurality of consecutively captured thermal images.

9. A control method of a can container leak point early warning device, characterized by, comprising: analyzing the array of pixels to identify a plurality of connected abnormal pixels that collectively form an abnormal thermal feature having a determined boundary; based on one or more morphological properties derived from the determined boundary, classifying the anomalous thermal feature as a potential leak point or a transient disturbance event; generating an alarm signal when the anomalous thermal feature is classified as the potential leak point.

10. A control device characterized by comprising: comprising: a can container leak point early warning device as claimed in any one of claims 1 to 8, the can container leak point early warning device comprising: an analysis module configured to analyze the array of pixels to identify a plurality of interconnected anomalous pixels that collectively form an anomalous thermal feature having a determined boundary; a classification module configured to classify the anomalous thermal feature as a potential leak point or a transient disturbance event based on one or more morphological properties derived from the determined boundary; a generation module configured to generate an alarm signal when the anomalous thermal feature is classified as the potential leak point.

Images