An artificial intelligence-based anti-oil theft warning system and method for oil tank trucks

Abstract

The present invention relates to the field of oil theft warning technology, and specifically, to an oil tanker oil theft warning system and method based on artificial intelligence. It includes: an oil quantity real-time monitoring unit monitors the oil level in real time through an ultrasonic level sensor; a data acquisition and processing unit identifies abnormal behavior by collecting and analyzing video data, and comprehensively judges the oil theft risk in combination with the oil quantity data, generates an early warning instruction according to the oil theft risk, considers the influence of the angle of reflected light on the light intensity in the judgment process, and designs the correction item of the reflected light angle in the judgment process, and designs the correction item of the reflected light angle in combination with the relative position of the reflected light source, the normal direction of the surface of the tanker truck, and the surface roughness; the early warning response unit receives the early warning instruction and triggers the alarm system. The design of the present invention can more accurately judge whether there is a risk of oil theft by combining the ultrasonic level sensor to monitor the oil quantity changes in real time and the identification of abnormal behavior in the video data.

Description

Technical Field

[0001] The present invention relates to the technical field of oil theft prevention and early warning technology, and in particular to an artificial intelligence-based oil tanker oil theft prevention and early warning system and method. Background Art

[0002] First, in terms of monitoring oil theft, traditional methods are difficult to achieve real-time and accurate monitoring of oil theft activities, especially in complex environments or under specific lighting conditions, where ordinary monitoring equipment may not be able to effectively identify abnormal situations. Secondly, if only a single data source, such as a liquid level sensor, is relied upon to detect changes in oil volume, it is easy to lead to a high false alarm rate, because this method cannot fully consider all factors that may cause oil volume fluctuations. In addition, external environmental factors such as strong direct or reflected light may interfere with the normal operation of monitoring equipment and affect the quality of monitoring images, thereby reducing the ability to identify abnormal behavior. Finally, the acceleration and speed changes of tank trucks during driving can cause normal oil volume fluctuations, which makes it particularly difficult to distinguish these normal fluctuations from actual oil theft. Therefore, an artificial intelligence-based anti-oil theft warning system and method for tank trucks are provided. Summary of the invention

[0003] The purpose of the present invention is to provide an artificial intelligence-based anti-oil theft warning system and method for tank trucks to solve the problems raised in the above background technology that oil theft is difficult to monitor, traditional monitoring methods have a high false alarm rate, and environmental factors have strong interference.

[0004] To achieve the above objectives, on the one hand, the present invention aims to provide an artificial intelligence-based anti-oil theft warning system for tank trucks, comprising: An oil level real-time monitoring unit, wherein the oil level real-time monitoring unit monitors the oil level in real time through an ultrasonic level sensor; A data acquisition and processing unit, wherein the data acquisition and processing unit identifies abnormal behavior by collecting and analyzing video data, and comprehensively judges the risk of oil theft in combination with the oil volume data, generates an early warning instruction according to the risk of oil theft, considers the influence of the angle of reflected light on the light intensity in the judgment process, and designs the correction item of the reflected light angle in the judgment process in combination with the relative position of the reflected light source, the normal direction of the surface of the tanker truck, and the surface roughness; An early warning response unit receives the early warning instruction and triggers the alarm system.

[0005] As a further improvement of the technical solution, the oil quantity real-time monitoring unit includes an oil level monitoring module and an oil quantity conversion module; Among them, the oil level monitoring module monitors the oil level in real time through an ultrasonic level sensor installed at the bottom of the oil tank; The oil volume conversion module converts the oil level data into oil volume through geometric formula method.

[0006] As a further improvement of the technical solution, the data acquisition and processing unit includes a data acquisition module and a data analysis and early warning module; Among them, the data acquisition module collects video data in real time through the camera at the oil tank port of the oil tanker; The data analysis and early warning module analyzes video data through convolutional neural network models and long short-term memory network models, identifies abnormal behavior, and comprehensively judges the risk of oil theft based on oil volume data.

[0007] As a further improvement of the technical solution, the data analysis and early warning module analyzes the video data through a convolutional neural network model and a long short-term memory network model, identifies abnormal behavior, and comprehensively determines the risk of oil theft in combination with the oil volume data, including the following steps: S1.1. Use the video dataset with labeled normal operation and oil theft behavior to train the convolutional neural network model to identify abnormal behavior. S1.2, use convolutional neural network to extract features from each frame of video image; S1.3, taking the features of each frame extracted from the convolutional neural network as the input of the long short-term memory network model; S1.4, sending the output of the long short-term memory network model to the fully connected layer to predict whether there is abnormal behavior in the current video image frame; S1.5. Real-time monitoring of oil volume changes through sensors, forming a time series of oil volume data , indicating the amount of oil in the tank at each moment; S1.6. Determine whether a sudden drop has occurred by calculating the change in oil volume; S1.7. Combine the oil volume change and abnormal video behavior to calculate the comprehensive risk score and generate early warning instructions.

[0008] As a further improvement of the technical solution, in S1.2, a convolutional neural network is used to extract features from each frame of video image, including the following steps: S1.21. Convolve the input image with multiple convolution kernels through a convolution operation. To address the problem of oil thieves interfering with monitoring by using strong direct light, a dynamic illumination compensation term is introduced into the convolution operation for optimization. The correction term of the reflected light angle is introduced into the calculation of the illumination compensation term. The correction term of the reflected light angle is designed in combination with the relative position of the reflected light source, the normal direction of the surface of the tanker truck, and the surface roughness, to further optimize the convolution operation. S1.22. Apply the ReLU activation function to the output of the convolution operation. S1.23, after the convolutional layer, a pooling operation is performed using the maximum pooling method; S1.24, the output layer classifies the behavior according to the result of feature extraction, and determines whether there is oil theft through classification: ; in, Indicates that given an input image The probability of identifying oil theft behavior; represents the activation function; represents the weight of the fully connected layer; represents the bias term of the fully connected layer; Represents the vector after flattening the feature map after convolution and pooling; Represents the input image.

[0009] As a further improvement of the technical solution, in S1.21, the convolution operation is: ; in, Represents the output feature map after the convolution operation; Represents the input image In Location The pixel value at ; Indicates the current frame; represents the convolution kernel; Indicates the position of the current image; Indicates the horizontal offset of the convolution kernel; Indicates the vertical offset of the convolution kernel; To solve the problem of oil thieves interfering with monitoring by using strong direct light, the light intensity at each pixel position is calculated, the light intensity of each pixel value is normalized, and a dynamic light compensation term is introduced in the convolution operation for optimization based on the calculated light intensity: ; in, Represents the result of the convolution operation after the dynamic lighting compensation term is introduced; Indicates the current pixel position The estimated local light intensity at ; represents a constant; Represents the weight coefficient of the lighting compensation item; Represents the sigmoid function; represents the light threshold; In view of the influence of the angle of reflected light on the light intensity, the correction term of the reflected light angle is introduced into the calculation of the light compensation term. The correction term of the reflected light angle is designed in combination with the relative position of the reflected light source, the normal direction of the tanker surface and the surface roughness to further optimize the convolution operation: ; ; in, The correction term representing the reflected light angle is the result of the convolution operation after introducing the illumination compensation term; A correction term for the angle of reflected light; represents the cosine of the angle between the incident light and the normal line of the tanker surface; Represents the angle between the incident light direction and the surface normal; Indicates the surface roughness correction factor; Indicates the distance between the surface of the tanker and the reflected light source; Indicates the maximum influence distance.

[0010] As a further improvement of the technical solution, in S1.7, the oil volume change and the abnormal behavior of the video are combined to calculate the comprehensive risk score, including the following steps: S1.71. Combining the oil volume change and the video abnormal behavior score to form a weighted score. For the oil volume change of the tanker during movement, the acceleration and speed of the tanker are introduced into the weighted score to distinguish between normal oil volume fluctuations caused by movement and oil theft; S1.72, calculate the cumulative risk score based on the oil volume changes and abnormal video behaviors at multiple time points through sliding window technology; S1.73. Calculate the cumulative risk score. When the cumulative risk score exceeds the preset threshold Rt, generate an early warning instruction.

[0011] As a further improvement of the technical solution, in S1.71, the weighted score is: ; in, represents the risk score; Indicates the weight of video data in the comprehensive risk score; Indicates that given video data The probability of oil theft under the condition of Indicates the weight coefficient of oil volume change data in the comprehensive risk score; Represents video data; Indicates changes in oil volume; Indicates the maximum capacity of the tanker truck's fuel tank; Indicates time; In view of the change in oil volume during the movement of the tanker, the acceleration and speed of the tanker are introduced into the weighted score to distinguish between normal oil volume fluctuations caused by movement and oil theft: ; in, represents the risk score after introducing the acceleration and speed of the tanker truck; represents the adjustment coefficient; represents the acceleration of the tanker truck; Indicates the speed of the tanker truck.

[0012] As a further improvement of the technical solution, the early warning response unit includes an instruction sending module and an early warning module; Wherein, the instruction sending module receives the generated warning instruction and sends the warning instruction to the warning module; The early warning module receives the early warning instruction and triggers an alarm.

[0013] On the other hand, the present invention provides an artificial intelligence-based anti-oil theft warning method for oil tankers, which is used in the artificial intelligence-based anti-oil theft warning system for oil tankers, and comprises the following steps: S2.1, real-time monitoring of oil level height through ultrasonic level sensor; S2.2. Identify abnormal behaviors by collecting and analyzing video data, and comprehensively judge the risk of oil theft in combination with oil volume data. Generate early warning instructions based on the risk of oil theft. Consider the impact of the angle of reflected light on the light intensity during the judgment process. Design the correction item of the reflected light angle in combination with the relative position of the reflected light source, the normal direction of the surface of the tanker truck, and the surface roughness. S2.3. Receive early warning instructions and trigger the alarm system.

[0014] Compared with the prior art, the present invention has the following beneficial effects: 1. An AI-based anti-oil theft warning system and method for oil tankers can more accurately determine whether there is a risk of oil theft by combining ultrasonic level sensors to monitor oil volume changes in real time and identify abnormal behaviors in video data. In particular, the use of convolutional neural networks (CNN) and long short-term memory networks (LSTM) models to analyze video data can not only capture instantaneous abnormal behaviors, but also analyze the persistence and pattern of abnormal behaviors based on time series, greatly improving the accuracy and reliability of oil theft detection. In addition, considering the impact of environmental factors such as reflected light angle and surface roughness on light intensity, the image processing process is further optimized to reduce the possibility of false alarms.

[0015] 2. An AI-based anti-oil theft warning system and method for tanker trucks takes into account complex and changeable practical application scenarios, such as monitoring interference caused by strong direct light or reflected light. By introducing dynamic light compensation items and reflected light angle correction items, monitoring failures or misjudgments caused by changes in external light are effectively dealt with. This not only enhances the robustness of the system under different lighting conditions, but also ensures that oil theft can be accurately identified even in the presence of strong light or reflected light interference, thereby improving the overall stability and practicality of the system. At the same time, the system uses acceleration and velocity information to distinguish between normal oil volume fluctuations and oil theft, further improving the applicability and accuracy of early warnings in a mobile state. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is the overall flow chart of the present invention; The meaning of each number in the figure is: 1. Oil quantity real-time monitoring unit; 11. Oil level monitoring module; 12. Oil quantity conversion module; 2. Data acquisition and processing unit; 21. Data acquisition module; 22. Data analysis and early warning module; 3. Early warning response unit; 31. Command sending module; 32. Early warning module. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] Example 1: Please refer to Figure 1 As shown, an artificial intelligence-based anti-oil theft warning system for tank trucks is provided, comprising: The oil level real-time monitoring unit 1 monitors the oil level in real time through an ultrasonic level sensor; In this embodiment, the oil quantity real-time monitoring unit 1 includes an oil level monitoring module 11 and an oil quantity conversion module 12; The oil level monitoring module 11 monitors the oil level in real time through an ultrasonic level sensor installed at the bottom of the oil tank; The oil quantity conversion module 12 converts the oil level height data into the oil quantity value through a geometric formula method. The geometric formula method is based on the geometric shape of the oil tank and uses the oil level height to calculate the volume of the corresponding shape to obtain the oil quantity value.

[0019] The data acquisition and processing unit 2 identifies abnormal behavior by collecting and analyzing video data, and comprehensively judges the risk of oil theft in combination with the oil volume data, generates an early warning instruction according to the risk of oil theft, considers the influence of the angle of reflected light on the light intensity in the judgment process, and designs the correction item of the reflected light angle in combination with the relative position of the reflected light source, the normal direction of the surface of the tanker truck and the surface roughness during the judgment process; In this embodiment, the data acquisition and processing unit 2 includes a data acquisition module 21 and a data analysis and early warning module 22; The data acquisition module 21 collects video data in real time through the camera at the oil tank port of the oil tanker; The data analysis and early warning module 22 analyzes video data through a convolutional neural network (CNN) model and a long short-term memory network (LSTM) model to identify abnormal behaviors (CNN is used to extract spatial features of video frames, while LSTM analyzes dynamic changes in time series based on these features to identify abnormal behaviors), and comprehensively judges the risk of oil theft in combination with oil volume data; CNN is good at capturing spatial features in images and can effectively extract visual information related to oil theft from video frames, such as human actions and equipment operations. LSTM can process time series data and capture the dynamic changes of these behaviors in the time dimension, thereby improving the accuracy of identifying oil theft behaviors; it not only relies on video data, but also takes into account data on oil volume changes. By comparing the change pattern of oil volume under normal conditions with the pattern under detected abnormal conditions, it can more comprehensively assess whether there is a risk of oil theft. This method reduces the misjudgment rate that may be caused by a single data source; The data analysis and early warning module 22 analyzes the video data through the convolutional neural network model and the long short-term memory network model, identifies abnormal behavior, and comprehensively determines the oil theft risk in combination with the oil volume data, including the following steps: S1.1. Using a video dataset labeled with normal operation and oil theft, a convolutional neural network (CNN) model is trained to identify abnormal behavior (in this embodiment, the abnormal behavior is oil theft); S1.2. Use a convolutional neural network to extract features from each frame of video image (detailed steps: perform convolution operations through convolution kernels to extract local features of the image; use the pooling layer for downsampling to reduce the amount of calculation while maintaining important spatial information; use the fully connected layer to convert the features extracted by the convolution and pooling layers into a fixed-length feature vector); Furthermore, a convolutional neural network is used to extract features from each frame of video image, including the following steps: S1.21. Convolve the input image with multiple convolution kernels (filters) through convolution operations to extract local features in the image. Each convolution kernel is responsible for capturing different local features (such as edges, textures, etc.). The purpose of the convolution operation is to extract useful features from the input image, such as edges, textures and other local features. Each convolution kernel (filter) slides on the image and performs a convolution operation, which can identify specific types of patterns or features of the image. In this way, the convolution layer can capture the spatial hierarchy in the image and convert the original pixel information into a more representative and abstract feature expression. These features are crucial for subsequent tasks such as classification and detection. In order to solve the problem that oil thieves use strong direct light environments to interfere with monitoring, dynamic illumination compensation items are introduced in the convolution operation for optimization, and the correction item of the reflected light angle is introduced into the calculation of the illumination compensation item. The correction item of the reflected light angle is designed in combination with the relative position of the reflected light source, the normal direction of the tanker surface and the surface roughness, and the convolution operation is further optimized. Furthermore, the convolution operation is: ; in, Represents the output feature map after the convolution operation; Represents the input image In Location The pixel value at ; Indicates the current frame; Represents the convolution kernel, which represents the filter used in the convolution operation to extract spatial features in the image; Indicates the position of the current image; Indicates the horizontal offset of the convolution kernel; Indicates the vertical offset of the convolution kernel; Strong direct light may cause some areas in the image to be overexposed (white areas have no details at all), while other areas may become too dim (lose visible details). Strong or uneven lighting conditions may obscure or distort the true shape of the target, making it difficult for the convolutional neural network (CNN) to accurately extract the characteristics of the target (whether the oil tank has been illegally opened, whether oil theft has occurred, etc.). Oil theft is often sudden and may be carried out under unfavorable lighting conditions. Oil thieves may use the environment of strong direct light to cover up their behavior. For example, oil thieves may choose to operate during the period of direct sunlight, resulting in the monitoring screen being illuminated by strong light, affecting the accuracy of the monitoring system. After introducing the dynamic lighting compensation term, CNN can be optimized under different lighting environments to extract more stable features; In order to solve the problem that oil thieves use strong direct light to interfere with monitoring, the light intensity at each pixel position is calculated to compensate for the influence of light, the light intensity of each pixel value is normalized, and according to the calculated light intensity, a dynamic light compensation term is introduced in the convolution operation for optimization to deal with strong light interference: ; in, represents the result of the convolution operation after the dynamic lighting compensation term is introduced, and represents the output feature map after compensation; Indicates the current pixel position The local light intensity estimate at represents the light level of the area, which is calculated by the brightness of the local area; Represents a constant, used to avoid division by zero errors; Represents the weight coefficient of the illumination compensation item, which controls the influence of the compensation item; Represents the sigmoid function, which is used to map the illumination difference to a smooth range; Indicates the light threshold, which is the preset light intensity threshold used to trigger light compensation. When it is greater than this threshold, the light compensation item will be larger, reducing the impact of overexposure or direct strong light; The angle of reflected light directly affects the light intensity at a specific location, especially when there is a reflected light source on the surface of the tanker truck. The reflected light may cause abnormal light intensity in some areas, resulting in overexposure or uneven brightness distribution. When the angle between the reflected light source and the normal of the tanker truck surface changes, the intensity of the reflected light changes. For example, when the light source directly illuminates the surface, the reflection intensity is the strongest, and when the angle deviates from the normal, the reflected light intensity decreases. By correcting the reflected light angle, it can ensure that the convolution operation can more accurately understand the changes in light at different angles and avoid misjudgment of strong light interference caused by reflected light. The surface of the tanker truck may not be completely smooth, and its surface roughness will affect the distribution of reflected light. A rougher surface will scatter light, resulting in uneven reflected light. By considering the influence of surface roughness, the correction term can enable the convolutional neural network to more accurately capture the propagation effect of reflected light on different surfaces and reduce the interference of light on the image. The relative position of the incident light source and the reflected light source will directly affect the intensity of the surface reflected light. By introducing this factor, the convolution operation can automatically adjust its sensitivity to different light source positions, avoiding over-reliance on lighting changes in certain areas; the surface normal direction determines the incident angle of the reflected light, which in turn affects the intensity of the reflected light. By combining the normal direction with the reflected light angle, the convolutional neural network can more accurately restore the actual impact of the reflected light on the image; In view of the influence of the angle of reflected light on the light intensity, the correction term of the reflected light angle is introduced into the calculation of the light compensation term. The correction term of the reflected light angle is designed in combination with the relative position of the reflected light source, the normal direction of the tanker surface and the surface roughness to further optimize the convolution operation: ; ; in, The correction term representing the reflected light angle is the result of the convolution operation after introducing the illumination compensation term; The correction term for the reflected light angle takes into account the effect of the reflection angle between the light source and the surface of the tanker on the local illumination; It represents the cosine value of the angle between the incident light and the surface normal of the tanker truck. This value reflects the angle of the incident light source relative to the surface normal. The smaller the angle (i.e. the more direct the light), the stronger the intensity of the reflected light and the larger the cosine value. Represents the angle between the incident light direction and the surface normal; It represents the surface roughness correction factor, which indicates the effect of the roughness of the tanker truck surface on the intensity of reflected light. A rough surface will scatter more light, resulting in a decrease in the intensity of reflected light, while a smooth surface will strongly reflect light. Indicates the distance between the surface of the tanker and the reflected light source. This means the distance from a certain position on the surface of the tanker to the reflected light source. The farther the distance, the weaker the intensity of the reflected light is usually. Therefore, the distance affects the intensity of the reflected light and the distribution of the illumination. Indicates the maximum impact distance; S1.22. Apply the ReLU activation function to the output of the convolution operation. ReLU can make the network model nonlinear and help the network learn more complex features. S1.23. After the convolutional layer, the pooling operation is performed by the maximum pooling method. The maximum pooling method reduces the spatial size of the feature map by taking the maximum value in the local area while retaining the key features. The purpose of the pooling operation is to reduce the size of the feature map and the amount of calculation while retaining important feature information. The pooling layer reduces the size of the feature map, reduces the computational complexity, and increases the robustness of the network to the spatial position. S1.24, the output layer classifies the behavior according to the result of feature extraction, and determines whether there is oil theft through classification: ; in, Indicates that given an input image The probability of identifying oil theft behavior; represents the activation function; represents the weight of the fully connected layer; represents the bias term of the fully connected layer; Represents the vector after flattening the feature map after convolution and pooling; Represents the input image.

[0020] S1.3. The features of each frame extracted from the convolutional neural network are used as the input of the long short-term memory network model (LSTM) (LSTM controls the transmission and forgetting of information through an internal gating mechanism); S1.4, sending the output of the long short-term memory network model (LSTM) to the fully connected layer to predict whether the current video image frame has abnormal behavior; S1.5. Real-time monitoring of oil volume changes through sensors, forming a time series of oil volume data , indicating the amount of oil in the tank at each moment; S1.6. Determine whether a sudden drop has occurred by calculating the change in oil volume; S1.7, combining the oil volume change and the abnormal behavior in the video, calculating the comprehensive risk score and generating early warning instructions; The oil volume change and the abnormal behavior in the video are combined to calculate the comprehensive risk score, which includes the following steps: S1.71. Combining the oil volume change and the video abnormal behavior score to form a weighted score. For the oil volume change of the tanker during movement, the acceleration and speed of the tanker are introduced into the weighted score to distinguish between normal oil volume fluctuations caused by movement and oil theft; Furthermore, the weighted score is: ; in, Represents the risk score, which is used to measure the current moment Is there a risk of oil theft? Indicates the weight of video data in the comprehensive risk score, that is, the degree of influence of video data (abnormal behavior detection) on the overall risk judgment; Indicates that given video data The probability of oil theft under the condition of , the higher the probability value, the greater the risk of oil theft; Indicates the weight coefficient of oil volume change data in the comprehensive risk score; Represents video data; Indicates changes in oil volume; Indicates the maximum capacity of the tanker truck's fuel tank; Indicates time; During the transportation of the tanker, the movement of the vehicle will cause the oil to shift in the tank. This physical phenomenon may cause changes in the oil volume reading. If these normal oil volume fluctuations caused by acceleration and speed changes are not taken into account, relying only on the data of the oil volume sensor to determine whether there is oil theft may result in a high false alarm rate. By combining the movement state (acceleration and speed) of the tanker, the system can more accurately distinguish these normal oil volume fluctuations caused by movement from the oil volume changes caused by actual oil theft. In view of the change in oil volume during the movement of the tanker, the acceleration and speed of the tanker are introduced into the weighted score to distinguish between normal oil volume fluctuations caused by movement and oil theft: ; in, represents the risk score after introducing the acceleration and speed of the tanker truck; Indicates the adjustment coefficient, which is used to control the effect of the motion state on the change of oil volume; represents the acceleration of the tanker truck; Indicates the speed of the tanker truck; Considering that wind resistance, tire pressure and road slope will affect the acceleration of the tanker truck during its movement, the weighted score is further optimized (strong wind or sudden wind speed changes will increase the resistance of the vehicle, causing fluctuations in the acceleration or deceleration of the tanker truck, which may appear as sudden changes in the oil volume in the oil volume monitoring data. If the effect of wind speed on vehicle acceleration is not considered, the oil volume fluctuation caused by wind speed changes may be misjudged as oil theft; the tire pressure of the tanker truck directly affects the rolling resistance of the tire. If the tire pressure is too low, the rolling resistance of the tire will increase, resulting in abnormal acceleration changes during vehicle driving. This change will affect the movement state of the tanker truck, resulting in oil volume fluctuations. If the effect of tire pressure is not considered, Acceleration changes caused by improper tire pressure may be misjudged as oil theft; when a tanker truck is traveling uphill or downhill, the acceleration of the vehicle will change due to gravity. For example, when going uphill, the acceleration of the tanker truck will decrease; while when going downhill, the acceleration may increase. If the influence of the road slope is not considered, these normal oil volume fluctuations will be misjudged as oil theft; by introducing the contribution of wind resistance, tire pressure and slope to acceleration, the traditional single acceleration is expanded to a comprehensive acceleration coupled with multiple factors, solving the problem of misjudgment caused by external environmental interference (including oil volume fluctuations caused by strong winds or slopes); dynamically adjust the rolling resistance coefficient according to the real-time tire pressure to avoid the sudden increase in rolling resistance when the tire pressure is insufficient and being misjudged as oil theft): ; The acceleration Decomposed into multiple components, including: tanker truck acceleration , wind resistance acceleration , the rolling resistance acceleration affected by tire pressure , road slope acceleration ; Among them, wind resistance acceleration for: ; In the formula, is the air density, is the drag coefficient, is the windward area, The quality of the tanker truck; Tire pressure affects rolling resistance acceleration for: ; In the formula, Indicates the current tire pressure The rolling resistance coefficient under is the coefficient under standard tire pressure, is the acceleration due to gravity; Road Slope Acceleration for: ; In the formula, is the slope angle; S1.72. Calculate the cumulative risk score based on the oil volume changes and video abnormal behaviors at multiple time points through sliding window technology. The sliding window technology performs rolling analysis on the oil volume changes and video abnormal behavior data within a fixed time period, and cumulatively calculates a score reflecting the degree of oil theft risk, which helps reduce false alarms and improve the accuracy of the system; S1.73. Calculate the cumulative risk score. When the cumulative risk score exceeds the preset threshold Rt, generate an early warning instruction.

[0021] The early warning response unit 3 receives the early warning instruction and triggers the alarm system; In this embodiment, the early warning response unit 3 includes an instruction sending module 31 and an early warning module 32; The instruction sending module 31 receives the generated warning instruction and sends the warning instruction to the warning module 32; The warning module 32 receives the warning instruction and triggers an alarm. The alarm takes various forms, including but not limited to sound alarm (such as alarm bell), visual alarm (such as flashing warning light), SMS or email notification to relevant personnel.

[0022] Embodiment 2: Embodiment 2 of the present invention differs from Embodiment 1 in that this embodiment introduces an anti-oil theft warning method used in an anti-oil theft warning system for oil tankers based on artificial intelligence.

[0023] An artificial intelligence-based anti-oil theft warning method for oil tankers, used in the artificial intelligence-based anti-oil theft warning system for oil tankers, comprises the following steps: S2.1, real-time monitoring of oil level height through ultrasonic level sensor; S2.2. Identify abnormal behaviors by collecting and analyzing video data, and comprehensively judge the risk of oil theft in combination with oil volume data. Generate early warning instructions based on the risk of oil theft. Consider the impact of the angle of reflected light on the light intensity during the judgment process. Design the correction item of the reflected light angle in combination with the relative position of the reflected light source, the normal direction of the surface of the tanker truck, and the surface roughness. S2.3. Receive early warning instructions and trigger the alarm system.

[0024] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments, and the above embodiments and descriptions are only preferred examples of the present invention, and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, and these changes and improvements all fall within the scope of the present invention to be protected.

Claims

1. An artificial intelligence-based anti-oil theft warning system for oil tankers, characterized in that: include: An oil level real-time monitoring unit (1), wherein the oil level real-time monitoring unit (1) monitors the oil level in real time through an ultrasonic liquid level sensor; A data acquisition and processing unit (2), wherein the data acquisition and processing unit (2) identifies abnormal behavior by acquiring and analyzing video data, and comprehensively determines the risk of oil theft in combination with the oil volume data, generates an early warning instruction according to the oil theft risk, considers the influence of the angle of reflected light on the light intensity during the judgment process, and designs the correction item of the reflected light angle in combination with the relative position of the reflected light source, the normal direction of the surface of the tanker truck, and the surface roughness during the judgment process; An early warning response unit (3), wherein the early warning response unit (3) receives an early warning instruction and triggers an alarm system.

2. The artificial intelligence-based anti-oil theft warning system for tank trucks according to claim 1 is characterized by: The oil quantity real-time monitoring unit (1) comprises an oil level monitoring module (11) and an oil quantity conversion module (12); The oil level monitoring module (11) monitors the oil level in real time through an ultrasonic level sensor installed at the bottom of the oil tank; The oil quantity conversion module (12) converts the oil level height data into oil quantity by means of a geometric formula method.

3. The artificial intelligence-based anti-oil theft warning system for tank trucks according to claim 2 is characterized by: The data acquisition and processing unit (2) comprises a data acquisition module (21) and a data analysis and early warning module (22); The data acquisition module (21) collects video data in real time through a camera at the oil tank port of the oil tank truck; The data analysis and early warning module (22) analyzes the video data through a convolutional neural network model and a long short-term memory network model, identifies abnormal behavior, and comprehensively determines the risk of oil theft in combination with the oil volume data.

4. The artificial intelligence-based anti-oil theft warning system for tank trucks according to claim 3 is characterized by: The data analysis and early warning module (22) analyzes the video data through a convolutional neural network model and a long short-term memory network model, identifies abnormal behavior, and comprehensively determines the oil theft risk in combination with the oil volume data, including the following steps: S1.

1. Use the video dataset with labeled normal operation and oil theft behavior to train the convolutional neural network model to identify abnormal behavior. S1.2, use convolutional neural network to extract features from each frame of video image; S1.3, taking the features of each frame extracted from the convolutional neural network as the input of the long short-term memory network model; S1.4, sending the output of the long short-term memory network model to the fully connected layer to predict whether there is abnormal behavior in the current video image frame; S1.

5. Real-time monitoring of oil volume changes through sensors, forming a time series of oil volume data , indicating the amount of oil in the tank at each moment; S1.

6. Determine whether a sudden drop has occurred by calculating the change in oil volume; S1.

7. Combine the oil volume change and abnormal video behavior to calculate the comprehensive risk score and generate early warning instructions.

5. The artificial intelligence-based anti-oil theft warning system for tank trucks according to claim 4 is characterized by: In S1.2, a convolutional neural network is used to extract features from each frame of video image, including the following steps: S1.

21. Convolve the input image with multiple convolution kernels through a convolution operation. To address the problem of oil thieves interfering with monitoring by using strong direct light, a dynamic illumination compensation term is introduced into the convolution operation for optimization. The correction term of the reflected light angle is introduced into the calculation of the illumination compensation term. The correction term of the reflected light angle is designed in combination with the relative position of the reflected light source, the normal direction of the surface of the tanker truck, and the surface roughness, to further optimize the convolution operation. S1.

22. Apply the ReLU activation function to the output of the convolution operation. S1.23, after the convolutional layer, a pooling operation is performed using the maximum pooling method; S1.24, the output layer classifies the behavior according to the result of feature extraction, and determines whether there is oil theft through classification: ； in, Indicates that given an input image The probability of identifying oil theft behavior; represents the activation function; represents the weight of the fully connected layer; represents the bias term of the fully connected layer; Represents the vector after flattening the feature map after convolution and pooling; Represents the input image.

6. The artificial intelligence-based anti-oil theft warning system for tank trucks according to claim 5 is characterized by: In S1.21, the convolution operation is: ； in, Represents the output feature map after the convolution operation; Represents the input image In Location The pixel value at ; Indicates the current frame; represents the convolution kernel; Indicates the position of the current image; Indicates the horizontal offset of the convolution kernel; Indicates the vertical offset of the convolution kernel; To solve the problem of oil thieves interfering with monitoring by using strong direct light, the light intensity at each pixel position is calculated, the light intensity of each pixel value is normalized, and a dynamic light compensation term is introduced in the convolution operation for optimization based on the calculated light intensity: ； in, Represents the result of the convolution operation after the dynamic lighting compensation term is introduced; Indicates the current pixel position The estimated local light intensity at ; represents a constant; Represents the weight coefficient of the lighting compensation item; Represents the sigmoid function; represents the light threshold; In view of the influence of the angle of reflected light on the light intensity, the correction term of the reflected light angle is introduced into the calculation of the light compensation term. The correction term of the reflected light angle is designed in combination with the relative position of the reflected light source, the normal direction of the tanker surface and the surface roughness to further optimize the convolution operation: ； ； in, The correction term representing the reflected light angle is the result of the convolution operation after introducing the illumination compensation term; A correction term for the angle of reflected light; represents the cosine of the angle between the incident light and the normal line of the tanker surface; Represents the angle between the incident light direction and the surface normal; Indicates the surface roughness correction factor; Indicates the distance between the surface of the tanker and the reflected light source; Indicates the maximum influence distance.

7. The artificial intelligence-based anti-oil theft warning system for tank trucks according to claim 6 is characterized by: In S1.7, the oil volume change and the abnormal behavior in the video are combined to calculate the comprehensive risk score, including the following steps: S1.

71. Combining the oil volume change and the video abnormal behavior score to form a weighted score. For the oil volume change of the tanker during movement, the acceleration and speed of the tanker are introduced into the weighted score to distinguish between normal oil volume fluctuations caused by movement and oil theft; S1.72, calculate the cumulative risk score based on the oil volume changes and abnormal video behaviors at multiple time points through sliding window technology; S1.

73. Calculate the cumulative risk score. When the cumulative risk score exceeds the preset threshold Rt, generate an early warning instruction.

8. The artificial intelligence-based anti-oil theft warning system for tank trucks according to claim 7 is characterized by: In S1.71, the weighted score is: ； in, represents the risk score; Indicates the weight of video data in the comprehensive risk score; Indicates that given video data The probability of oil theft under the condition of Indicates the weight coefficient of oil volume change data in the comprehensive risk score; Represents video data; Indicates changes in oil volume; Indicates the maximum capacity of the tanker truck's fuel tank; Indicates time; In view of the change in oil volume during the movement of the tanker, the acceleration and speed of the tanker are introduced into the weighted score to distinguish between normal oil volume fluctuations caused by movement and oil theft: ； in, represents the risk score after introducing the acceleration and speed of the tanker truck; represents the adjustment coefficient; represents the acceleration of the tanker truck; Indicates the speed of the tanker truck.

9. The artificial intelligence-based anti-oil theft warning system for tank trucks according to claim 8 is characterized by: The early warning response unit (3) comprises an instruction sending module (31) and an early warning module (32); The instruction sending module (31) receives the generated warning instruction and sends the warning instruction to the warning module (32); The early warning module (32) receives the early warning instruction and triggers an alarm.

10. An artificial intelligence-based anti-oil theft warning method for oil tanker trucks, used in the artificial intelligence-based anti-oil theft warning system for oil tanker trucks as claimed in any one of claims 1 to 9, characterized in that: The steps include: S2.1, real-time monitoring of oil level height through ultrasonic level sensor; S2.

2. Identify abnormal behaviors by collecting and analyzing video data, and comprehensively judge the risk of oil theft in combination with oil volume data. Generate early warning instructions based on the risk of oil theft. Consider the impact of the angle of reflected light on the light intensity during the judgment process. Design the correction item of the reflected light angle in combination with the relative position of the reflected light source, the normal direction of the surface of the tanker truck, and the surface roughness. S2.

3. Receive early warning instructions and trigger the alarm system.

Images