An intelligent camera control method and control system based on the Internet of Things

Abstract

The present invention discloses an intelligent camera control method and control system based on the Internet of Things, which relates to the technical field of camera control and includes the following steps: obtaining polarized reflection light images and thermal imaging data respectively through a low-light sensor and an infrared sensor working synchronously, and dynamically adjusting the fusion weight coefficient of thermal imaging and visible light images; deploying a Bayesian neural network at an edge computing node, extracting spatio-temporal features through a multi-channel feature map and a convolutional module, and updating the weights using the MCMC algorithm to optimize the target position prediction; compensating for the target position offset when the intersection over union (IoU) decrease rate between consecutive frames of the target bounding box exceeds a threshold; dynamically switching the camera working mode based on the target tracking confidence quantile; in a haze environment, generating a target trajectory through thermal imaging data and reducing positioning noise using a time series window; the present invention significantly improves the accuracy and robustness of target tracking.

Description

Technical Field

[0001] The present invention relates to the technical field of camera control, and specifically to an intelligent camera control method and control system based on the Internet of Things. Background Art

[0002] With the continuous development and popularization of Internet of Things technology, more and more intelligent devices have started to be connected to the Internet of Things, including intelligent cameras. Intelligent cameras are gradually developing towards the direction of intelligence and networking. So far, intelligent cameras can already form a network, such as direct connection through PC5 connection. In this way, multiple intelligent cameras can cooperate for video acquisition and processing through network transmission, such as fusing the videos collected by multiple intelligent cameras into one for playback, and the user experience is also better.

[0003] However, traditional target tracking algorithms mostly rely on single-sensor data (such as infrared or visual images), and are prone to target loss in the case of drastic environmental changes (such as occlusion, sudden light changes, etc.). In order to improve the tracking accuracy, it is usually necessary to maintain high frame rate and high resolution modes, which leads to a significant increase in power consumption. Existing technologies mostly deal with the occlusion problem through some simple filtering or compensation methods, but most methods have limited processing effects on complex environments (such as haze, strong light reflection, etc.). Summary of the Invention

[0004] To solve the defects existing in the prior art, the present invention provides an intelligent camera control method and control system based on the Internet of Things.

[0005] To solve the above technical problems, the present invention provides the following technical solutions:

[0006] The present invention provides an intelligent camera control method based on the Internet of Things, including the following steps:

[0007] S1: Multi-modal sensor data fusion acquisition, obtaining polarized reflection light images and thermal imaging data of the monitoring scene through a low-light sensor and an infrared sensor working synchronously respectively;

[0008] S2: Dual-modal feature correlation modeling, dividing the temperature topology partition of the thermal imaging data according to the pixel grid. When the absolute value of the average temperature difference between adjacent partitions exceeds the threshold, a dynamic target tracking signal is triggered, and the polarized angle parameter of the material surface is analyzed for the polarized reflection light image. The support vector machine classification algorithm is used to identify the material type, and a temperature-material feature correlation database is established;

[0009] S3: Dynamic weighted image fusion, dynamically adjusting the fusion weight coefficient of the thermal imaging and visible light images according to the light intensity change rate parameter monitored by the ambient light sensor in real time;

[0010] S4: Probabilistic inference network processing. Deploy a Bayesian neural network at the edge computing node, extract spatio-temporal features through multi-channel feature maps and convolutional modules, and use the MCMC algorithm to update the weights to optimize the target position prediction;

[0011] S5: Occlusion compensation and trajectory correction. When the intersection over union (IoU) decrease rate between consecutive frames of the target bounding box exceeds the threshold, call the historical temperature data to construct a Markov random field model, calculate the probability distribution of candidate regions based on the temperature spatial correlation, and compensate for the target position offset;

[0012] S6: Energy efficiency adaptive regulation. Dynamically switch the camera working mode based on the target tracking confidence quantile;

[0013] S7: Multi-node collaborative verification. In a haze environment, generate the target trajectory through thermal imaging data, use the time series window to reduce the positioning noise, and upload the positioning results and parameters to the cloud platform through the MQTT protocol.

[0014] As a preferred technical solution of the present invention, the fusion formula in step S3 is:

[0015] ;

[0016] where α represents the dynamic adjustment coefficient of environmental illumination, T thermal (x, y) is the temperature value of the thermal imaging image at the position (x, y), and I visual (x, y) is the pixel value of the visual image at the position (x, y).

[0017] As a preferred technical solution of the present invention, step S4 specifically includes the following steps:

[0018] S41: Deploy a four-layer Bayesian neural network at the edge computing node. The input layer is connected to a four-channel feature map, including the thermal imaging gradient map, the material reflection feature map, the motion vector map, and the environmental illumination parameters;

[0019] S42: The residual convolutional layer extracts spatio-temporal features, the fully connected layer generates the posterior probability distribution of the target position, and the output layer updates the weight parameters based on Markov chain Monte Carlo sampling;

[0020] S43: Perform posterior probability distribution inference through the edge computing device. The posterior probability distribution inference formula is as follows:

[0021] ;

[0022] where ω represents the network weights, D represents the training data, p(w∣D) represents the posterior probability distribution, p(D∣w) represents the likelihood function, and p(w) represents the prior distribution.

[0023] As a preferred technical solution of the present invention, the weight update of the Bayesian neural network follows the following formula:

[0024] ;

[0025] where ω t represents the node weight value at the t-th iteration, ε represents the learning rate, and P(D∣w t ) is the likelihood probability of the training data under the current parameters.

[0026] As a preferred technical solution of the present invention, the fusion result F fusion (x, y) of step S1 is used as the input of the lightweight Bayesian neural network for real-time inference of target tracking. The Bayesian neural network infers through the posterior probability distribution and outputs the probability distribution of the target position:

[0027] ;

[0028] where W 1 , W 2 , W 3 , W 4 are the weight matrices of the Bayesian neural network, ReLU is the activation function, and softmax is used for probability normalization.

[0029] As a preferred technical solution of the present invention, the weights of the Markov random field model in step S5 are defined as follows:

[0030] ;

[0031] where T R is the average temperature of the candidate region, T hist is the mean value of the target historical temperature, ρ is the spatial correlation factor, and σ is the standard deviation of the temperature distribution.

[0032] As a preferred technical solution of the present invention, in step S6, when the confidence quantile is lower than 0.9, the high-definition mode is enabled, and when the confidence quantile is higher than 0.9, the low-power mode is switched to, and at the same time, the power supply module of the unactivated sensor is turned off.

[0033] As a preferred technical solution of the present invention, the uploaded data transmission in step S7 uses the AES-256 encryption algorithm, and the key pair is regenerated every 15 seconds.

[0034] The present invention also provides an intelligent camera control system, and the control system includes a dual-spectrum sensing module, an edge computing unit, an IoT communication module, and a dynamic power management unit;

[0035] The dual-spectrum sensing module integrates a low-light sensor and an infrared sensor, and is built-in with a weighted fusion preprocessing chip;

[0036] The edge computing unit is built-in with a programmable logic array for performing Bayesian neural network inference, and the storage module contains a temperature-material feature association database;

[0037] The IoT communication module supports a dual-protocol switching module for the MQTT protocol and the LoRaWAN protocol, and real-time uploads target location data and environmental parameters;

[0038] The dynamic power management unit dynamically adjusts the power consumption of the sensor and the main frequency of the computing unit according to the operating mode.

[0039] As a preferred technical solution of the present invention, the edge computing unit is equipped with a hardware acceleration module and supports simultaneous floating-point operations of four-channel feature maps. The IoT communication module includes a data encryption sub-module, and the dynamic power management unit integrates an adaptive voltage regulator.

[0040] The beneficial effects of the present invention are:

[0041] 1. Even in complex environments (such as light changes or occlusions), the method of the present invention significantly improves the accuracy and robustness of target tracking through multi-modal data fusion, dynamic weighted image fusion, and occlusion compensation.

[0042] 2. In the present invention, through intelligent energy efficiency adaptive regulation, the working mode of the camera is dynamically adjusted to ensure that the high frame rate / high resolution mode is only used when necessary, thereby significantly reducing the power consumption of the device while ensuring tracking accuracy.

[0043] 3. In the present invention, through multi-node collaborative verification and real-time cloud data update, the reliability of the system in harsh environments is enhanced, and different complex situations can be effectively dealt with. Description of the Drawings

[0044] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the present invention, and do not constitute a limitation to the present invention. In the drawings:

[0045] Figure 1 It is a schematic flowchart of the intelligent camera control method of the present invention. Detailed Embodiments

[0046] The following describes the preferred embodiments of the present invention with reference to the drawings. It should be understood that the preferred embodiments described herein are only used to explain and illustrate the present invention, and are not used to limit the present invention.

[0047] Example 1: As Figure 1As shown in the figure, an intelligent camera control method based on the Internet of Things includes the following steps:

[0048] S1: Multi-modal sensor data fusion acquisition. The polarized reflected light image and thermal imaging data of the monitoring scene are respectively obtained through a low-light sensor (covering the visible light to near-infrared band) and an infrared sensor (generally using a low-power uncooled long-wave infrared sensor with a wavelength of 8-14 μm) that work synchronously. The low-light sensor supports the analysis of material reflectivity in low-light scenarios, such as distinguishing the types of clothing fibers or the characteristics of vehicle surface coatings.

[0049] S2: Dual-modal feature correlation modeling. The temperature topology partition is divided for the thermal imaging data according to a (preferably 16×16) pixel grid. When the absolute value of the average temperature difference between adjacent partitions exceeds a threshold (such as 3°C), a dynamic target tracking signal is triggered, and the polarization angle parameter of the material surface is analyzed for the polarized reflected light image. The support vector machine classification algorithm is used to identify the material type, and a temperature-material feature correlation database is established. By dividing the temperature topology partition and calculating the temperature difference between adjacent partitions, moving targets can be dynamically identified and tracked.

[0050] S3: Dynamic weighted image fusion. According to the light intensity change rate parameter real-time monitored by the ambient light sensor, the fusion weight coefficient of the thermal imaging and visible light images is dynamically adjusted.

[0051] S4: Probabilistic inference network processing. A Bayesian neural network is deployed at the edge computing node. Spatiotemporal features are extracted through multi-channel feature maps and convolutional modules, and the MCMC algorithm is used to update the weights to optimize the target position prediction.

[0052] S5: Occlusion compensation and trajectory correction. When the intersection over union (IoU) decline rate between consecutive frames of the target bounding box exceeds a threshold (50-85%, preferably 75%), the historical temperature data is called to construct a Markov random field model, and the candidate region probability distribution is calculated based on the temperature spatial correlation to compensate for the target position offset. By calculating the change in the IoU decline rate between consecutive frames of the target bounding box, the target occlusion phenomenon can be detected and corrected. The Markov random field model can use historical temperature data and spatial correlation to infer the potential position of the target when the target is occluded, and compensate for the target position offset through the data of multi-modal sensors (such as infrared sensors and low-light sensors). In this way, even if the target is briefly occluded, the system can maintain a high tracking accuracy.

[0053] S6: Energy efficiency adaptive regulation. The camera working mode is dynamically switched based on the target tracking confidence quantile.

[0054] S7: Multi-node collaborative verification. In a haze environment, the target trajectory is generated through thermal imaging data, the timing window is used to reduce positioning noise, and the positioning results and parameters are uploaded to the cloud platform through the MQTT protocol. This multi-node collaboration can effectively reduce the error of a single camera in complex scenes and further improve the accuracy and stability of target tracking.

[0055] Among them, the polarized reflected light image and thermal imaging data of the target are obtained through the synchronous working of the low-light sensor and infrared sensor. The weighted adaptive fusion algorithm is used to dynamically adjust the fusion weight of thermal imaging and visual data according to the changes in ambient light, thereby enhancing the target tracking capability in complex environments (such as low light, haze, etc.) and reducing the target loss caused by sudden changes in light.

[0056] When the ambient light intensity changes, the fusion weight coefficient of thermal imaging and visible light images is dynamically adjusted. This ensures that the camera can always provide the best target recognition capability under different lighting conditions, thereby reducing target tracking failures caused by instantaneous lighting changes;

[0057] When the target encounters occlusion (for example, when the bounding box intersection-to-union ratio drops below a threshold), occlusion compensation is performed using historical temperature data and the Markov random field model. The model uses the temperature spatial correlation of the target to predict and correct the target position offset, maintain continuous tracking of the target, and avoid tracking loss due to short-term occlusion.

[0058] The camera working mode is dynamically switched based on the confidence quantile of target tracking. When the confidence of the target is lower than a certain threshold, the system automatically switches to high-definition mode to improve accuracy; when the confidence is high, it switches to low-power mode to effectively reduce device power consumption and extend usage time;

[0059] In haze environments, thermal imaging data is used to generate target trajectories, and timing windows are used to reduce positioning noise to further improve tracking accuracy in non-ideal environments. This is achieved by updating and adjusting resolution and alarm threshold parameters in real time on the cloud platform to ensure the stability of the system under various environmental conditions.

[0060] Furthermore, the fusion weight of thermal imaging and visual data is dynamically adjusted according to the change of ambient light by weighted averaging. The fusion formula of step S3 is:

[0061] ;

[0062] Among them, α is the dynamic adjustment coefficient of ambient light, and its value is adjusted according to the change of ambient light. Generally speaking, the range of α is 0.3~0.7. thermal (x, y) is the temperature value of the thermal imaging image at position (x, y), I visual(x, y) is the pixel value of the visual image at position (x, y).

[0063] For example, when the material identified in step S2 is metal, the thermal imaging weight α in step S3 is forced to increase by 0.15 to compensate for the imaging deviation caused by the heat absorption of the metal.

[0064] Further, step S4 specifically includes the following steps:

[0065] S41: Deploy a four-layer Bayesian neural network at the edge computing node. The four-layer Bayesian neural network includes an input layer, a residual convolutional layer, a fully connected layer, and an output layer. The input layer is connected to a four-channel feature map, including a thermal imaging gradient map, a material reflection feature map, a motion vector map, and environmental light parameters; the thermal imaging gradient map is a composite map of horizontal and vertical gradients extracted from thermal imaging data using the Sobel operator, the material reflection feature map is a grayscale image generated after normalizing the polarization angle data collected by a low-light sensor, the motion vector map is a pixel displacement vector field calculated based on the adjacent frame difference method, and the environmental light parameter is the real-time reading of a visible light sensor in lux;

[0066] These feature maps provide the key information required for the camera to perform target tracking in complex environments, including temperature changes, material surface information, target motion states, and surrounding light conditions. Through multi-channel input, the fusion of multiple types of information can be achieved, reducing the errors caused by a single data source and improving the accuracy of target recognition and tracking.

[0067] S42: The residual convolutional layer extracts spatio-temporal features. The convolutional operation is used to extract the spatio-temporal information in the multi-channel features, further improving the expression ability of the feature map. Through the residual structure, the problem of gradient disappearance or explosion during the training process can be effectively alleviated, improving the stability and performance of network training. The fully connected layer generates the posterior probability distribution of the target position. The output layer updates the weight parameters based on the Markov chain Monte Carlo sampling. At the output layer of the Bayesian neural network, through the Markov chain Monte Carlo (MCMC) sampling algorithm, the posterior probability distribution of the target position is updated. This algorithm infers the target position probability under the given data based on Bayes' theorem, thus avoiding target loss in complex environments. Combining the posterior probability distribution of Bayesian inference, the target position can be dynamically updated according to the current observation data, solving the problem of target loss caused by occlusion, sudden light changes, etc. Especially in the case of occlusion and sudden light changes, Bayesian inference can reasonably infer the possible position of the target, helping to maintain the stability and accuracy of tracking;

[0068] S43: Perform posterior probability distribution inference through the edge computing device. The posterior probability distribution inference formula is as follows:

[0069] ;

[0070] Among them, ω represents the network weight, D represents the training data, p(w∣D) represents the posterior probability distribution, p(D∣w) represents the likelihood function, and p(w) represents the prior distribution;

[0071] The weight update of the Bayesian neural network follows the following formula:

[0072] ;

[0073] Among them, ω t represents the node weight value at the t-th iteration, ε represents the learning rate, and P(D∣w t ) is the likelihood probability of the training data under the current parameters;

[0074] This formula means using the gradient ascent method to update the network weight. At each iteration, the network weight is adjusted according to the gradient of the loss function to maximize the likelihood function P(D∣w t ) given the current dataset D. In this way, the network gradually learns how to more accurately predict the position of the target. Especially when performing target tracking in a complex environment, it can reduce the risk of target loss.

[0075] In practical applications, due to the interference of factors such as environmental light and occlusion, the appearance and motion state of the target may change suddenly. The Bayesian neural network uses the above weight update formula to dynamically adjust the posterior probability distribution of the target in the prediction of each frame of image, thereby reducing the target positioning error caused by environmental changes. Through repeated updates, the network can adapt to different environmental conditions and stabilize the target tracking process;

[0076] In addition, through the adaptive inference mechanism of the Bayesian neural network, combined with environmental features (such as light, motion state, etc.), the working mode of the camera can be dynamically adjusted on the premise of ensuring accuracy. For example, under good lighting conditions, the resolution or frame rate can be appropriately reduced to reduce power consumption; in the case of sudden light changes or target occlusion, the working accuracy can be increased in a timely manner to ensure the stability of tracking.

[0077] The above solution combines the advantages of the Bayesian neural network and the MCMC algorithm, providing a robust solution for target tracking of intelligent cameras in complex environments. Especially its ability to dynamically adaptively adjust weights and optimize the target prediction position can not only improve the target tracking accuracy but also avoid the device's excessive dependence on high-power consumption modes when the environment changes, thereby achieving energy efficiency optimization.

[0078] Furthermore, the fusion result F of the step S1 fusion(x, y) is used as the input of the lightweight Bayesian neural network for real-time inference of target tracking. The Bayesian neural network infers through the posterior probability distribution and outputs the probability distribution of the target position:

[0079] ;

[0080] Among them, W 1 , W 2 , W 3 , W 4 are the weight matrices of the Bayesian neural network, ReLU is the activation function, and softmax is used for probability normalization.

[0081] The lightweight Bayesian neural network can infer the target position through the posterior probability distribution. The input undergoes multiple weight matrix transformations (such as being processed by the ReLU activation function and then passed to different layers), and finally outputs the probability distribution of the target position. In this process, the core idea of the Bayesian neural network is to update the probability distribution of the target position through Bayesian inference, and this distribution can dynamically adjust the prediction of the target position according to different input features.

[0082] In this way, even in complex or dynamic environments, the camera can still maintain a high target tracking accuracy. Due to the adoption of the fused image input, the Bayesian neural network can quickly adapt and adjust the prediction under different environmental conditions (such as when there is occlusion or sudden light change), thereby reducing the high frame rate requirements due to target loss.

[0083] This method can not only improve the tracking accuracy but also effectively reduce the device power consumption and avoid the resource waste caused by high frame rates and high resolutions.

[0084] Further, the weights of the Markov random field model in step S5 are defined as follows:

[0085] ;

[0086] Among them, T R is the current average temperature of the candidate region R, T hist is the mean value of the target historical temperature, ρ is the spatial correlation factor, and the range of ρ is 0.5 - 1.2. σ is the standard deviation of the temperature distribution, and the temperature standard deviation is dynamically calculated through 30 consecutive frames of data. It is usually used to balance the influence of temperature differences. exp is the exponential function, which is used to adjust the weights according to the magnitude of the temperature difference. The larger the temperature difference, the smaller the exponential value, indicating that the weight of this position is lower; the smaller the temperature difference, the larger the exponential value, indicating that the weight of this position is higher;

[0087] The Markov random field model is used to model the skin temperature distribution of the target. By considering the temperature changes at different positions on the target surface, the candidate positions of the target are inferred. In this embodiment, the Markov random field model combines spatio-temporal correlation to generate a candidate position probability map, and the Pearson correlation coefficient is introduced to measure the linear relationship between the temperature at the current position and the target historical temperature data;

[0088] For example, if the Pearson correlation coefficient between the temperature feature and the historical data is ≥ 0.85, it indicates a strong positive correlation between them, and ∣T R -T hist ∣ (temperature difference) will be small, indicating that the temperature change trend in this area is consistent with the historical data. At this time, it is necessary to increase the weight of this area to improve the prediction accuracy, and the current weight W(R) will be multiplied by an additional coefficient of 1.2.

[0089] Further, in step S6, when the confidence quantile is lower than 0.9 (when the target is clear and reliably tracked), the high-definition mode (higher frame rate and resolution) is enabled to ensure that high-quality video data can still be obtained in a complex environment and increase the accuracy of target recognition;

[0090] When the confidence quantile is higher than 0.9 (when the system detects that the target is lost or uncertain), it switches to the low-power mode (reducing the frame rate or resolution), and at the same time turns off the unactivated sensor power supply module, thereby reducing the power consumption of the device. And in this mode, the processor operating current does not exceed 1.2A, and the sensor sampling frequency is reduced to 60% of the reference value.

[0091] Through this strategy, the system can adjust the working mode according to actual needs to save battery life or extend the working time of the device, especially when continuous high frame rate / high resolution images are not required.

[0092] It should be noted that in order to further improve the performance of the system, hardware acceleration can be combined. For example, an FPGA or GPU can be used to accelerate the calculation of the target tracking algorithm, which is particularly important for real-time processing at high frame rates and high resolutions, and can significantly improve the algorithm processing efficiency and reduce power consumption.

[0093] Further, the upload data transmission in step S7 uses the AES-256 encryption algorithm, and a key pair is regenerated every 15 seconds.

[0094] To ensure the security of the transmitted data, all data uploaded to the cloud through the MQTT protocol uses a strong encryption algorithm - AES-256. AES-256 uses a 256-bit key length, which can effectively prevent data from being eavesdropped or tampered with, ensure that there are no security vulnerabilities in the data transmission process, and is suitable for the Internet of Things environment with high security requirements.

[0095] Dynamic Key Update Mechanism: To further enhance the security of the system, a mechanism is designed to regenerate the key pair every 15 seconds. This mechanism ensures that even if a malicious attacker obtains the current key, they can only crack it within a short period, greatly improving the security of data transmission. The frequency and intensity of key updates enable the system to withstand external attacks and prevent security risks caused by long-term key leakage.

[0096] Embodiment 2: Components that are the same or corresponding to those in Embodiment 1 are labeled with the corresponding reference numerals as in Embodiment 1. For the sake of simplicity, only the differences from Embodiment 1 will be described below. The difference between this Embodiment 2 and Embodiment 1 lies in:

[0097] The present invention also provides an intelligent camera control system, and the control system includes a dual-spectrum sensing module, an edge computing unit, an IoT communication module, and a dynamic power management unit;

[0098] The dual-spectrum sensing module integrates a low-light sensor and an infrared sensor, ensures the time alignment of the data of the two sensors through a synchronous trigger circuit, and is built with a weighted fusion preprocessing chip to perform weighted adaptive fusion of thermal imaging and visible light images in real time, reducing the computational burden on the backend and improving data processing efficiency. At the same time, it has an automatic calibration function to adjust the sensor parameters according to the environmental temperature and light conditions, improving the quality of image acquisition;

[0099] The edge computing unit is configured with a quad-core heterogeneous processor (integrating a GPU and an FPGA module, specifically for efficiently executing Bayesian neural network inference tasks), and a built-in programmable logic array (supporting a customized neural network accelerator to optimize the real-time processing ability of multi-modal data) for performing Bayesian neural network inference. The storage module contains a temperature-material feature association database for quick retrieval to optimize target recognition and tracking performance, and also supports local model training and inference, reducing the dependence on cloud computing resources and enhancing data privacy and system response speed;

[0100] The IoT communication module supports a dual-protocol switching module for the MQTT protocol and the LoRaWAN protocol, ensuring data transmission stability and wide-area coverage ability in different network environments, and uploading target location data and environmental parameters in real time. It also has an end-to-end data encryption function, using the AES-256 encryption algorithm to ensure the security of data transmission;

[0101] The dynamic power management unit dynamically adjusts the power consumption of the sensor and the main frequency of the computing unit according to the operating mode.

[0102] The edge computing unit is equipped with hardware acceleration modules (such as Tensor cores and FPGA accelerators), which significantly improve the deep learning inference speed and support simultaneous floating-point operations on four-channel feature maps, optimizing the parallel processing ability of multi-modal data, including thermal imaging gradient maps, material reflection feature maps, motion vector maps, and environmental light parameters. The IoT communication module includes a data encryption sub-module that encrypts the transmitted data using the AES-256 encryption algorithm and regenerates the key pair every 15 seconds to ensure communication security. The dynamic power management unit integrates an adaptive voltage regulator that intelligently adjusts the voltage and current according to the device operating load, optimizing power consumption to the greatest extent and facilitating the system to adjust the working mode (high-definition mode or low-power mode) according to actual needs.

[0103] Finally, it should be noted that the above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions recorded in the foregoing embodiments or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A smart camera control method based on the Internet of Things, characterized in that: The following steps are involved: S1: Multimodal sensor data fusion acquisition, using synchronous low-light sensors and infrared sensors to obtain polarized reflected light images and thermal imaging data of the monitored scene respectively; S2: Dual-modal feature association modeling, divide the thermal imaging data into temperature topological partitions according to the pixel grid, trigger the dynamic target tracking signal when the absolute value of the average temperature difference between adjacent partitions exceeds the threshold, and analyze the polarization angle parameters of the material surface from the polarized reflected light image, use the support vector machine classification algorithm to identify the material type, and establish a temperature-material feature association database; S3: Dynamic weighted image fusion, dynamically adjusts the fusion weight coefficient of thermal imaging and visible light images according to the light intensity change rate parameter monitored in real time by the ambient light sensor; S4: Probabilistic reasoning network processing, deploying Bayesian neural networks on edge computing nodes, extracting spatiotemporal features through multi-channel feature maps and convolution modules, and using MCMC algorithms to update weights to optimize target position prediction; S5: Occlusion compensation and trajectory correction. When the intersection-over-combination ratio of the target bounding box between consecutive frames decreases beyond the threshold, the historical temperature data is used to build a Markov random field model, and the probability distribution of the candidate area is calculated based on the temperature spatial correlation to compensate for the target position offset. S6: Adaptive energy efficiency control, dynamically switching the camera working mode based on the target tracking confidence quantile; S7: Multi-node collaborative verification, in a haze environment, the target trajectory is generated through thermal imaging data, the timing window is used to reduce positioning noise, and the positioning results and parameters are uploaded to the cloud platform through the MQTT protocol.

2. The method for controlling an intelligent camera based on the Internet of Things according to claim 1, characterized in that: The fusion formula of step S3 is: ； Among them, α represents the dynamic adjustment coefficient of ambient light, T thermal (x, y) is the temperature value of the thermal imaging image at position (x, y), I visual (x,y) is the pixel value of the visual image at position (x,y).

3. The method for controlling an intelligent camera based on the Internet of Things according to claim 1, characterized in that: The step S4 specifically comprises the following steps: S41: A four-layer Bayesian neural network is deployed on the edge computing node, and the input layer is connected to a four-channel feature map, including a thermal imaging gradient map, a material reflection feature map, a motion vector map, and ambient lighting parameters; S42: The residual convolution layer extracts spatiotemporal features, the fully connected layer generates the posterior probability distribution of the target location, and the output layer updates the weight parameters based on Markov chain Monte Carlo sampling; S43: Perform posterior probability distribution inference through edge computing devices. The posterior probability distribution inference formula is as follows: ； Among them, ω represents the network weight, D represents the training data, p(w|D) represents the posterior probability distribution, p(D|w) represents the likelihood function, and p(w) represents the prior distribution.

4. The method for controlling an intelligent camera based on the Internet of Things according to claim 3, characterized in that: The weight update of the Bayesian neural network follows the following formula: ； Among them, ω t is the node weight of the tth iteration, ε is the learning rate, P(D|w t ) is the likelihood probability of the training data under the current parameters.

5. The method for controlling an intelligent camera based on the Internet of Things according to claim 3, characterized in that: The fusion result F of step S1 fusion (x, y) is used as the input of the lightweight Bayesian neural network for real-time inference of target tracking. The Bayesian neural network infers the posterior probability distribution and outputs the probability distribution of the target position: ； Among them, W1, W2, W3, and W4 are the weight matrices of the Bayesian neural network, ReLU is the activation function, and softmax is used for probability normalization.

6. The method for controlling an intelligent camera based on the Internet of Things according to claim 1, characterized in that: The weights of the Markov random field model in step S5 are defined as follows: ； Among them, T R is the average temperature of the candidate area, T hist is the mean historical temperature of the target, ρ is the spatial correlation factor, and σ is the standard deviation of the temperature distribution.

7. The method for controlling an intelligent camera based on the Internet of Things according to claim 1, characterized in that: In the step S6, when the confidence quantile is lower than 0.9, the high-definition mode is enabled, and when the confidence quantile is higher than 0.9, the mode is switched to the low-power mode, and the inactivated sensor power supply modules are turned off.

8. The method for controlling an intelligent camera based on the Internet of Things according to claim 1, characterized in that: The uploading data transmission in step S7 adopts the AES-256 encryption algorithm, and the key pair is regenerated every 15 seconds.

9. An intelligent camera control system, used to implement the intelligent camera control method according to any one of claims 1 to 8, characterized in that: The control system includes a dual-spectrum sensing module, an edge computing unit, an IoT communication module, and a dynamic power management unit; The dual-spectrum sensing module integrates a low-light sensor and an infrared sensor, and has a built-in weighted fusion pre-processing chip; The edge computing unit is configured with a quad-core heterogeneous processor with a built-in programmable logic array for performing Bayesian neural network reasoning, and the storage module includes a temperature-material feature association database; The IoT communication module supports dual-protocol switching modules of MQTT protocol and LoRaWAN protocol, and uploads target positioning data and environmental parameters in real time; The dynamic power management unit dynamically adjusts the sensor power consumption and the main frequency of the computing unit according to the operation mode.

10. The intelligent camera control system according to claim 9, characterized in that: The edge computing unit is equipped with a hardware acceleration module and supports simultaneous floating-point operations of four-channel feature maps. The Internet of Things communication module includes a data encryption submodule, and the dynamic power management unit integrates an adaptive voltage regulator.

Images