A temperature control system and method for furnaces in high-temperature zones

Abstract

This invention discloses a furnace temperature measurement and control system and method for high-temperature areas, including a dual-spectrum imaging module, an onboard processing unit, a gimbal module, a power supply module, and an industrial IoT communication module. The dual-spectrum imaging module is used to acquire thermal image data and visible light image data, and output thermal image frame streams. The onboard processing unit receives continuous thermal image frame streams, and for each frame, executes a hotspot identification algorithm to determine the pixel coordinates of the highest temperature point in the image, and calculates a two-dimensional error vector e between these coordinates and a preset image frame center point. The error vector is input into a PID controller, and control commands are sent to the gimbal module to drive the camera to adjust its aiming direction to reduce and ultimately eliminate the error vector. Key telemetry data and real-time video streams are accessible to operators via the industrial IoT communication module for manual intervention when necessary. This invention features the ability to automatically lock onto and track high-temperature areas without manual intervention.

Description

Technical Field

[0001] This invention relates to the fields of non-contact thermal imaging temperature measurement technology and industrial process control technology, specifically to a furnace and kiln temperature measurement control system and method in high-temperature areas. Background Technology

[0002] In heavy industrial production processes such as metallurgy, building materials, and chemicals, furnaces are core thermal equipment. They operate under harsh conditions of high temperature, heavy load, and continuous operation for extended periods, and the temperature state of the furnace shell (kiln wall) directly affects equipment safety, production efficiency, and product quality.

[0003] However, existing furnace temperature measurement technologies have significant limitations: 1. Inherent Defects of Manual Inspection. Currently, many factories still rely on manual handheld infrared thermometers or thermal imagers for scheduled inspections. This method has many problems: First, it is inefficient and lacks real-time performance. Manual inspections are time-consuming and cannot achieve continuous and accurate monitoring of temperature changes, resulting in significant data feedback delays. Second, manual inspections are prone to missed or false detections, leading to untimely early warnings and making it difficult to effectively predict potential faults and intervene in risks. Finally, temperature measurement stations are usually located near high-temperature kilns, creating harsh working environments that not only pose safety threats to workers but also make recruitment difficult, leading to continuously rising labor costs.

[0004] 2. Limitations of Fixed Online Monitoring Systems. To overcome the shortcomings of manual inspections, the industry has developed fixed online monitoring systems. These systems typically install thermal imagers at specific locations near the kiln, enabling continuous monitoring and data recording 24 / 7. However, their "fixed" nature also brings new problems. First, deployment flexibility is poor; a single system can only monitor a fixed area. For large kilns or factories with multiple kilns, achieving full coverage requires a huge investment in equipment, resulting in low economic efficiency. Second, the monitoring field of view is limited; if a hotspot develops abnormally or moves out of the camera's fixed field of view due to kiln rotation, the system loses its monitoring capability.

[0005] 3. Functional Gaps in Traditional Portable Devices. While existing portable temperature measurement devices have solved the problem of deployment flexibility, their functionality remains at the level of "advanced manual tools." Operators still need to be physically present on-site, manually aiming the device at the target area and visually identifying and judging high-temperature points. Descriptions of existing portable solutions do not mention any automatic tracking functions, indicating that they are essentially measurement tools requiring continuous human intervention, failing to achieve true automation and intelligence.

[0006] In summary, among existing technologies, manual inspection offers portability and low per-cycle cost, but sacrifices automation and data quality; fixed systems achieve automation but sacrifice portability and cost-effectiveness; traditional portable devices retain portability but lack any automation capabilities. Summary of the Invention

[0007] In order to overcome the defects of the existing technology, the purpose of this invention is to provide a furnace temperature measurement and control system and method in a high-temperature area, which has the characteristics of being able to automatically lock and track the high-temperature area without manual intervention.

[0008] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A furnace temperature measurement and control system for high-temperature areas includes a dual-spectrum imaging module, an onboard processing unit, a gimbal module, a power supply module, and an industrial Internet of Things (IIoT) communication module. The dual-spectrum imaging module is used to acquire thermal image data and visible light image data, and output thermal image frame stream; The onboard processing unit receives a continuous stream of thermal image frames. For each frame, it executes a hotspot identification algorithm to determine the pixel coordinates of the highest temperature point in the image. Then, it calculates the two-dimensional error vector e between the coordinates and the preset image frame center point. The error vector is input into the PID controller, which calculates the angular velocity required to drive the pan and tilt motors of the gimbal based on the magnitude, accumulation and rate of change of the error. The control command is sent to the gimbal module to drive the camera to adjust its aiming direction in order to reduce and eventually eliminate the error vector. Key telemetry data and real-time video streams are transmitted to a remote monitoring center via a 5G network through an Industrial Internet of Things (IIoT) communication module, allowing operators to view them or intervene manually when necessary.

[0009] The gimbal module is a multi-axis gimbal, and the dual-spectrum imaging module is connected to the housing through the multi-axis gimbal. The multi-axis gimbal is used to drive the dual-spectrum imaging module to rotate in at least the horizontal and pitch directions. The dual-spectrum imaging module and the multi-axis gimbal are communicatively connected to the onboard processing unit. A quick-installation module is installed at the bottom of the base of the housing. The quick-installation module is magnetically attached to a ferromagnetic surface to achieve quick fixation and disassembly of the device. The quick-installation module has a switchable magnet and includes a base plate, the bottom of which may be designed with a V-groove.

[0010] An active thermal management module is installed inside the housing. The active thermal management module autonomously tracks the amount of heat generated by the high-performance onboard processing unit during operation. The active thermal management module is a solid-state semiconductor cooler. The cold end absorbs and collects internal heat through an internal heat-conducting plate that is in close contact with the onboard processing unit; the hot end is rigidly thermally connected to the external heat dissipation fins on the device casing by welding or thermal paste.

[0011] The active thermal management module actively pumps internal heat to the external environment, ensuring the stable operation of core electronic components in high-temperature environments.

[0012] A method for temperature measurement and control of a furnace in a high-temperature zone includes the following steps; Step (1): Receive a frame of thermal image data acquired by the thermal imaging sensor in the dual-spectrum imaging module; Step (2): The onboard processing unit analyzes the thermal image data, identifies the pixel coordinates corresponding to the highest temperature value by traversing the temperature matrix of the thermal image data; or extracts the high-temperature connected region with the largest area in the image by using an adaptive threshold segmentation and contour detection algorithm, and calculates its geometric centroid as the position coordinates to determine the location of a highest temperature region in the frame. Step (3): The onboard processing unit calculates the error vector between the location of the highest temperature region and a predetermined center point of the frame; and generates and sends a control signal to the multi-axis gimbal to drive the multi-axis gimbal to move, thereby reducing the magnitude of the error vector; Step (4): Based on the magnitude of the error vector, a future location of the highest temperature region is predicted using a Kalman filter based on a series of historical locations.

[0013] In step (1), the active thermal management module continues to operate, actively pumping internal heat to the external environment to provide high-temperature protection for the thermal imaging sensor and ensure its data acquisition accuracy and working stability under harsh conditions.

[0014] In step (2), a more robust image processing method is adopted: first, adaptive threshold segmentation is performed on the thermal image to dynamically generate a binary mask that retains only the high temperature region. Then, the contour detection algorithm is used to find all high temperature connected regions from the mask and filter out the connected region with the largest area. Finally, the geometric centroid of the connected region is calculated to accurately determine its position coordinates, thereby stably locating the most representative highest temperature region in the frame.

[0015] Step (3) specifically involves: The formula for calculating the error vector e is: ; in( , ) represents the coordinates of the highest temperature region. , () represents the coordinates of the center point of the image frame; The error vector is input into a PID (proportional-integral-derivative) controller. Based on the magnitude, accumulation, and rate of change of the error, the controller calculates the angular velocity required to drive the pan and tilt motors of the gimbal. Control commands are sent to the gimbal module to drive the camera to adjust its aiming direction in order to reduce and eventually eliminate the error vector. This closed-loop process is repeated continuously at a high frequency (e.g., 30 times per second) to ensure that the camera can lock the center of its field of view on the hottest area on the kiln wall in real time and automatically. At the same time, telemetry data (such as maximum temperature value, equipment status, etc.) and real-time video streams are sent to the remote monitoring center via the 5G network through the IIoT communication module, so that operators can view them or manually intervene when necessary.

[0016] The onboard processing unit is also configured to generate the control signal via a proportional-integral-derivative (PID) controller. The proportional (P) stage generates a fast response based on the current error magnitude; the integral (I) stage accumulates historical errors to eliminate steady-state deviations; and the derivative (D) stage predicts trends based on the error change rate to suppress overshoot and oscillations. Finally, the controller outputs angular velocity signals in the horizontal and pitch directions to drive the gimbal. By precisely tuning the PID parameters, fast, accurate, and stable tracking of the target is achieved. The IIoT communication module uses the MQTT protocol to transmit telemetry data and the WebRTC protocol to transmit video data streams. This module connects the onboard processing unit to the remote monitoring center, enabling remote transmission of device data, status monitoring, and reception of control commands. In step (4), the Kalman filter makes an optimal estimate of the dynamic state of the system and predicts its future state. (1) State Modeling: The filter establishes a state vector for the hotspot, which includes not only its current position (x, y) but also its velocity (x, y). , ); (2) Prediction steps: Before processing a new frame of image, the Kalman filter uses the state and motion model of the previous moment to predict the position where the hot spot should appear in the current frame; for rotary kilns, this motion model can be a uniform circular motion model. (3) Update step: After the hotspot identification module provides the actual measured location of the hotspot in the new frame image, the filter compares this "measured value" with the previous "predicted value"; it uses the difference between the two to correct and update the internal state vector to obtain a more accurate current state estimate. Step (4) specifically includes: A state vector containing position and velocity is established. A prediction step is performed to estimate the current state based on the previous state and the motion model. Then, the state is corrected by actual position measurement to obtain the optimal estimate. Finally, the predicted future position is output. ).

[0017] The predicted future location is used as the location of the highest temperature region for calculating the error vector; specifically, the predicted coordinates ( Substitute into the error vector calculation formula: The forward error vector is obtained.

[0018] The beneficial effects of this invention are: 1. Combining dual-spectral imaging and real-time image processing algorithms, the system can automatically identify high-temperature areas and dynamically adjust the attitude of the multi-axis gimbal. The onboard processing unit generates angular velocity control signals in the horizontal and pitch directions through a proportional-integral-derivative (PID) controller to achieve fast, accurate and stable target tracking. At the same time, through a Kalman filter, the future position of the highest temperature area is predicted based on historical positions. The optimal estimate is obtained through state prediction and measurement correction, avoiding errors and lags in manual operation.

[0019] 2. With the help of the communication assembly, users can remotely monitor temperature data and video streams in the control center or on mobile terminals, and intervene in control in real time, which significantly improves operational safety and response efficiency.

[0020] 3. The system has good anti-interference capabilities and is suitable for harsh industrial environments such as high temperature and high dust. At the same time, the algorithm has adaptive capabilities and can cope with various furnace types and operating conditions.

[0021] 4. It not only provides real-time temperature monitoring, but also records historical data, generates temperature field distribution maps and trend reports, providing data support for process optimization and predictive maintenance. Attached Figure Description

[0022] Figure 1 This is a system architecture block diagram of an embodiment of the present invention.

[0023] Figure 2 Detailed cross-sectional view of the protective casing and active thermal management module.

[0024] Figure 3 This is a software flowchart for the "hotspot identification and adaptive gimbal control" process.

[0025] Figure 4 This is a diagram showing the operational status of the predictive tracking submodule. Detailed Implementation

[0026] The present invention will now be described in further detail with reference to the accompanying drawings.

[0027] A furnace temperature measurement and control system for high-temperature areas includes a dual-spectrum imaging module for acquiring thermal image data and visible light image data; the dual-spectrum imaging module is connected to the housing via a multi-axis gimbal, and combined with a real-time image processing algorithm, automatically identifies high-temperature areas and dynamically adjusts the gimbal attitude; the multi-axis gimbal is used to drive the dual-spectrum imaging module to rotate in at least the horizontal and pitch directions; the dual-spectrum imaging module and the multi-axis gimbal are communicatively connected to an onboard processing unit.

[0028] The base of the housing is equipped with a quick-installation module. The quick-installation module is magnetically attached to a ferromagnetic surface to achieve quick fixation and disassembly of the equipment. The quick-installation module has a switchable magnet.

[0029] An active thermal management module containing a semiconductor cooler is installed inside the housing and close to the onboard processing unit. The active thermal management module actively pumps internal heat to the external environment to ensure the stable operation of core electronic components in high-temperature environments.

[0030] A method for temperature measurement and control of a furnace in a high-temperature zone includes the following steps; Step (1): Receive a frame of thermal image data acquired by the thermal imaging sensor in the dual-spectrum imaging module; during this process, the active thermal management module continues to operate, actively pumping internal heat to the external environment, providing high-temperature protection for core electronic components (including the thermal imaging sensor), and ensuring its data acquisition accuracy and working stability under harsh conditions.

[0031] Step (2): Analyze the thermal image data and identify the pixel coordinates corresponding to the highest temperature value by traversing the temperature matrix of the thermal image data; or, in order to deal with complex scenarios such as noise interference and multiple hot spots, adopt a more robust image processing method: first, perform adaptive threshold segmentation on the thermal image to dynamically generate a binary mask that only retains the high temperature area, then use the contour detection algorithm to find all high temperature connected components from the mask, and filter out the connected component with the largest area. Finally, calculate the geometric centroid of the connected component to accurately determine its position coordinates, thereby stably locating the most representative highest temperature area in the frame; Step (3): Calculate the error vector between the location of the highest temperature region and a predetermined center point of the frame; and generate and send a control signal to the gimbal to drive the gimbal movement; Step (4): Based on the magnitude of the error vector obtained in the above steps, a future location of the highest temperature region is predicted using a Kalman filter based on a series of historical locations.

[0032] Step (3) specifically involves: The formula for calculating the error vector e is: ; in( , ) represents the coordinates of the highest temperature region. , () represents the coordinates of the center point of the image frame; The error vector is input into a PID controller, which calculates the angular velocity required to drive the pan-tilt and tilt motors based on the magnitude, accumulation, and rate of change of the error. Control commands are sent to the pan-tilt module to drive the camera to adjust its aiming direction in order to reduce and eventually eliminate the error vector. This closed-loop process is repeated continuously at a high frequency to ensure that the camera can lock the center of its field of view on the hottest area on the kiln wall in real time and automatically. At the same time, key telemetry data and real-time video streams are transmitted to the remote monitoring center via the 5G network through the IIoT communication module, allowing operators to view them or intervene manually when necessary.

[0033] The IIoT communication module uses the MQTT protocol to transmit telemetry data and the WebRTC protocol to transmit video data streams. This module connects the onboard processing unit with the remote monitoring center to realize remote transmission of device data, status monitoring, and reception of control commands.

[0034] Step (4) specifically involves: A state vector containing position and velocity is established. A prediction step is performed to estimate the current state based on the previous state and the motion model. Then, the state is corrected by actual position measurement to obtain the optimal estimate. Finally, the predicted position is output. ); The predicted future location is used as the location of the highest temperature region for calculating the error vector; specifically, the predicted coordinates ( Substitute into the error vector calculation formula: The forward error vector is obtained.

[0035] Reference Figure 1 This invention demonstrates the interconnectivity between a dual-spectrum image acquisition module, an onboard processing unit, a gimbal module, a power supply module, a thermal management module, a quick-installation module, and an Industrial Internet of Things (IIoT) communication module. The core working principle of this invention is a closed-loop vision servo control system. In practical applications, operators use the quick-installation module to fix the device to a ferromagnetic surface near the furnace.

[0036] After the device is started, the onboard processing unit receives a continuous stream of thermal image frames from the dual-spectrum image acquisition module. For each frame, the processing unit executes a hotspot identification algorithm to determine the pixel coordinates of the highest temperature point in the image. Then, it calculates the two-dimensional error vector e between the coordinates and the preset image frame center point. The error vector is input into a PID (Proportional-Integral-Derivative) controller. Based on the magnitude, accumulation, and rate of change of the error, the controller calculates the angular velocities required to drive the pan and tilt motors of the gimbal. Control commands are sent to the gimbal module, which adjusts the camera's aiming direction to reduce and ultimately eliminate the error vector. This closed-loop process is repeated continuously at a high frequency (e.g., 30 times per second), ensuring that the camera can automatically and in real-time lock its field of view onto the hottest area on the kiln wall. Simultaneously, critical telemetry data (such as maximum temperature and equipment status) and real-time video streams are transmitted via an IIoT communication module to a remote monitoring center through a 5G network, allowing operators to review the data or intervene manually when necessary.

[0037] like Figure 2 As shown, the portable quick-installation module serves as the physical interface between the device and industrial site facilities. Its core design principle lies in resolving the conflict between the "stability" required for autonomous operation and the "mobility" needed for a portable device. The module's base plate structurally acts as the carrier of the magnets and the guide component for the magnetic circuit. Through a built-in linkage mechanism or electrical control interface, one or more high-power switching permanent magnets are rigidly fixed to it, thus forming a complete adsorption unit capable of withstanding enormous shear forces and rapidly switching on and off. These magnets are designed to have a holding force exceeding 175 pounds (approximately 80 kilograms), sufficient to withstand the vibrations and shocks commonly encountered in industrial environments, ensuring absolute stability of the device during autonomous tracking.

[0038] Its key technical feature lies in its "switch-type" design. A mechanical handle or electrical switch allows for easy switching on or off of the magnetic circuit, controlling the presence or absence of magnetic force. This enables a single operator to install, reposition, or disassemble the equipment within seconds without any tools. Furthermore, the base can be designed with V-grooves, allowing it to not only adhere firmly to flat surfaces but also reliably mount on common curved surfaces such as rotary kilns, greatly enhancing on-site adaptability. This design perfectly combines the stability of bolt fixing with the convenience of temporary installation, solving a problem that traditional solutions could not achieve simultaneously.

[0039] The multi-axis gimbal module provides the imaging module with precise and smooth motion capabilities. It is a two-axis (horizontal and pitch) gimbal driven by a high-torque stepper motor or servo motor. Its design references ruggedized PTZ (Pan-Tilt-Zoom) systems used in harsh industrial environments, ensuring high reliability and durability. The gimbal achieves its wide range of motion through a compact and robust two-stage mechanical structure: horizontal rotation is supported by a high-rigidity crossed roller bearing base, allowing for unrestricted 360° continuous rotation; while pitch motion is achieved through a pair of precision-machined trunnions combined with a high-strength harmonic reducer, converting motor power into smooth and highly torsional-rigid pitch action. This ensures that the imaging module remains stable within a range of ±90 degrees of elevation, as if firmly held. This design, integrating unlimited rotation and large-angle pitch, creates a near-perfect, blind-spot-free observation hemisphere in space. The motor and transmission mechanism are precisely designed to have extremely small backlash and high repeatability, which is the physical basis for ensuring the stability and tracking accuracy of the PID control algorithm.

[0040] The housing and active thermal management module provide a safe and stable operating environment for the internal precision electronic components. The housing is constructed of die-cast aluminum or 304 stainless steel and achieves an IP67 protection rating through the use of silicone seals at the joints. This means the device is completely dustproof and can withstand brief immersion in 1 meter of water without damage, making it suitable for weathering, washing, and dusty environments in industrial settings.

[0041] This module integrates an active thermal management system. The high-performance onboard processing unit, which autonomously tracks the required heat output, generates a significant amount of heat during operation. Simultaneously, the equipment operates near a high-temperature furnace, resulting in a high ambient temperature. The combined effect of internal and external heat sources renders traditional passive cooling methods (such as heat sinks) completely inadequate. Therefore, this invention employs a solid-state thermoelectric cooler (TEC), also known as a Peltier module. Figure 2 As shown, the TEC (Active Heat Pump) is centrally located, with its cold end efficiently absorbing and collecting internal heat through an internal heat-conducting plate in close contact with the onboard processing unit. Its hot end is rigidly connected to external heat sink fins on the device casing via soldering or thermal paste. When DC current passes through the TEC, it acts like a miniature heat pump, actively "pumping" the internally generated heat to the external environment. Even in high ambient temperatures, it forces the internal core chip temperature to remain within a safe operating range. This active cooling solution is an enabling technology for reliable operation of high-performance edge computing in extreme environments. Without it, powerful onboard processing capabilities would be impossible, and the device's "intelligent" core would be inoperable.

[0042] The dual-spectrum image acquisition module serves as the system's "eyes," integrating a thermal imaging sensor and a visible light sensor near the same optical axis. The thermal imaging section uses an uncooled microbolometer as the detector, achieving a resolution of 640x512 pixels in the long-wave infrared (LWIR) spectral range. Crucially, this thermal imager possesses "radiometric" capability, meaning each pixel value directly corresponds to a precise temperature reading, providing essential quantification data for subsequent hotspot identification algorithms. The visible light section employs a high-resolution CMOS sensor (e.g., 64 megapixels) to capture high-definition color images of the site. These two images can be fused and overlaid or presented side-by-side, allowing remote operators to intuitively understand the specific physical structure corresponding to thermal anomalies (such as welds, refractory brick detachment, etc.), significantly improving the accuracy of fault diagnosis.

[0043] The onboard processing unit is the "brain" of the system, responsible for executing all complex computational tasks locally. It utilizes a System-on-Chip (SoC) optimized for edge AI and computer vision applications, such as the Qualcomm QCS8550. These chips integrate a powerful multi-core CPU, a graphics processing unit (GPU), and a dedicated neural network processing unit (NPU), sufficient to run the entire software stack locally in real time: including image data decoding, hotspot identification algorithms, PID closed-loop control, Kalman filter prediction, and the encapsulation and transmission of communication protocols. Performing all computational tasks on the device avoids the significant network latency associated with transmitting video streams to the cloud for analysis, a prerequisite for achieving highly dynamic, low-latency autonomous tracking control.

[0044] Reference Figure 3 The hotspot identification and localization module accurately and reliably identifies target hotspots from the thermal image data stream. In the simplest embodiment, the algorithm only needs to traverse the temperature matrix of the current frame to find the maximum temperature value and its corresponding (x,y) pixel coordinates. However, to cope with complex situations such as noise and multiple hotspots in a small area in real-world applications, more robust image processing methods can be employed.

[0045] One approach is to use adaptive thresholding to extract all regions in the image above a certain dynamic threshold, forming a binary mask. Then, a contour detection algorithm is used to find the largest connected component within this mask, and the geometric centroid of that component is calculated as the hotspot coordinates.

[0046] Another more advanced method is to use the Maximum Stable Extreme Region (MSER) ​​algorithm, which can detect bright spots (i.e., hotspots) in the image very stably and has stronger anti-interference capabilities. Regardless of the method used, the final output of this module is a precise (x,y) coordinate pair, which serves as the tracking target for the next-level control module.

[0047] Adaptive gimbal control module This is the core control algorithm for achieving automatic tracking. This invention uses a proportional-integral-derivative (PID) controller to achieve precise control of the gimbal.

[0048] (1) Input (error vector e): The input to the controller is the deviation between the hotspot target position and the image center position, i.e. .

[0049] (2) Proportional (P) stage: The output of this stage is proportional to the current error. The larger the error, the faster the gimbal moves to quickly approach the target. It determines the system's response speed.

[0050] (3) Integral (I) term: This term accumulates the errors from all past moments. Its function is to eliminate the steady-state error of the system. For example, even with continuous small disturbances (such as wind), the integral term can ensure that the gimbal eventually stops precisely at the center of the target, rather than having a small fixed deviation to the side.

[0051] (4) Differential (D) stage: The output of this stage is proportional to the rate of change of the error. Its function is to predict the future trend of the error, thereby playing the role of "braking" and "damping", effectively suppressing the overshoot and oscillation of the system when approaching the target, making the tracking process smoother and more stable.

[0052] The PID controller outputs two independent control variables: the horizontal angular velocity and the pitch angular velocity. These two control variables are sent to the gimbal's motor driver, thus completing one closed-loop control cycle. This is achieved by carefully tuning the three parameters of the PID controller (…). , , This allows the system's tracking response to reach an optimal state that is fast, accurate, and stable.

[0053] Predictive Tracking Submodule (Advanced Implementation): For equipment like rotary kilns where the main body undergoes periodic rotational motion, the trajectory of surface hotspots is predictable. A standard PID controller is a purely "reactive" controller; it only begins to correct after an error is detected. For fast-moving targets, this lag causes the gimbal to constantly "chase" behind the target, making perfect locking impossible.

[0054] To address this issue, a high-level embodiment of the present invention introduces a predictive tracking submodule, referring to... Figure 4 .

[0055] The core of this module is the Kalman Filter, a powerful recursive estimation algorithm that can make optimal estimates of the dynamic state of a system based on a series of noisy observations and predict its future state.

[0056] (1) State Modeling: The filter establishes a state vector for the hotspot, which includes not only its current position (x, y) but also its velocity (x, y). , ).

[0057] (2) Prediction Step: Before processing a new frame of image, the Kalman filter uses the state and motion model of the previous moment to predict the location where the hotspot should appear in the current frame. For a rotary kiln, this motion model can be a uniform circular motion model.

[0058] (3) Update step: After the hotspot identification module provides the actual measured location of the hotspot in the new frame image, the filter compares this "measured value" with the previous "predicted value". It uses the difference between the two to correct and update the internal state vector to obtain a more accurate current state estimate (including more accurate location and velocity).

[0059] The introduction of this mechanism represents a qualitative leap in control systems—upgrading from "reactive" control to "active" or "predictive" control. Now, the target of the PID controller is no longer the measured position of the hotspot from the previous moment, but rather the predicted position of the hotspot at the current moment, given by the Kalman filter. This means the gimbal can begin moving in advance to "meet" the approaching target, rather than passively chasing it. This proactive control method greatly eliminates tracking delay, enabling silky-smooth and highly accurate target locking even when facing a high-speed rotating kiln body. This is a significant innovation and improvement over simple vision servo systems.

[0060] The IIoT communication module is responsible for data exchange between the device and the remote monitoring platform. Considering the characteristics of Industrial Internet of Things (IIoT) scenarios, this invention employs a specific communication protocol stack. The device has a built-in 5G cellular network module to meet the requirements of high bandwidth and low latency communication. For structured telemetry data, such as maximum temperature values, device GPS coordinates, battery level, and operating status, the Message Queuing Telemetry Transport (MQTT) protocol is used. MQTT is a lightweight messaging protocol based on a publish / subscribe model, well-suited for use in industrial environments with limited bandwidth and unstable networks. The device acts as an MQTT client, publishing various types of data to different "topics," while the remote server subscribes to these topics to receive data. MQTT also provides different Quality of Service (QoS) levels to ensure reliable delivery of critical alarm and other information.

[0061] Low-latency video streaming module For real-time video streaming, this invention employs the Web Real-Time Communication (WebRTC) protocol. Compared to traditional streaming media protocols such as RTSP, WebRTC offers significant advantages. RTSP typically requires dedicated player plugins and suffers from high latency when transmitting over public networks. WebRTC, on the other hand, is designed for peer-to-peer real-time audio and video communication between browsers or between a browser and a device. It is natively supported by modern mainstream browsers and requires no plugins. Through WebRTC, this device can directly push thermal imaging and visible light video streams to a remote operator's webpage or mobile application with ultra-low latency of less than 500 milliseconds. This dual-protocol communication architecture, combining MQTT and WebRTC, is a well-thought-out collaborative design. It doesn't simply choose two trendy technologies, but rather selects the optimal transmission channel for different types of data: lightweight and reliable MQTT for transmitting "small data" (telemetry and control signaling), and efficient, low-latency WebRTC for transmitting "big data" (video streams). This architecture maximizes the utilization efficiency of 5G network bandwidth, providing remote users with a perfect experience that combines data reliability and visual immediacy, constituting a major innovation of this invention at the communication level.

Claims

1. A furnace temperature measurement and control system for high-temperature zones, characterized in that, It includes a dual-spectrum imaging module, an onboard processing unit, a gimbal module, a power supply module, and an industrial IoT (IIoT) communication module; The dual-spectrum imaging module is used to acquire thermal image data and visible light image data, and output thermal image frame stream; The onboard processing unit receives a continuous stream of thermal image frames. For each frame, it executes a hotspot identification algorithm to determine the pixel coordinates of the highest temperature point in the image. Then, it calculates the two-dimensional error vector e between the coordinates and the preset image frame center point. The error vector is input into the PID controller, which calculates the angular velocity required to drive the pan and tilt motors of the gimbal based on the magnitude, accumulation and rate of change of the error. Control commands are sent to the gimbal module to drive the camera to adjust its aiming direction in order to reduce and eventually eliminate the error vector. Key telemetry data and real-time video streams are transmitted to a remote monitoring center via a 5G network through an Industrial Internet of Things (IIoT) communication module, allowing operators to view them or intervene manually when necessary.

2. The furnace temperature measurement and control system for high-temperature zones according to claim 1, characterized in that, The gimbal module is a multi-axis gimbal, and the dual-spectrum imaging module is connected to the housing through the multi-axis gimbal. The multi-axis gimbal is used to drive the dual-spectrum imaging module to rotate in at least the horizontal and pitch directions. The dual-spectrum imaging module and the multi-axis gimbal are communicatively connected to the onboard processing unit. A quick-installation module is installed at the bottom of the base of the housing. The quick-installation module is magnetically attached to a ferromagnetic surface to achieve quick fixation and disassembly of the device. The quick-installation module has a switchable magnet and includes a base plate with a V-groove at the bottom.

3. The furnace temperature measurement and control system for high-temperature zones according to claim 2, characterized in that, An active thermal management module is installed inside the housing; The active thermal management module is a solid-state semiconductor cooler. The cold end absorbs and collects internal heat through an internal heat-conducting plate that is in close contact with the onboard processing unit; the hot end is rigidly thermally connected to the external heat dissipation fins on the device casing by welding or thermal paste. The active thermal management module actively pumps internal heat to the external environment, ensuring the stable operation of core electronic components in high-temperature environments.

4. A method for temperature measurement and control of a furnace / kiln in a high-temperature zone based on the system described in any one of claims 1-3, characterized in that, Includes the following steps; Step (1): Receive a frame of thermal image data acquired by the thermal imaging sensor in the dual-spectrum imaging module; Step (2): The onboard processing unit analyzes the thermal image data, identifies the pixel coordinates corresponding to the highest temperature value by traversing the temperature matrix of the thermal image data; or extracts the high-temperature connected region with the largest area in the image by using an adaptive threshold segmentation and contour detection algorithm, and calculates its geometric centroid as the position coordinates to determine the location of a highest temperature region in the frame. Step (3): The onboard processing unit calculates the error vector between the location of the highest temperature region and a predetermined center point of the frame; and generates and sends a control signal to the multi-axis gimbal to drive the multi-axis gimbal to move, thereby reducing the magnitude of the error vector; Step (4): Based on the magnitude of the error vector, a future location of the highest temperature region is predicted using a Kalman filter based on a series of historical locations.

5. The method for temperature measurement and control of a furnace in a high-temperature zone according to claim 4, characterized in that, In step (1), the active thermal management module continues to operate, actively pumping internal heat to the external environment to provide high-temperature protection for the thermal imaging sensor and ensure its data acquisition accuracy and working stability under harsh conditions.

6. The method for temperature measurement and control of a furnace in a high-temperature zone according to claim 4, characterized in that, In step (2), a more robust image processing method is adopted: first, adaptive threshold segmentation is performed on the thermal image to dynamically generate a binary mask that retains only the high temperature region. Then, the contour detection algorithm is used to find all high temperature connected regions from the mask and filter out the connected region with the largest area. Finally, the geometric centroid of the connected region is calculated to accurately determine its position coordinates, thereby stably locating the most representative highest temperature region in the frame.

7. The method for temperature measurement and control of a furnace in a high-temperature zone according to claim 4, characterized in that, Step (3) specifically involves: The formula for calculating the error vector e is: ； in( , ) represents the coordinates of the highest temperature region. , () represents the coordinates of the center point of the image frame; The error vector is input into a PID controller, which calculates the angular velocity required to drive the pan and tilt motors of the gimbal based on the magnitude, accumulation and rate of change of the error. The control command is sent to the gimbal module to drive the camera to adjust its aiming direction in order to reduce and eventually eliminate the error vector. This closed-loop process is repeated continuously at a high frequency, ensuring that the camera can lock the center of its field of view on the hottest area on the kiln wall in real time and automatically. At the same time, telemetry data and real-time video streams are transmitted to the remote monitoring center via the 5G network through the IIoT communication module, so that operators can view them or manually intervene when necessary.

8. The method for temperature measurement and control of a furnace in a high-temperature zone according to claim 7, characterized in that, The onboard processing unit is also configured to generate the control signal via a proportional-integral-derivative (PID) controller. The proportional (P) stage generates a fast response based on the current error magnitude; the integral (I) stage accumulates historical errors to eliminate steady-state deviations; and the derivative (D) stage predicts trends based on the error change rate to suppress overshoot and oscillations. Finally, the controller outputs angular velocity signals in the horizontal and pitch directions to drive the gimbal. By precisely tuning the PID parameters, fast, accurate, and stable tracking of the target is achieved. The IIoT communication module uses the MQTT protocol to transmit telemetry data and the WebRTC protocol to transmit video data streams. This module connects the onboard processing unit with the remote monitoring center to realize remote transmission of device data, status monitoring, and reception of control commands.

9. The method for temperature measurement and control of a furnace in a high-temperature zone according to claim 8, characterized in that, In step (4), the Kalman filter makes an optimal estimate of the dynamic state of the system and predicts its future state. (1) State Modeling: The filter establishes a state vector for the hotspot, which includes not only its current position (x, y) but also its velocity (x, y). , ); (2) Prediction step: Before processing a new frame of image, the Kalman filter uses the state and motion model of the previous moment to predict the location where the hotspot should appear in the current frame; (3) Update step: After the hot spot recognition module gives the actual measurement location of the hot spot in the new frame image, the filter compares the "measured value" with the previous "predicted value"; using the difference between the two, the internal state vector is corrected and updated to obtain a more accurate current state estimate.

10. The method for temperature measurement and control of a furnace in a high-temperature zone according to claim 9, characterized in that, Step (4) specifically involves: A state vector containing position and velocity is established. A prediction step is performed to estimate the current state based on the previous state and the motion model. Then, the state is corrected by actual position measurement to obtain the optimal estimate. Finally, the predicted future position is output. ); The predicted future location is used as the location of the highest temperature region for calculating the error vector; specifically, the predicted coordinates ( Substitute into the error vector calculation formula: The forward error vector is obtained.

Images