CN-122023767-A - Automatic target tracking and monitoring system in high-temperature furnace based on multispectral infrared

CN122023767ACN 122023767 ACN122023767 ACN 122023767ACN-122023767-A

Abstract

The invention is suitable for the technical field of target tracking, and discloses an automatic target tracking and monitoring system in a high-temperature furnace based on multispectral infrared, which comprises a tracking acquisition module, a control module and a control module, wherein the tracking acquisition module is used for identifying and tracking the target in the furnace based on a YOLO target detection algorithm, presetting initial acquisition frequency and synchronously acquiring radiation image data of the target in different infrared wave bands and acquiring environmental parameters in the furnace; and the identification and judgment module is used for synchronously extracting the apparent morphology and deformation rate characteristics of the target based on the collected multiband target radiation image data by inverting the target surface temperature distribution field through a multispectral radiation transmission model. According to the invention, the primary label is obtained through deformation rate characteristic matching, and then the temperature distribution field and the apparent morphology are combined to generate the correction factor to complete secondary calibration, so that multi-source information cross verification is realized, the false alarm rate and the false alarm rate are greatly reduced, clear and reliable stage basis is provided for subsequent temperature control, the internal tissue defect of the workpiece caused by stage judgment errors is reduced, and the yield of finished products is improved.

Inventors

WANG CHENGWEI
ZHOU PAN
ZHANG YONG
LI ZHIFENG
MENG FEI

Assignees

马鞍山市科泰电气科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260114

Claims (8)

1. The automatic tracking and monitoring system for the target in the high-temperature furnace based on multispectral infrared comprises a tracking and acquisition module, a detection module and a control module, wherein the tracking and acquisition module is used for identifying and tracking the target in the furnace based on a YOLO target detection algorithm, presetting initial acquisition frequency and synchronously acquiring radiation image data of the target in different infrared bands and acquiring environmental parameters in the furnace, and is characterized by further comprising the following components; The identification and judgment module is used for inverting a target surface temperature distribution field through a multispectral radiation transmission model based on collected multiband target radiation image data, synchronously extracting target apparent morphology and deformation rate characteristics, matching the deformation rate characteristics with a preset deformation rate characteristic vector library to obtain a preliminary label, fusing the surface temperature distribution field and the apparent morphology to generate a correction factor, and correcting the preliminary label by the correction factor to generate a final label; The adjusting and comparing module dynamically adjusts the initial acquisition frequency according to the importance degree judged by the current process state, and performs point-by-point comparison on the target surface temperature distribution field and a preset process temperature reference field to calculate an average temperature deviation value; And the execution module is used for dividing the abnormal grades according to the average temperature deviation value, generating a differential regulation instruction and executing the differential regulation instruction.
2. The automatic tracking and monitoring system for the target in the high-temperature furnace based on the multispectral infrared according to claim 1, wherein the specific steps of identifying and tracking the target in the furnace based on the YOLO target detection algorithm are as follows: collecting multispectral infrared images containing targets at different forging stages and different interference degrees, marking the targets in the images by using a boundary box to form a special sample library of the targets in the furnace, and performing targeted training on YOLOv models based on the sample library; Acquiring a current image of a target in a furnace from a multispectral infrared camera according to a preset or dynamically adjusted frequency, preprocessing the current image, inputting the preprocessed current image into a trained YOLOv model, and outputting target information detected in the current image through the YOLOv model, wherein the template information comprises boundary frame coordinates, credibility and target types; Inputting target information of a current image into a BoT-SORT tracker, predicting the boundary frame position of the current image based on the previous image position of each existing target track by utilizing a Kalman filter, and optimally matching a detection frame in the current image with the existing predicted track by the BoT-SORT tracker through calculating motion similarity and appearance similarity; Updating the matched detection frame information to the corresponding track, wherein the unique ID of the track is kept unchanged, track continuation is completed, the unmatched detection frame is initialized to be a new track, a BoT-SORT tracker allocates a new unique ID and re-matches the new unique ID in the subsequent continuous images, if matching is successful, track continuation is confirmed, and otherwise, the track is terminated.
3. The automatic tracking and monitoring system for the target in the high-temperature furnace based on the multispectral infrared according to claim 2, wherein the specific steps of inverting the target surface temperature distribution field through the multispectral radiation transmission model are as follows: The collected multiband target radiation image is subjected to pixel alignment by adopting a sub-pixel level phase correlation method, and then the factory calibration coefficient and the field calibration parameter of the multispectral infrared camera are utilized to carry out image pixel gray value Converted into the actual radiation brightness value of the corresponding wave band The specific conversion formula is as follows: wherein For the different infrared bands of wavelengths, For the image pixel coordinates, Is of wave band Is used for the radiation gain coefficient of (a), Is of wave band Is a radiation offset coefficient of (2); establishing an actual radiance value True temperature of target Is defined by the relation: , wherein, For the smoke to the wave band Is used for the optical fiber, the transmittance of the optical fiber, For smoke to wave band Is used for the optical fiber, the transmittance of the optical fiber, In-band for target acquisition by laboratory calibration Is used for the optical fiber, the emissivity of (a), Is of wave band Lower temperature The corresponding blackbody radiation intensity is used to determine, For the brightness of the incident radiation of the environment, The radiation emitted by the flue gas is used for the flue gas; For each pixel And combining the relational expressions of all wave bands, and then carrying out iterative solution by using a least square method.
4. The automatic tracking and monitoring system for the target in the high-temperature furnace based on multispectral infrared as claimed in claim 3, wherein the specific steps of extracting the apparent morphology and deformation rate characteristics of the target are as follows: Firstly, selecting a wave band with highest contrast between a target and a background in a furnace from a multiband image as a characteristic image I (x, y, t), filtering the characteristic image to inhibit noise, stretching the contrast to highlight the edge of the target, determining the outline of the target on the characteristic image by adopting an edge detection algorithm according to a region which is higher than a background preset temperature threshold value in a target surface temperature distribution field as an initial position, obtaining a binarized image B (x, y, t), and finally calculating the area, perimeter and compactness characteristic parameters of the binarized image B (x, y, t) based on the binarized image B (x, y, t); Invoking contour features of two adjacent images of the same target ID, completing contour matching by adopting an iterative nearest point algorithm, and calculating the width change rate of the minimum circumscribed rectangle of the two images based on a matching result Height variation Area variation And combined with time intervals The deformation rate of width, height, and area, collectively referred to as deformation rate characteristics, is obtained.
5. The automatic tracking and monitoring system for the target in the high-temperature furnace based on multispectral infrared rays according to claim 4, wherein the specific acquisition steps of the initial tag are as follows: Summarizing standard deformation rate characteristics of typical process stages in m in historical data to construct a vector library, and identifying process stage names corresponding to each deformation rate characteristic; The Euclidean geometric distance between the current deformation rate characteristic and the standard deformation rate characteristic of each process stage in the vector library is calculated through an Euclidean geometric distance formula, and the Euclidean geometric distance is used for measuring the similarity degree between the current deformation rate characteristic and the standard deformation rate characteristic of each process stage in the vector library; and selecting the standard deformation rate characteristic with the minimum European geometric distance with the current rate deformation characteristic, and outputting the process stage name corresponding to the standard deformation rate characteristic as a primary label.
6. The automatic tracking and monitoring system for the target in the high-temperature furnace based on multispectral infrared rays according to claim 5, wherein the specific acquisition steps of the correction factors are as follows: Defining an ideal characteristic range and an acceptable characteristic range for each process state; extracting average temperature and temperature uniformity parameters in a temperature distribution field, and area, perimeter and compactness parameters in apparent morphology features; Comparing the extracted parameters with a preset ideal characteristic range and an acceptable characteristic range one by one, if the extracted parameters are in the ideal characteristic range, the score is 1, if the extracted parameters are not in the ideal range but in the acceptable characteristic range, the score is 0.5, and otherwise, the score of each parameter is 0; and (5) distributing weight to each parameter, and obtaining a correction factor through weighted fusion.
7. The automatic tracking and monitoring system for the target in the high-temperature furnace based on multispectral infrared rays according to claim 6, wherein the specific steps of correcting the preliminary label by using the correction factor to generate the final label are as follows: Presetting a high threshold Y1 and a low threshold Y2, and And the correction factor is denoted as Y; When (when) When the final label is confirmed as the preliminary label When the final label is judged to be abnormal in process At this point the final label maintains the preliminary label, but a low confidence flag is appended.
8. The automatic tracking and monitoring system for the target in the high-temperature furnace based on multispectral infrared rays according to claim 7, wherein the specific acquisition step of the average temperature deviation value is as follows: according to the current process state, a corresponding preset process temperature reference field is called, and the coordinate of each measuring point in the target surface temperature distribution field is aligned with the coordinate in the process temperature reference field; And sequentially calculating absolute differences of measured temperatures and reference temperatures of all measurement points on the surface of the target, finally summing the absolute differences of all the measurement points, dividing the sum by the total number of the measurement points, and finally obtaining an average temperature deviation value.

Description

Automatic target tracking and monitoring system in high-temperature furnace based on multispectral infrared Technical Field The invention relates to the technical field of target tracking, in particular to an automatic target tracking and monitoring system in a high-temperature furnace based on multispectral infrared. Background High-temperature forging is a core process in the fields of metallurgy, heavy machinery manufacturing, aerospace part machining and the like, and the core requirement is that the mechanical property and the forming precision of a workpiece are ensured by accurately controlling the temperature and the deformation of the workpiece in different forging stages. The automatic tracking and monitoring of the target in the high-temperature furnace is a core support for realizing the precise control of the process, the safety protection of equipment and the intelligent upgrading of production by high-temperature forging. In the forging furnace scene, automatic tracking and monitoring can continuously lock the workpiece, accurately judge the stages of initial forging, middle-stage forming, later cooling and the like, and realize staged temperature closed loop control. If tracking monitoring is not available, internal tissue defects are easily generated in the workpiece due to stage judgment errors and excessive temperature fluctuation, so that the yield of finished products is increased; In the current production process of the forging heating furnace, the judgment of the forging stage of the workpiece mainly depends on personal experience of operators, and the judgment has stronger subjectivity and inconsistency, so that the stage identification precision is low and the result fluctuation is large. Because of the lack of clear and consistent stage judgment basis, the follow-up temperature control is difficult to realize precision and stability, thereby influencing the heating uniformity, the metal structure performance and the final forming quality of the workpiece and causing the defects of oxidization, overburning, deformation and the like. Disclosure of Invention The invention aims to provide a multispectral infrared-based automatic tracking and monitoring system for targets in a high-temperature furnace, which aims to solve the problems that the subjectivity is strong and the judgment error is large in the conventional forging stage of judging a workpiece by depending on the experience of an operator, so that the follow-up temperature control has no clear stage basis and the forming quality of the workpiece is affected. The automatic tracking and monitoring system for the target in the high-temperature furnace based on multispectral infrared comprises a tracking and acquisition module, a radiation image acquisition module and a monitoring module, wherein the tracking and acquisition module is used for identifying and tracking the target in the furnace based on a YOLO target detection algorithm, presetting initial acquisition frequency and synchronously acquiring radiation image data of the target in different infrared bands and acquiring environmental parameters in the furnace; The identification and judgment module is used for inverting a target surface temperature distribution field through a multispectral radiation transmission model based on collected multiband target radiation image data, synchronously extracting target apparent morphology and deformation rate characteristics, matching the deformation rate characteristics with a preset deformation rate characteristic vector library to obtain a preliminary label, fusing the surface temperature distribution field and the apparent morphology to generate a correction factor, and correcting the preliminary label by the correction factor to generate a final label; The adjusting and comparing module dynamically adjusts the initial acquisition frequency according to the importance degree judged by the current process state, and performs point-by-point comparison on the target surface temperature distribution field and a preset process temperature reference field to calculate an average temperature deviation value; And the execution module is used for dividing the abnormal grades according to the average temperature deviation value, generating a differential regulation instruction and executing the differential regulation instruction. Preferably, the specific steps of identifying and tracking the target in the furnace based on the YOLO target detection algorithm are as follows: Firstly, acquiring multispectral infrared images containing targets at different forging stages and different interference degrees, marking the targets in the images by using a boundary box to form a special sample library of the targets in the furnace, and performing targeted training on YOLOv models based on the sample library; Secondly, acquiring a current image of a target in the furnace from a multispectral infrared camera according to a preset or dynamically adj