CN-121977701-A - Motion robot double-light fusion furnace bottom non-stop temperature measurement method and system

CN121977701ACN 121977701 ACN121977701 ACN 121977701ACN-121977701-A

Abstract

The invention discloses a non-stop temperature measurement method and system for a double-light fusion furnace bottom of a moving robot, and relates to the technical field of metallurgical equipment temperature monitoring. The method comprises the steps of controlling a moving robot carrying a double-light camera to stop and cruise along a preset path, synchronously collecting infrared and visible light images and robot pose data, conducting anti-bump stabilization processing on an image sequence based on gyroscope data, processing the stabilized double-light images, achieving target segmentation, tracking and association of irregular areas, outputting a plurality of areas with unique identity marks, extracting temperature values from the infrared images according to the identity marks and space information, and conducting calculation and early warning. The system comprises corresponding functional modules. According to the invention, the segmentation precision is improved through double-light fusion, and the problems of low temperature measurement efficiency and insufficient precision of an irregular area of the ultra-large furnace bottom in a dynamic scene are solved by combining the bumpy stabilization and target tracking correlation technology, so that the rapid, accurate and automatic temperature monitoring without stopping is realized.

Inventors

ZHANG MENG
ZHANG TAO
SUN YUPENG
DENG CHENGCHENG

Assignees

杭州艾铂特智能科技有限公司

Dates

Publication Date: 20260505
Application Date: 20260407

Claims (10)

1. The double-light fusion furnace bottom non-stop temperature measurement method for the moving robot is characterized by comprising the following steps of: Controlling a moving robot carrying a double-light camera and a sensor, and executing non-stop cruising according to a preset routing inspection path; Synchronously acquiring infrared and visible light double-light images and robot pose data; Based on the robot pose data, performing anti-bump stabilization treatment on the collected double-light image sequence to obtain a stabilized double-light image; performing target detection and target tracking processing on the stabilized double-light image, and outputting a plurality of irregular areas with unique identity marks and motion state parameters; dividing all targets into a first target set which is stably tracked and a second target set which needs total updating based on the tracking state of the irregular area; For targets in the first target set, generating a search range for the next frame image segmentation processing based on the motion state parameters of the targets, and calculating to obtain an incremental temperature value based on the temperature data of the current frame and the historical frame of the targets; And extracting and calculating the full temperature value from the infrared images synchronously acquired based on the spatial information of the targets in the second target set.
2. The method of claim 1, wherein the simultaneous acquisition of infrared and visible dual light images comprises: Performing high-frequency vibration correction on the double-light image sequence to obtain an image sequence subjected to high-frequency vibration correction; And carrying out low-frequency offset correction on the image sequence subjected to the high-frequency vibration correction based on an optical flow method so as to obtain the stabilized double-light image.
3. The method of claim 1, wherein the registering, fusing segmentation and object tracking of the stabilized bi-optic image comprises: Registering the infrared image and the visible light image in the stabilized double-light image to obtain a registered double-light image; inputting the registered double-light images into a pre-trained fusion segmentation model, carrying out feature fusion and irregular region segmentation, and outputting segmented irregular regions; and carrying out target tracking and identity management on the partitioned irregular areas, and outputting unique identity identification and motion state parameters of each area.
4. The method of claim 1, wherein the performing the state study and preliminary scheduling on all targets based on the unique identity of the irregular area and the motion state parameter comprises: Judging each target as a stable tracking state or a full-quantity updating state; The state of 'needing full quantity updating' refers to that the target is the first occurrence of the current frame or the tracking confidence level is lower than a preset threshold value; outputting all targets judged to be in a stable tracking state as the first target set, and outputting all targets judged to be in a full-quantity updating state as the second target set.
5. The method of claim 1, wherein the generating a search range for a next frame image segmentation process for the objects in the first set of objects based on their motion state parameters comprises: predicting the image position of the target in the next frame according to the motion state parameters of the target by using a Kalman filter; And generating a rectangular coordinate area set with a preset size by taking the predicted position as the center, and taking the rectangular coordinate area set as a search range of the next frame image segmentation processing.
6. The method of claim 1, wherein calculating and outputting delta temperature values based on temperature data of a current frame and a historical frame of the target comprises: acquiring a current frame original temperature measured value of a target and a last frame final output temperature value, and calculating a temperature variation between the current frame original temperature measured value and the last frame final output temperature value; when the absolute value of the temperature variation is smaller than a preset temperature variation rate threshold value, calculating an incremental temperature value of the current frame by carrying out nonlinear attenuation processing on the variation; and when the absolute value of the temperature change amount is larger than or equal to the temperature change rate threshold value, determining the original temperature measured value of the current frame as an increment temperature value of the current frame.
7. The method of claim 1, wherein the extracting and calculating the full-scale temperature value from the synchronously acquired infrared images based on the spatial information thereof for the targets in the second target set comprises: extracting a temperature value of a corresponding pixel from the infrared image data according to the spatial information of the targets in the second target set; and calculating the temperature parameter of each region based on the temperature value of the corresponding pixel, correcting the temperature according to a preset model, and outputting a full-scale temperature calculation result.
8. The utility model provides a motion robot double light fuses no-stop temperature measurement system in stove bottom which characterized in that includes: the path planning and control module is used for planning a routing inspection path and controlling the moving robot to execute non-stop cruising along the path; the synchronous data acquisition module is used for synchronously acquiring an infrared image, a visible light image and robot pose data containing gyroscope data in the cruising process; The image stabilization processing module is used for performing anti-bump stabilization processing on the collected double-light image sequence based on the gyroscope data to obtain a stabilized double-light image; the target segmentation, tracking and optimizing module is used for registering, fusing and segmenting the stabilized double-light image and carrying out target tracking processing on the stabilized double-light image, outputting an irregular area with a unique identity and a motion state parameter and carrying out efficiency optimization based on the motion state parameter, wherein the target segmentation, tracking and optimizing module comprises: the image registration unit is used for registering the stabilized infrared image and the visible light image; The double-light fusion segmentation unit is used for inputting the registered double-light images into a pre-trained fusion segmentation model to carry out irregular region segmentation; the multi-target tracking unit is used for tracking and managing the identity of the segmented area and outputting the unique identity and the motion state parameter; the efficiency optimization scheduling unit is used for generating an adaptive region of interest coordinate set for the next frame segmentation based on the motion state parameters, feeding back the coordinate set to the double-light fusion segmentation unit to limit the search range of the region of interest, and directly calculating and outputting an increment temperature value of the stably tracked region; The temperature calculation module is used for executing temperature extraction and correction calculation from the infrared image data based on the space information of the target area which is identified by the efficiency optimization scheduling unit and needs full-scale calculation to obtain a full-scale temperature result; And the result summarizing and early warning module is used for receiving and summarizing the increment temperature value and the full-quantity temperature result and triggering early warning when the summarized temperature result exceeds a set threshold value.
9. An electronic device, comprising: A memory for storing a computer program; a processor for executing a computer program stored on the memory, implementing the method according to any one of claims 1 to 7.
10. A computer readable storage medium, characterized in that the computer readable storage medium has stored therein a computer program which, when executed by a processor, implements the method according to any of claims 1 to 7.

Description

Motion robot double-light fusion furnace bottom non-stop temperature measurement method and system Technical Field The invention relates to the technical field of metallurgical equipment temperature monitoring, in particular to a non-stop temperature measurement method and system for a double-light fusion furnace bottom of a moving robot. The invention also relates to the electronic equipment and the storage medium for realizing the method. Background The furnace bottom temperature is a core index of the safe operation of the metallurgical furnace, and for the ultra-large furnace bottom, the ultra-large furnace bottom comprises a great number of irregular temperature measuring areas, the efficiency and the precision are difficult to be considered in the prior temperature measuring technology, and two types of technical paths and obvious defects mainly exist: One type is a traditional manual operation mode, wherein an operator holds the temperature measuring device to stop and collect the temperature at fixed points, or the operator draws the temperature measuring area on an infrared image by hand. Aiming at the problems that the ultra-large furnace bottom has extremely low efficiency (60 minutes are needed for the furnace bottom of a 120m2 blast furnace), more manual intervention, strong subjectivity of temperature measurement results, easy omission of irregular areas and the like when aiming at the ultra-large furnace bottom extremely many irregular areas, the mode can not completely meet the requirement of batch rapid temperature measurement. The other type is an automatic temperature measurement technology based on image segmentation, related patents form preliminary exploration, but are not adapted to the core requirements of 'ultra-large furnace bottom + irregular area + no-stop acquisition', and the specific defects are as follows: the prior art is cited as follows: CN112539843B (method, device and computer equipment) is characterized by that it adopts gray threshold segmentation and variance maximization method to extract high-low temperature region, has no deep learning participation, and depends on manual setting threshold value, and does not relate to motion acquisition equipment, and adopts parking fixed acquisition point position, and can not implement continuous temperature measurement of furnace bottom, and the dynamic acquisition and accurate temperature measurement logic of motion robot and cradle head are completely missing. CN103196564B (an infrared thermal imaging temperature measurement method for correcting the surface emissivity through image segmentation) improves the temperature measurement accuracy through the image segmentation and correction emissivity, but an emissivity measuring instrument is required to be additionally used, so that 'one-image rapid measurement' can not be realized, and the scene adaptation for high temperature and dust interference of the furnace bottom is not realized. CN118823027B (a method for detecting the running state of a decomposing furnace) adopts a clustering algorithm to divide an abnormal region, and combines the entropy value correction probability, so that the method is mainly aimed at fault diagnosis of the decomposing furnace, temperature parameter calculation of a fixed-point region of the unfocused furnace bottom, and has insufficient instantaneity. Summary of core defects of the prior art: 1. The high-temperature area segmentation precision is low, namely infrared image pixels are saturated due to the large-area high temperature of the furnace bottom, area offset easily occurs in single infrared image segmentation, and the boundary recognition capability of an irregular area is poor; 2. Poor dynamic adaptability, namely bumpy motion robot and serious image shake caused by uneven furnace bottom, the prior art cannot adapt to a non-stop acquisition scene, and the temperature measurement stability under the dynamic state cannot be ensured; 3. the efficiency bottleneck is obvious, no video continuity association mechanism exists, the same target is repeatedly detected, the traditional parking acquisition mode is long in time consumption, and the efficiency is extremely low when aiming at extremely many irregular areas of the oversized furnace bottom; 4. And the irregular area is poor in adaptation, priori knowledge of the irregular structure of the furnace bottom is not combined, the temperature measuring area is positioned inaccurately, and the incomplete problem of the detection area is easy to occur. Disclosure of Invention The invention aims to solve the core problems of low temperature measurement efficiency, insufficient precision and poor dynamic adaptability of the ultra-large furnace bottom in the prior art aiming at a very large number of irregular areas, and realizes rapid and accurate temperature measurement by automatically identifying the irregular areas on the premise of taking non-stop acquisition as a core, and specifically aims