WO2020098299A1 - 一种高温熔融流体流速检测方法及系统 - Google Patents

一种高温熔融流体流速检测方法及系统 Download PDF

Info

Publication number
WO2020098299A1
WO2020098299A1 PCT/CN2019/096143 CN2019096143W WO2020098299A1 WO 2020098299 A1 WO2020098299 A1 WO 2020098299A1 CN 2019096143 W CN2019096143 W CN 2019096143W WO 2020098299 A1 WO2020098299 A1 WO 2020098299A1
Authority
WO
WIPO (PCT)
Prior art keywords
molten fluid
feature block
module
flow rate
feature
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2019/096143
Other languages
English (en)
French (fr)
Inventor
蒋朝辉
何磊
陈致蓬
桂卫华
潘冬
阳春华
谢永芳
张海峰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Central South University
Original Assignee
Central South University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Central South University filed Critical Central South University
Priority to US16/970,344 priority Critical patent/US11605169B2/en
Publication of WO2020098299A1 publication Critical patent/WO2020098299A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/50—Depth or shape recovery
    • G06T7/507—Depth or shape recovery from shading
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01F—MEASURING VOLUME, VOLUME FLOW, MASS FLOW OR LIQUID LEVEL; METERING BY VOLUME
    • G01F1/00—Measuring the volume flow or mass flow of fluid or fluent solid material wherein the fluid passes through a meter in a continuous flow
    • G01F1/704—Measuring the volume flow or mass flow of fluid or fluent solid material wherein the fluid passes through a meter in a continuous flow using marked regions or existing inhomogeneities within the fluid stream, e.g. statistically occurring variations in a fluid parameter
    • G01F1/708—Measuring the time taken to traverse a fixed distance
    • G01F1/712—Measuring the time taken to traverse a fixed distance using auto-correlation or cross-correlation detection means
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01F—MEASURING VOLUME, VOLUME FLOW, MASS FLOW OR LIQUID LEVEL; METERING BY VOLUME
    • G01F1/00—Measuring the volume flow or mass flow of fluid or fluent solid material wherein the fluid passes through a meter in a continuous flow
    • G01F1/704—Measuring the volume flow or mass flow of fluid or fluent solid material wherein the fluid passes through a meter in a continuous flow using marked regions or existing inhomogeneities within the fluid stream, e.g. statistically occurring variations in a fluid parameter
    • G01F1/708—Measuring the time taken to traverse a fixed distance
    • G01F1/7086—Measuring the time taken to traverse a fixed distance using optical detecting arrangements
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01F—MEASURING VOLUME, VOLUME FLOW, MASS FLOW OR LIQUID LEVEL; METERING BY VOLUME
    • G01F15/00—Details of, or accessories for, apparatus of groups G01F1/00 - G01F13/00 insofar as such details or appliances are not adapted to particular types of such apparatus
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/22—Matching criteria, e.g. proximity measures
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/20—Analysis of motion
    • G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
    • G06T7/248—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments involving reference images or patches
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/70—Determining position or orientation of objects or cameras
    • G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/80—Analysis of captured images to determine intrinsic or extrinsic camera parameters, i.e. camera calibration
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/42—Global feature extraction by analysis of the whole pattern, e.g. using frequency domain transformations or autocorrelation
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/40—Scenes; Scene-specific elements in video content
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/40—Scenes; Scene-specific elements in video content
    • G06V20/49—Segmenting video sequences, i.e. computational techniques such as parsing or cutting the sequence, low-level clustering or determining units such as shots or scenes
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01F—MEASURING VOLUME, VOLUME FLOW, MASS FLOW OR LIQUID LEVEL; METERING BY VOLUME
    • G01F25/00—Testing or calibration of apparatus for measuring volume, volume flow or liquid level or for metering by volume
    • G01F25/10—Testing or calibration of apparatus for measuring volume, volume flow or liquid level or for metering by volume of flowmeters
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10016—Video; Image sequence
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P10/00—Technologies related to metal processing
    • Y02P10/20—Recycling

Definitions

  • the invention mainly relates to the technical field of high-temperature molten fluid flow rate detection, in particular to a high-temperature molten fluid flow rate detection method and system.
  • ferrous metals and non-ferrous metals are mostly extracted from ores or concentrates by pyrometallurgy under high temperature conditions, producing molten crude metals or metal enrichments and slag from high-temperature closed reaction furnaces
  • High-speed outflow such as molten iron flowing out of the blast furnace outlet in the iron and steel industry process, matte copper outflow from the reverberatory furnace in the pyrometallurgical process, and crude zinc outflow from the smelting furnace in the pyrolysis process.
  • the change trend of the pressure in the reactor is an important indicator to indicate whether the reactor is smooth and forward. However, due to the harsh environment in the reactor, it is difficult to directly detect the pressure change in the reactor.
  • Detecting the flow rate of the molten fluid at the outlet of the reactor can characterize the reaction
  • the pressure in the furnace can also reflect the proportional relationship between the metal and slag produced, which helps to find and eliminate abnormal working conditions in time, improve the permeability of the reactor, and ensure the smooth and smooth production of the reactor. Therefore, detecting the flow rate of the molten fluid at the outlet of the reaction furnace is particularly important for the safe production of the reaction furnace and the improvement of quality and quantity.
  • the detection object of the present invention is the high-temperature high-gloss molten fluid flowing out of the outlet of the reaction furnace, and at the same time, there is unavoidable vibration and a large amount of unevenly distributed dust on the site, which greatly increases the difficulty of detection.
  • the mainstream detection methods are divided into two types: contact measurement method and non-contact measurement method.
  • the contact detection method for detecting high-temperature molten fluid requires the use of high-temperature resistant materials to directly contact the high-temperature molten fluid and high-speed flowing High-temperature fluids will gradually wear out and erode high-temperature resistant materials, resulting in poor repeatability and short service life of the device.
  • non-contact detection methods are mainly through the establishment of a mechanism model to achieve non-invasive detection of flow rate, but also due to the ultra-high temperature of the detection object and the environment Harshness seriously affects the accuracy of detection.
  • Patent Publication No. CN103480813A The invention patent is a continuous casting crystallizer high-temperature molten steel flow rate measuring device and measuring method. Its working principle is that the bearing is fixed on the fixing device through a fixed shaft, and the spring and the measuring rod are respectively installed at the upper and lower symmetrical positions of the bearing.
  • the spring is installed on the "T" type fixing device, the bearing is installed with a bearing sleeve, the bearing sleeve and the angular displacement sensor are connected through a coupling, the angular displacement sensor is powered by the power supply, and the real-time deflection angle of the measuring rod in the flowing molten steel is recorded and passed
  • the data line is transmitted to the data acquisition and analysis system, which converts the angle data into the molten steel flow rate value.
  • the patent re-registers according to the different detection objects of the device. Before use, the device needs to be preheated to 1200 to 1400 °C. It is complicated to use, the detection range is small, and the detection error is large for fluids with excessive flow rates, and after the test is completed The device cannot directly detect the next object and is limited in the repeatability of use.
  • Patent Publication No. CN104131126A The invention patent is a fuzzy model-based blast furnace slag flow detection method.
  • This patent establishes a fuzzy inference model of blast furnace slag flow, combined with the characteristics of the impact of the slag surface height at the i-th moment on the blast furnace slag flow , Set fuzzy membership function about the height of the slag surface at the i-th time, use fuzzy inference model and fuzzy membership function to establish a blast furnace slag flow calculation model, and use the blast furnace slag flow calculation model for online detection of blast furnace real-time slag total flow .
  • the initial value in this patent is obtained by the process personnel from the knowledge of manual operation experience. Human factors have a greater influence and are subjective, and the design process is an open loop, which cannot guarantee the accuracy of long-term operation results.
  • the method and system for detecting the flow velocity of the high-temperature molten fluid provided by the present invention solve the existing technical problem that the detection accuracy of the flow velocity of the molten fluid with high temperature, high speed and high light is not high.
  • the high-temperature molten fluid flow rate detection method proposed by the present invention includes:
  • a feature block of the outline of the molten fluid is extracted, and the flow rate of the molten fluid is obtained based on the feature block.
  • the feature block is specifically a ripple or a shadow that appears during the high-speed molten fluid outflow at a high speed.
  • extracting the molten fluid contour of the molten fluid region of interest includes:
  • the first-order partial derivative finite difference is used to calculate the gradient amplitude and direction of the preprocessed image, and the gradient amplitude is non-maximum suppressed;
  • a double threshold algorithm is used to detect and connect the molten fluid contour of the molten fluid area of interest, and a skeleton extraction algorithm is used to refine the molten fluid contour.
  • the feature block for extracting the outline of the molten fluid includes:
  • the contour of the feature block is completed based on the contour of the molten fluid, and the feature block of the contour of the molten fluid is accurately positioned based on the centroid of the feature block after the contour is completed.
  • acquiring the flow rate of the molten fluid based on the characteristic block includes:
  • the features include size feature, angle feature and position feature;
  • S W is the moving distance of the feature block in the world coordinate system
  • R is the diameter of the outlet hole of the reactor
  • R c is the diameter pixel of the outlet of the reactor on the image
  • S C is the feature block on the molten fluid adjacent to each other The pixel distance moved within two frames
  • the flow rate of the high-temperature high-speed molten fluid is obtained based on the moving distance and the speed formula.
  • acquiring the features of the feature block includes:
  • ⁇ ij is the i + j order moment of the boundary point of the feature block
  • M is the number of pixel points on the boundary of the feature block
  • m 0, 2, ..., M
  • (x m , y m ) is the mth on the boundary of the feature block Pixels
  • (x t , y t ) is the centroid coordinate
  • the angle between the equivalent ellipse of the feature block and the horizontal direction is calculated based on the i + j order moment, and the calculation formula is specifically:
  • ⁇ is the angle between the equivalent ellipse of the feature block and the horizontal direction
  • centroid formula to calculate the centroid of the feature block, specifically:
  • using the features of the feature block to perform similarity matching between two adjacent frames to obtain the pixel distance that the feature block moves in the two adjacent frames includes:
  • step1 The horizontal coordinate x t of the feature block centroid of the previous frame image T (x, y) is the starting point, and the backward frame image T ′ (x, y) searches the feature block centroid x ′ t backward along the horizontal coordinate;
  • step2 In the image T ′ (x, y), mark the centroid that satisfies the condition of x ′ t -x t ⁇ l, l is the threshold set according to observation;
  • step3 Determine whether the ordinate of the centroid of the marker satisfies
  • h ′ is the set threshold
  • y ′ t and y t are the ordinate and the centroid of the feature block of the next image frame
  • step4 Calculate the long semi-axis r 1 ′, short semi-axis r 2 ′ and included angle ⁇ ′ of the marked feature block of the image T ′ (x, y), and convert them into three parameters of the same dimension (r 1 ′ cos ⁇ ′ , r 1 ′ sin ⁇ ′, r ′ 2 ), the feature block parameters are recorded as (x ′ 1 , x ′ 2 , x ′ 3 ), and similarly, the feature block parameters of the image T (x, y) are recorded as (x 1 , x 2 , x 3 );
  • step5 To evaluate the similarity of two feature blocks online in real time, calculate the similarity coefficient, specifically:
  • ⁇ is the similarity coefficient
  • the range of ⁇ is [-1,1]
  • x k is the k-th feature block parameter of the image T (x, y)
  • x ′ k is the image T ′ (x, y)
  • S C is calculated as follows:
  • the high-temperature molten fluid flow rate detection system proposed by the present invention includes:
  • High-speed camera capture video unit used to collect video stream of high-temperature high-speed molten fluid
  • the video acquisition unit is used to decompose the video stream into a sequence of frame images in time sequence and extract the region of molten fluid of interest in the frame image sequence;
  • the contour extraction unit is used to extract the molten fluid contour of the molten fluid area of interest
  • the molten fluid flow rate detection unit is used to extract the characteristic block of the outline of the molten fluid and obtain the flow rate of the molten fluid based on the characteristic block.
  • the characteristic block is specifically a ripple or a shadow that occurs during the high-speed outflow of the high-temperature molten fluid.
  • the high-speed camera capturing video unit includes a high-speed camera for capturing a video stream of high-temperature and high-speed molten fluid, a lens dust-proof cleaning module and a high-speed camera air-cooling module provided on the high-speed camera, wherein,
  • the lens dust-proof cleaning module is used to clean the lens of the high-speed camera when the high-speed camera is not working;
  • the high-speed camera air-cooling module is used to completely cover the high-speed camera body, and the air flow rate between the air-cooling device and the high-speed camera body is accelerated by an external fan to achieve heat dissipation and cooling.
  • the video acquisition unit includes a video acquisition module and a video processing module, where the video acquisition module includes:
  • a / D module used to convert the analog signal of the video source of the video unit captured by the high-speed camera into a digital signal
  • Memory module for storing video signals
  • the video compression module is used to compress the video signal and send the compressed video signal to the video processing module;
  • the video processing module includes:
  • the video stream decomposition module is used to divide the compressed video stream passed by the video acquisition module into frame images in time sequence;
  • the frame image ROI extraction module is used to extract the fused fluid region of interest in the frame image and transfer the extracted frame image group to the contour extraction unit.
  • the molten fluid flow rate detection unit includes a feature block fine positioning module, a feature block matching module and a flow rate detection module sequentially connected to the feature block fine positioning module;
  • the fine positioning module of the feature block includes a sub-module for removing horizontal pixels, a sub-module for identifying and coarse positioning of the feature block, and a sub-module for accurately positioning the feature block.
  • the culling-like horizontal pixel sub-module is used to eliminate the horizontally-distributed pixels in the contour of the molten fluid, to retain the pixels of the feature block with a larger inclination from the horizontal, and to separate the contour of the feature block from the contour line of the molten fluid;
  • the feature block recognition and coarse positioning sub-module is used to calculate the minimum distance, the height difference between the upper endpoints and the angle difference between two adjacent curves in the feature block outline, and based on the minimum distance, the height difference between the upper endpoints and the angle difference to melt Rough positioning of the feature block of the fluid contour;
  • the feature block precise positioning submodule is used to complete the contour of the feature block based on the molten fluid contour, and to accurately locate the feature block of the molten fluid contour based on the centroid of the feature block after completing the contour;
  • the feature block matching module is used to perform similarity matching between the features of the feature block between two adjacent frames to obtain the pixel distance that the feature block moves in the time between two adjacent frames;
  • the flow velocity detection module includes a camera calibration stator module and a flow velocity output submodule.
  • the camera calibration stator module is used to calibrate the high-speed camera and solve the moving distance of the feature block in the world coordinate system.
  • the flow velocity output submodule is used to calculate and output the flow velocity of the molten fluid.
  • the method and system for detecting the flow rate of a high-temperature molten fluid collects a video stream of a high-temperature and high-speed molten fluid, decomposes the video stream into a sequence of frame images in a time sequence, and extracts a region of molten fluid of interest in the frame image sequence. Extract the molten fluid contour of the molten fluid area of interest, extract the feature block of the molten fluid contour, and obtain the flow rate of the molten fluid based on the feature block.
  • the feature block is specifically the ripples or shadows that appear during the high-temperature molten fluid high-speed outflow process.
  • the flow rate detection accuracy of the molten fluid with high temperature, high speed, and high light is not high.
  • small targets on the molten fluid are accurately tracked and located in real time Significant feature block, so as to realize the flow rate detection process of molten fluid with high temperature, high speed and high light.
  • the method and system have the advantages of high accuracy, strong stability, long periodicity, suitable for high-speed or excessively high-speed flowing fluid, and low investment cost.
  • FIG. 1 is a flowchart of a method for detecting a flow rate of a high-temperature molten fluid according to Embodiment 1 of the present invention
  • FIG. 2 is a flowchart of a method for detecting a flow rate of a high-temperature molten fluid according to Embodiment 2 of the present invention
  • FIG. 3 is a flowchart of a method for extracting a molten fluid contour of a molten fluid region of interest in Embodiment 2 of the present invention
  • FIG. 4 is a schematic diagram of a matrix of a horizontal-like pixel culling module according to Embodiment 2 of the present invention.
  • FIG. 5 is a state diagram of adjacent curves of a feature block and an inverted feature block according to Embodiment 2 of the present invention.
  • FIG. 6 is a program block diagram of a feature block fine positioning module according to Embodiment 2 of the present invention.
  • FIG. 7 is a structural block diagram of a high-temperature molten fluid flow rate detection system of the present invention.
  • FIG. 8 is a schematic diagram of a molten fluid flow rate detection device of the present invention and a scene.
  • High-speed camera capture video unit 20, video acquisition unit; 30, contour extraction unit; 40, molten fluid flow rate detection unit; 101, high-speed camera; 102, lens dust cleaning module; 103, high-speed camera air cooling module; 201 , Video acquisition module; 202, video processing module; 401, feature block fine positioning module; 402, feature block matching module; 403, flow rate detection module.
  • a method for detecting a flow rate of a high-temperature molten fluid according to Embodiment 1 of the present invention includes:
  • Step S101 Collect a video stream of a high-temperature and high-speed molten fluid
  • Step S102 decompose the video stream into a sequence of frame images in time sequence, and extract regions of molten fluid of interest in the frame image sequence;
  • Step S103 extract the molten fluid contour of the molten fluid region of interest
  • step S104 a feature block of the outline of the molten fluid is extracted, and the flow velocity of the molten fluid is obtained based on the feature block.
  • the feature block is specifically a ripple or a shadow that occurs during the high-speed outflow of the high-temperature molten fluid.
  • the high-temperature molten fluid flow rate detection method collects the video stream of the high-temperature and high-speed molten fluid, decomposes the video stream into a time-series frame image sequence, and extracts the region of the molten fluid of interest in the frame image sequence. Extract the molten fluid contour of the molten fluid area of interest, extract the feature block of the molten fluid contour, and obtain the flow rate of the molten fluid based on the feature block.
  • the feature block is specifically the ripples or shadows that appear during the high-temperature molten fluid high-speed outflow process. There is a technical problem that the flow rate detection accuracy of the molten fluid with high temperature, high speed, and high light is not high.
  • the embodiment of the present invention is relatively new It is proposed to extract the feature block of the molten fluid contour, that is, the ripples or shadows that appear during the high-speed molten fluid high-speed outflow, and accurately track and locate the small target salient feature block on the molten fluid in real time, so as to achieve High-speed, high-gloss molten fluid flow rate detection process.
  • the method has the advantages of high accuracy, strong stability, long periodicity, and is suitable for high-temperature or excessively high-speed flowing fluids, and has low investment cost.
  • a method for detecting a flow rate of a high-temperature molten fluid according to Embodiment 2 of the present invention includes:
  • Step S201 Collect a video stream of a high-temperature and high-speed molten fluid.
  • the embodiment of the present invention collects a video stream of a high-temperature and high-speed molten fluid through a high-speed camera.
  • step S202 the video stream is decomposed into a sequence of frame images in time sequence, and the molten fluid region of interest in the frame image sequence is extracted.
  • the video stream information of the high-temperature and high-speed molten fluid object is collected by the high-speed camera; the collected video stream for a period of time is decomposed into a time-series frame image sequence and the fusion of interest in the frame image sequence is extracted Fluid region; then, the image enhancement operation is performed on the region of interest and the contour of the molten fluid feature block is extracted by the single-pixel contour extraction module to obtain a binary frame contour image of the molten fluid.
  • Step S203 Extract the molten fluid contour of the molten fluid region of interest.
  • pre-process the frame image group including gradation processing and image enhancement processing.
  • gradation processing there are multiple color spaces, of which RGB data processing is the most convenient to convert to gray space.
  • RGB data processing is the most convenient to convert to gray space.
  • the RGB-to-gray space conversion is first performed to obtain a gray signal for subsequent processing.
  • the conversion formula is as follows,
  • the edge of the image corresponds to the high-frequency part of the Fourier transform of the image in the frequency domain of the image, and the background region of the image corresponds to the low-frequency part.
  • the image enhancement program is implemented by an exponential high-pass filter, and its transfer function is:
  • D (u, v) is the distance from the origin of the gray (x, y) Fourier transform center, and D 0 is the cutoff frequency.
  • the Gray (x, y) Fourier transform of the image results in Gray (u, v), and the product of H (u, v) performs the inverse Fourier transform to obtain a high-pass filtered image, ie,
  • the single pixel contour of the molten fluid region of interest is extracted.
  • the Canny operator has the advantages of high positioning accuracy, accurate detection, and strong anti-interference ability in detecting contours, but the extracted contour is a non-single pixel contour. For this reason, after the Canny operator is output A skeleton extraction program for morphological processing is added to refine the outline.
  • the embodiment of the present invention extracts the single-pixel contour of the molten fluid region of interest mainly including Gaussian filtering U521, gradient and gradient direction calculation U522, non-maximum suppression U523, dual threshold segmentation U524, and skeleton extraction U525, as shown in FIG.
  • the first step is to smooth the image with a Gaussian filter to reduce noise interference, and then use the first-order partial derivative finite difference to calculate the gradient amplitude and direction of the image and perform non-maximum suppression of the gradient amplitude.
  • Step S204 extract the feature block of the outline of the molten fluid.
  • the feature block for extracting the outline of the molten fluid in the embodiment of the present invention includes:
  • step S2041 pixels with horizontal distribution in the outline of the molten fluid are eliminated, and the pixel points of the feature block at a predetermined inclination angle to the horizontal are retained, and the feature block contour is separated from the outline of the molten fluid.
  • this embodiment eliminates the horizontal level in the image
  • the pixel points of the distribution retain the pixel points of the feature block at a larger inclination from the horizontal, and separate the contour of the feature block from the contour line of the molten fluid.
  • the gray value b (x, y) of the pixel point of the obtained image is determined by the set ⁇ , that is,
  • the pixels of the discrete small area are removed from the image to obtain the processed image B (x, y).
  • the horizontal horizontal pixels of the non-feature block part of the molten fluid contour are basically eliminated, and the complete feature block contour is retained, and the top part of the feature block line
  • the pixels are class-level and inevitably removed, and the feature block is divided into two curves.
  • Step S2042 Calculate the minimum distance, the height difference between the upper end points and the angle difference between the two adjacent curves in the contour of the feature block, and roughen the feature block of the molten fluid outline based on the minimum distance, the height difference between the upper end points and the angle difference Positioning.
  • the feature block To identify and locate the feature block, the feature block must be analyzed, so that the features of the feature block must be digitized.
  • the feature block For the feature block is split into two curves, as shown in Figure 5 shows the possible status of the two curves of the feature block and the inverted feature block, there is a certain feature relationship between the two adjacent curves of the feature block: minimum distance, height of the upper endpoint Poor, poor angle.
  • the minimum distance represents the minimum pixel distance between two adjacent curves
  • the angle difference represents the angle between the equivalent straight line of the two adjacent curves and the horizontal
  • the height difference at the upper end represents the height difference between the highest two points of the two adjacent curves, that is, the ordinate Difference.
  • ⁇ is the angle between the equivalent straight line and the horizontal.
  • X i and Y i are the abscissa and ordinate of the pixel point i on the curve respectively.
  • ⁇ value of the equivalent straight line the minimum square sum of the difference between the actual value and the calculated value of the equivalent straight line is used as the optimization criterion.
  • the expression of the angle ⁇ value is,
  • the gray value of the pixel points of the highest and lowest points of the feature block and the inverted feature block in the image are set to zero, so that the completely retained feature block and the inverted feature block are split into two curves.
  • the image after the above processing is B ′ (x, y).
  • the minimum distance between the two curves of the feature block is always within a certain small threshold, and the upper endpoint is also almost at the same height, that is, the height difference between the upper endpoint is also within the small threshold, and the two curves of the inverted feature block and the feature block are at the angle difference There is a big difference.
  • the block diagram of the realization of feature block recognition is shown in Figure 6. The detailed steps are as follows:
  • step1 There are multiple connected domains in the image B ′ (x, y), and the connected domains are numbered ⁇ r 1 , r 2 , ..., r n ⁇ from left to right (S611).
  • step2 Take a single connected domain r k (S612), calculate the minimum distance between r k and r k + 1 , the expression is,
  • r k (i) is the position of a pixel in the connected domain r k . If d k ⁇ , then continue with the following steps, otherwise go to r k + 1 and perform step step2;
  • step3 If the height difference h k of the upper endpoint satisfies h k ⁇ , then continue with the following steps, otherwise go to r k + 1 and execute step step2;
  • step4 Use formula (6) (7) to calculate the angle difference ⁇ k - ⁇ k + 1 between the connected domains r k and r k + 1. If ⁇ k - ⁇ k + 1 ⁇ is satisfied, the connected domain is marked r k and r k + 1 are the characteristic block curve pair (r k , r k + 1 ) (S613), keep the lower part of (r k , r k + 1 ) flush and go to r k + 1 , and perform the steps Step2, otherwise go to r k + 1 and execute step2 until it traverses all connected domains.
  • step5 Set the pixel gray value of the unlabeled connected domain to zero (S614).
  • the feature block image T (x, y) and the feature block curve pair set and ⁇ (r k , r k + 1 ), ... ⁇ are obtained to complete the feature block recognition and coarse positioning process.
  • Step S2043 the contour of the feature block is completed based on the contour of the molten fluid, and the feature block of the contour of the molten fluid is accurately positioned based on the centroid of the feature block after the contour is completed.
  • the coarse positioning function of the feature block is implemented in the image B (x, y) that has lost a lot of pixel information.
  • This step will complete the feature block contour on the complete contour image E (x, y) and further Filter feature blocks.
  • the image T (x, y) corresponds to the difference between the two highest points of the characteristic block curve pair (r k , r k + 1 )
  • Search for the connected domain in the difference image if there is a connected domain that can connect the feature block curve pair, fill the connected domain into the image T (x, y), complete the contour of the feature block, if it does not exist, then remove the feature Block curve pair.
  • centroid formula to calculate the centroid (x t , y t ) of the feature block, accurately locate the feature block, and achieve fast and precise positioning of the small target feature block in high frame rate video.
  • N is the number of pixels in the connected domain
  • (x n , y n ) is the coordinate of a pixel in the connected domain.
  • the feature block in each frame of the image can be identified and the complete contour of the feature block is retained.
  • this embodiment makes full use of the contour line of the molten fluid is generally in a horizontal state, and the feature block has a large angle between the contour line of the molten fluid.
  • the contour of the feature block is separated from the contour line of the molten fluid, and the separation is accurate. 3. High contour completeness.
  • Step S205 Obtain the features of the feature block, and use the features of the feature block to perform similarity matching between two adjacent frames to obtain the pixel distance that the feature block moves in the two adjacent frames.
  • the features include size features, angle features and Location characteristics.
  • the features of the feature block are first obtained, and the similarity matching is performed between the two adjacent frames by using the features of the feature block.
  • the features of the feature block in this embodiment are size feature, angle feature and position feature, respectively.
  • the size feature indicates the orthogonal axial length of the feature block, that is, the maximum length and width
  • the angle feature indicates the angle between the feature block and the horizontal direction
  • the position feature indicates the position coordinate of the feature block centroid.
  • M is the number of pixels on the boundary of the feature block, and the pixels on the boundary of the feature block are expressed as (x m , y m ).
  • the major semi-axis r 1 and the minor semi-axis r 2 of the equivalent ellipse of the feature block can be calculated, that is, the orthogonal axial length of the feature block, the expression is,
  • T (x, y) is similar to the feature block in the image T ′ (x, y) of the next frame.
  • the algorithm steps are as follows:
  • step1 The horizontal coordinate x t of the feature block centroid of the image T (x, y) is the starting point, and the feature block centroid x ′ t is searched backward along the horizontal coordinate in the image T ′ (x, y).
  • Step2 In the image T ′ (x, y), mark the centroid that satisfies the condition of x ′ t -x t ⁇ l, and l is the threshold set according to observation.
  • step3 In a short time, the shape of the feature block has not changed greatly, and the height of the center of mass will only fluctuate slightly. If the ordinate of the centroid of the mark satisfies
  • step4 Calculate the long semi-axis r 1 ′, short semi-axis r 2 ′ and included angle ⁇ ′ of the marked feature block of the image T ′ (x, y) using equations (11) (12), and convert them into three of the same dimension Parameters (r 1 ′ cos ⁇ ′, r 1 ′ sin ⁇ ′, r ′ 2 ), the feature block parameters are recorded as (x ′ 1 , x ′ 2 , x ′ 3 ), the same is true for the image T (x, y) The feature block parameters are recorded as (x 1 , x 2 , x 3 ).
  • step5 In order to judge the similarity of two feature blocks online in real time, give a similarity coefficient ⁇ , that is,
  • the value range of ⁇ is [-1,1].
  • the feature block with the largest matching similarity ⁇ and satisfying ⁇ > ⁇ in the image T ′ (x, y) is regarded as a successful match, thereby obtaining the horizontal pixel distance S C of the feature block movement, and jumping the image T (x, y) For the next centroid coordinate, continue to step1.
  • S C expression can be written as,
  • the feature block size feature, angle feature, and position feature are used to perform similarity matching on the positioning results of the feature blocks between adjacent frames from multiple angles, so that the tracking and positioning feature blocks can be accurately identified and the features can be further accurately obtained
  • the pixel distance that the block moves in two adjacent frames greatly improves the accuracy of subsequent flow rate detection.
  • step S206 the high-speed camera that collects the video stream of the high-temperature and high-speed molten fluid is calibrated according to the installation parameters on the spot to establish the relationship between the image coordinate system and the world coordinate system.
  • Step S207 the moving distance of the feature block in the world coordinate system is solved, and the calculation formula for calculating the moving distance is:
  • S W is the moving distance of the feature block in the world coordinate system
  • R is the diameter of the outlet hole of the reactor
  • R c is the diameter pixel of the outlet of the reactor on the image
  • S C is the feature block on the molten fluid adjacent to each other The distance of pixels moved in two frames.
  • step S208 the flow rate of the high-temperature and high-speed molten fluid is obtained based on the moving distance and the speed formula.
  • the horizontal velocity V x of the molten iron flow is obtained, that is,
  • the high-temperature molten fluid flow rate detection method collects the video stream of the high-temperature and high-speed molten fluid, decomposes the video stream into a time-series frame image sequence, and extracts the region of the molten fluid of interest in the frame image sequence. Extract the molten fluid contour of the molten fluid area of interest, extract the feature block of the molten fluid contour, and obtain the flow rate of the molten fluid based on the feature block.
  • the feature block is specifically the ripples or shadows that appear during the high-temperature molten fluid high-speed outflow process. There is a technical problem that the flow rate detection accuracy of the molten fluid with high temperature, high speed, and high light is not high.
  • the method By using non-invasive acquisition of high frame rate video stream of high temperature molten fluid outflow, small targets on the molten fluid are accurately tracked and located in real time Significant feature block, so as to realize the flow rate detection process of molten fluid with high temperature, high speed and high light.
  • the method has the advantages of high accuracy, strong stability, long periodicity, and is suitable for high-temperature or excessively high-speed flowing fluids, and has low investment cost.
  • the object of the present invention is to provide a method for achieving fast and precise positioning of the salient feature blocks of small targets in high-frame-rate video in contour images of high-speed outflow of molten fluid.
  • the object of the present invention is to provide a method for realizing the transient matching of salient feature blocks in two adjacent frames of images using the features of the feature blocks on the molten fluid.
  • the purpose of the present invention is that the molten fluid flow rate detection algorithm mentioned in the patent is transplanted in parallel to the parallel programming model and multi-GPU general computing architecture, which greatly improves the algorithm execution efficiency and meets the real-time requirements of flow rate detection.
  • a high-temperature molten fluid flow rate detection system provided by an embodiment of the present invention includes: a high-speed camera capture video unit 10, a video acquisition unit 20, a contour extraction unit 30, and a molten fluid flow rate detection connected in sequence to the high-speed camera capture video unit 10 Unit 40, where,
  • the high-speed camera captures the video unit 10, which is used to collect the video stream of the high-temperature and high-speed molten fluid;
  • the video acquisition unit 20 is used to decompose the video stream into a sequence of frame images in time sequence, and extract the region of molten fluid of interest in the frame image sequence;
  • the contour extraction unit 30 is used to extract the molten fluid contour of the molten fluid area of interest
  • the molten fluid flow rate detection unit 40 is used to extract a feature block of the contour of the molten fluid, and obtain the flow rate of the molten fluid based on the feature block.
  • the feature block is specifically a ripple or a shadow that occurs during the high-speed outflow of the high-temperature molten fluid.
  • the high-speed camera capturing video unit 10 includes a high-speed camera 101 for capturing a video stream of a high-temperature and high-speed molten fluid, a lens dust-proof cleaning module 102 and a high-speed camera air-cooling module 103 provided on the high-speed camera 101, wherein,
  • the lens dust-proof cleaning module 102 is used for cleaning the lens of the high-speed camera 101 when the high-speed camera 101 is not in operation;
  • the high-speed camera air-cooling module 103 is used to completely cover the high-speed camera body, and an external fan is used to accelerate the air flow rate between the air-cooling device and the high-speed camera body to achieve heat dissipation and cooling.
  • the lens dust-proof cleaning device 102 cleans the lens of the high-speed camera when the high-speed camera is not working. After the cleaning is completed, the protective lens cover is automatically covered to prevent dust from falling onto the lens until the high-speed camera starts to work
  • the high-speed camera air-cooling device 103 completely covers the high-speed camera body, and accelerates the air flow rate between the air-cooling device and the high-speed camera body by an external fan to achieve the purpose of heat dissipation and cooling.
  • the most important is the selection of high-speed camera frame rate and resolution parameters.
  • the appropriate frame rate ensures that the feature blocks of the high-speed flowing molten fluid can be tracked without loss, and sufficient resolution can provide more features on the image Block details.
  • the high-speed camera needs to be installed at a certain distance from the molten fluid, and a metal baffle is separated in the middle to reduce the thermal radiation.
  • the camera lens collects video images through the window of the baffle.
  • the molten fluid At the smelting site, during the outflow of molten fluid at the outlet of the reactor, the molten fluid emits strong thermal radiation outwards, and there is more dust on the site.
  • the dust covering the lens of the high-speed camera 101 will cause black shadows in imaging.
  • a lens cooling device 102 and a dust-proof cleaning device 103 must be installed on the lens of the high-speed camera to ensure the long-term stable operation of the high-speed camera in a complex and harsh site.
  • the video capture unit 20 includes a video capture module 201 and a video processing module 202, where the video capture module 201 includes:
  • the A / D module is used to convert the analog signal of the video source of the video unit 10 captured by the high-speed camera into a digital signal;
  • Memory module for storing video signals
  • the video compression module is used to compress the video signal and send the compressed video signal to the video processing module 202;
  • the video processing module 202 includes:
  • the video stream decomposition module is used to divide the compressed video stream passed by the video acquisition module 201 into frame images in order of time;
  • the frame image ROI extraction module is used to extract the fused fluid region of interest in the frame image, and transmit the extracted frame image group to the contour extraction unit 30.
  • the video acquisition module 201 in this embodiment converts the analog signal of the video source of the video unit captured by the high-speed camera 101 by the video acquisition card into a digital signal through the A / D conversion module on the high-speed acquisition card, and then sends it to the onboard memory In the module, after a period of storage, the video compression module on the video capture card compresses a large number of video signals, and sends the compressed video signal to the video processing module 202 for processing. At the same time, the video capture card will The collected video stream opens up a new storage space in the memory to store the current video stream and release the storage space of the previous video stream.
  • the video processing module in this embodiment uses a FPGA (Field-Programmable Gate Array) programmable hardware platform preloaded with the hardware logic circuit of the video stream decomposition program and the frame image ROI extraction program to pass the compressed video from the video acquisition module 201
  • FPGA Field-Programmable Gate Array
  • the stream is divided into time-series frame images, and the molten fluid region of interest in the frame image is extracted to reduce the post-processing time, and then the frame image group after the ROI extraction is sent to the contour extraction unit 30 to be processed
  • the frame image group completes the work flow of the unit.
  • the contour extraction unit and the video acquisition unit of this embodiment are implemented on the same FPGA programmable hardware platform.
  • the contour extraction unit is composed of an image preprocessing module and a single pixel contour extraction module.
  • the molten fluid flow rate detection unit 40 includes a feature block fine positioning module 401, a feature block matching module 402 connected in turn to the feature block fine positioning module 401, and a flow rate detection module 403;
  • the feature block fine positioning module 401 includes a culling-like horizontal pixel sub-module, a feature block recognition and coarse positioning sub-module, and a feature block precise positioning sub-module, wherein,
  • the culling-like horizontal pixel sub-module is used to eliminate the horizontally-distributed pixels in the contour of the molten fluid, to retain the pixels of the feature block with a larger inclination from the horizontal, and to separate the contour of the feature block from the contour line of the molten fluid;
  • the feature block recognition and coarse positioning sub-module is used to calculate the minimum distance, the height difference between the upper endpoints and the angle difference between two adjacent curves in the feature block outline, and based on the minimum distance, the height difference between the upper endpoints and the angle difference to melt Rough positioning of the feature block of the fluid contour;
  • the feature block precise positioning submodule is used to complete the contour of the feature block based on the molten fluid contour, and to accurately locate the feature block of the molten fluid contour based on the centroid of the feature block after completing the contour;
  • the feature block matching module 402 is used to perform similarity matching between the features of the feature block between two adjacent frames to obtain the pixel distance that the feature block moves in the time between two adjacent frames;
  • the flow velocity detection module 403 includes a camera calibration stator module and a flow velocity output sub-module.
  • the camera calibration stator module is used to calibrate the high-speed camera and solve the moving distance of the feature block in the world coordinate system.
  • the flow velocity output sub-module is used to calculate and output the flow velocity of the molten fluid .
  • the molten fluid flow rate detection unit is a key unit of the present invention. Its main purpose is to analyze the feature blocks of each frame and perform similarity matching on the feature blocks in the adjacent two frames of images after identifying and positioning the feature blocks in the contour. To complete the detection process.
  • the unit is a multi-GPU (Graphics Processing Units) as the hardware image platform architecture image processing software system, the hardware platform has a parallel computing processing architecture, real-time detection of molten fluid flow rate.
  • the unit is composed of three modules: feature block fine positioning module 401, feature block matching module 402 and flow velocity detection module 403.
  • the flow rate detection module 403 is the final result output part of the embodiment of the present invention. Its function is to use the first two modules of the molten fluid flow rate detection unit to process a frame of image groups and integrate the processing to calculate the flow rate of the molten fluid according to the actual site conditions.
  • the module is composed of two functional sub-modules: the camera stator sub-module and the flow velocity output sub-module.
  • the present invention is applied to a domestic 2650m 3 blast furnace.
  • High-speed cameras and other devices are installed at one of the three tapholes on the blast furnace according to FIG. 8.
  • the appropriate frame rate can capture the motion information of the molten iron flow, and sufficient resolution provides more detailed information on the image.
  • the molten iron flow rate of the blast furnace outlet is 5 ⁇ 6m / s.
  • the resolution of high-speed cameras is 1280 * 720, and the frame rate is 240.
  • the specific implementation steps of the complete detection process of the molten fluid flow rate are as follows:
  • Step 1 According to the installation parameters and field data of the high-speed camera, calibrate the camera to determine the relationship between the image coordinate system and the world coordinate system;
  • Step 2 During a period of tapping the blast furnace, the high-speed camera video capture unit captures the molten iron outflow video in real time.
  • the video collection unit 20 collects the molten iron outflow video in one minute in real time, compresses it and sends it to the video processing unit, and then the video collection unit 20 re-collects the current video and repeats the process;
  • Step 3 The video processing unit decomposes the compressed video of the molten iron into frame images, and extracts the molten iron region of interest in the frame image to obtain the group of frame images to be processed;
  • the image preprocessing process (U51) is completed, and then through the hot metal flow contour extraction process completed by the Canny contour detection algorithm and skeleton extraction processing, a single Pixel hot metal flow contour image for subsequent processing;
  • Step 5 Perform horizontal pixel culling on the obtained binary molten iron contour image, and finally split the target feature block into two adjacent curves. For these two adjacent curves, there is a minimum distance, angle difference and The three characteristics of the height difference of the endpoints, after analyzing the unique shape features of the salient feature block, the feature block recognition and coarse positioning process can be completed, and then the results of the rough positioning of the molten iron feature block are selectively filtered according to the precise block positioning module , Remove the feature blocks that have been identified incorrectly, and realize the fine positioning of the feature blocks;
  • Step 6 Using the three characteristics of the size, angle and position of the feature block, perform a transient matching process on the feature block on the adjacent two frames of images to match the feature block with similarity ⁇ > 0.6, and obtain the feature block from equation (14) The horizontal pixel distance that moves within two frames of time. Then, according to step 1, the camera calibration acquires the relationship between the image coordinate system and the world coordinate system, and converts the pixel distance into the real world distance;
  • Step 7 after processing a frame image group in the above steps, remove all the outliers in the distance data and take the average value, according to the elevation angle at the outlet of the reactor, to get the average of the outflow of molten iron within the time period of the current frame image group Flow rate value.
  • the specific working process and working principle of the high-temperature molten fluid flow rate detection system of this embodiment can refer to the working process and working principle of the high-temperature molten fluid flow rate detection method in this embodiment.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Multimedia (AREA)
  • Fluid Mechanics (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Image Analysis (AREA)

Abstract

一种高温熔融流体流速检测方法及系统，通过采集高温高速熔融流体的视频流(S101)，将视频流分解成以时间为序的帧图像序列，并提取帧图像序列中的感兴趣熔融流体区域(S102)，提取感兴趣熔融流体区域的熔融流体轮廓(S103)，提取熔融流体轮廓的特征块，并基于特征块获取熔融流体的流速(S104)，解决了对具有高温、高速、高光的熔融流体的流速检测精度不高的技术问题，通过利用非侵入式获取高温熔融流体出流的高帧率视频流，通过实时精确跟踪定位熔融流体上的小目标显著性特征块，从而实现对具有高温、高速、高光的熔融流体的流速检测。该方法及系统具有高精确性，强稳定性，长周期性，适用于高温高速流体，具有投资成本少等优点。

Description

一种高温熔融流体流速检测方法及系统 技术领域
本发明主要涉及高温熔融流体流速检测技术领域，特指一种高温熔融流体流速检测方法及系统。
背景技术
冶金行业中，黑色金属和有色金属多是在高温条件下通过火法冶金的方法从矿石或精矿中提取，产出熔融态的粗金属或金属富集物和炉渣从高温密闭的反应炉内高速流出，如在钢铁工业流程中铁水从高炉出铁口流出，在火法炼铜工业流程中冰铜从反射炉流出，在火法炼锌流程中粗锌从熔炼炉流出等。反应炉内压力的变化趋势是表征反应炉是否平稳顺行的重要指标，但由于反应炉内环境恶劣导致难以直接检测反应炉内压力的变化情况，检测反应炉出口处熔融流体的流速能表征反应炉炉内的压力，同时也能反映产出的金属和渣之间的比例关系，助于及时发现并排除异常工况，改善反应炉透气性，保证反应炉平稳顺行生产。因此，检测反应炉出口处熔融流体的流速对于反应炉安全生产、提质提量的意义尤为重要。
本发明检测对象为反应炉出口处出流的高温高光熔融流体，同时检测现场存在不可避免的震动及大量分布不均的粉尘，极大地增加了检测的难度。现有技术中，其主流检测的方法分为接触式测量法和非接触式测量法两种，检测高温熔融流体的接触式检测方法需要采用耐高温的材质与高温熔融流体直接接触，高速流动的高温流体会逐渐磨损侵蚀耐高温材质，导致装置重复性差、使用周期短，不仅如此，在大量粉尘及高温的恶劣环境下同样会影响装置的使用寿命和使用性能，投资成本高，检测准确度低，使得接触式检测方法受到了较大的限制；非接触式的检测方法主要是通过建立机理模型的方式来达到非侵入式地检测流速目的，但同样会因检测对象的超高温性及环境的恶劣性严重影响检测的准确性。
专利公开号CN103480813A发明专利是一种连铸结晶器高温钢液流速测量装置及测量方法，其工作原理是轴承通过固定轴固定在固定装置上，轴承上、下对称位置分别安装弹簧和测量杆，弹簧安装“T”型固定装置上，轴承安装轴承套，通过联轴器连接轴承套和角度位移感器，角度位移传感器由电源供电，记录测量杆在流动钢液中的实时偏转角度，并通过数据线传输到数据采集分析系统，将角度数据转化为钢液流速值。但该专利根据装置检测对象的不同进行重新配准，使用前需将装置预热至1200～1400℃，使用复杂，检测量程较小，对于流速过大的流体检测误差较大，且检测结束后装置不能直接检测下一个对象，在使用的可重复性上受到限制。
专利公开号CN104131126A发明专利是一种基于模糊模型的高炉熔渣流量检测方法，该专 利建立高炉熔渣流量的模糊推理模型，结合第i时刻的渣面高度的大小对高炉熔渣流量的影响特性，设定关于第i时刻的渣面高度的模糊隶属函数，利用模糊推理模型与模糊隶属函数，建立高炉熔渣流量计算模型，使用高炉熔渣流量计算模型进行高炉实时熔渣总流量的在线检测。但该专利中的初始值是通过工艺人员由人工操作经验知识中获得的，人为因素影响较大，主观性强，且设计流程是一个开环，无法保证长期运行结果的准确性。
发明内容
本发明提供的高温熔融流体流速检测方法及系统，解决了现有对具有高温、高速、高光的熔融流体的流速检测精度不高的技术问题。
为解决上述技术问题，本发明提出的高温熔融流体流速检测方法包括：
采集高温高速熔融流体的视频流；
将视频流分解成以时间为序的帧图像序列，并提取帧图像序列中的感兴趣熔融流体区域；
提取感兴趣熔融流体区域的熔融流体轮廓；
提取熔融流体轮廓的特征块，并基于特征块获取熔融流体的流速，特征块具体为高温熔融流体高速出流过程中出现的波纹或阴影。
进一步地，提取感兴趣熔融流体区域的熔融流体轮廓包括：
对感兴趣熔融流体区域进行预处理，获得预处理图像；
采用一阶偏导有限差分计算预处理图像的梯度幅值和方向，并对梯度幅值进行非极大值抑制；
采用双阈值算法检测和连接感兴趣熔融流体区域的熔融流体轮廓，并利用骨架提取算法细化熔融流体轮廓。
进一步地，提取熔融流体轮廓的特征块包括：
剔除熔融流体轮廓中类水平分布的像素点，保留与水平成预设倾角的特征块像素点，从熔融流体轮廓中分离出特征块轮廓；
计算特征块轮廓中两条相邻曲线之间的最小距离、上端点高度差以及夹角差，并基于最小距离、上端点高度差以及夹角差对熔融流体轮廓的特征块进行粗定位；
基于熔融流体轮廓对特征块的轮廓进行补全，并基于补全轮廓后的特征块的质心，对熔融流体轮廓的特征块进行精确定位。
进一步地，基于特征块获取熔融流体的流速包括：
获取特征块的特征，特征包括大小特征、角度特征和位置特征；
在相邻两帧之间利用特征块的特征进行相似性匹配，得到特征块在相邻两帧时间内移动的像素距离；
根据现场的安装参数对采集高温高速熔融流体的视频流的高速相机进行标定，建立图像坐标系与世界坐标系之间的关系；
求解特征块在世界坐标系中的移动距离，且计算移动距离的计算公式为：
Figure PCTCN2019096143-appb-000001
其中S W为特征块在世界坐标系中的移动距离，R为反应炉出口圆孔的直径，R c为反应炉出口在图像上的直径像素，S C为熔融流体上的特征块在相邻两帧时间内移动的像素距离；
基于移动距离以及速度公式获得高温高速熔融流体的流速。
进一步地，获取特征块的特征包括：
计算特征块边界点的i+j阶矩，计算公式具体为：
Figure PCTCN2019096143-appb-000002
其中μ ij为特征块边界点的i+j阶矩，M为特征块边界像素点的个数，m＝0,2,…,M，(x m,y m)为特征块边界上第m个像素点，(x t,y t)为质心坐标；
根据i+j阶矩计算特征块的等效椭圆的长半轴r 1和短半轴r 2，计算公式具体为：
Figure PCTCN2019096143-appb-000003
其中，r 1和r 2分别为特征块的等效椭圆的长半轴和短半轴，μ 2,0为i＝2，j＝0时μ ij的取值，μ 0,2为i＝0，j＝2时μ ij的取值，μ 1,1为i＝1，j＝1时μ ij的取值；
基于i+j阶矩计算特征块的等效椭圆与水平方向的夹角，计算公式具体为：
Figure PCTCN2019096143-appb-000004
其中β为特征块的等效椭圆与水平方向的夹角；
利用质心公式计算特征块的质心，具体为：
Figure PCTCN2019096143-appb-000005
其中(x t,y t)为特征块的质心，N为连通域像素点的个数，(x n,y n)为连通域中第n个像素点的坐标；
根据长半轴r 1、短半轴r 2、夹角β和质心(x t,y t)，获得特征块的特征。
进一步地，在相邻两帧之间利用特征块的特征进行相似性匹配，得到特征块在相邻两帧时间内移动的像素距离包括：
step1：以前一帧图像T(x,y)的特征块质心的横坐标x t为起点，在后一帧图像T′(x,y)沿横坐标向后搜索特征块质心x′ t；
step2：在图像T′(x,y)中，将满足x′ t-x t＜l条件的质心标记，l为根据观察设定的阈值；
step3：判断标记的质心纵坐标是否满足|y′ t-y t|≤h′，其中h′为设定的阈值，y′ t和y t分别为后一帧图像特征块质心纵坐标和前一帧图像特征块质心纵坐标，并在判定为否时去除特征块的标记，并跳转图像T(x,y)的下一个质心坐标，继续步骤step1；
step4：计算图像T′(x,y)被标记特征块的长半轴r 1′、短半轴r 2′和夹角β′，转化成量纲相同的三个参数(r 1′cosβ′,r 1′sinβ′,r′ 2)，特征块参数记为(x′ 1,x′ 2,x′ 3)，同理，图像T(x,y)的特征块参数记为(x 1,x 2,x 3)；
step5：为了在线实时评判两个特征块的相似度，计算相似度系数，具体为：
Figure PCTCN2019096143-appb-000006
其中ρ为相似度系数，ρ的取值范围为[-1,1]，x k为图像T(x,y)的第k个特征块参数，x′ k为图像T′(x,y)的第k个特征块参数，图像T′(x,y)中匹配相似度ρ最大且满足ρ＞ξ的特征块则认定为匹配成功，从而得到特征块移动的水平像素距离S C，跳转图像T(x,y)的下一个质心坐标，继续步骤step1，其中S C的计算公式为：
S C＝x′ t-x t。
本发明提出的高温熔融流体流速检测系统包括：
高速相机捕捉视频单元、与高速相机捕捉视频单元依次连接的视频采集单元、轮廓提取单元以及熔融流体流速检测单元，其中，
高速相机捕捉视频单元，用于采集高温高速熔融流体的视频流；
视频采集单元，用于将视频流分解成以时间为序的帧图像序列，并提取帧图像序列中的感兴趣熔融流体区域；
轮廓提取单元，用于提取感兴趣熔融流体区域的熔融流体轮廓；
熔融流体流速检测单元，用于提取熔融流体轮廓的特征块，并基于特征块获取熔融流体的流速，特征块具体为高温熔融流体高速出流过程中出现的波纹或阴影。
进一步地，高速相机捕捉视频单元包括用于采集高温高速熔融流体的视频流的高速相机以及设置于高速相机上的镜头防尘清扫模块以及高速相机风冷模块，其中，
镜头防尘清扫模块，用于在高速相机未工作时对高速相机的镜头进行清扫处理；
高速相机风冷模块，用于将高速相机机身完全覆盖，通过外接风机加快风冷装置与高速相机机身之间空气的流动速率达到散热冷却。
进一步地，视频采集单元包括视频采集模块和视频处理模块，其中视频采集模块包括：
A/D模块，用于将高速相机捕捉视频单元的视频源模拟信号转换成数字信号；
存储器模块，用于存储视频信号；
视频压缩模块，用于对视频信号进行压缩，并将已压缩的视频信号送至视频处理模块；
视频处理模块包括：
视频流分解模块，用于将视频采集模块传递过来的压缩视频流分割成以时间为序的帧图像；
帧图像感兴趣区域提取模块，用于提取帧图像中感兴趣熔融流体区域，并将提取后的帧图像组传送至轮廓提取单元。
进一步地，熔融流体流速检测单元包括特征块精细定位模块、与特征块精细定位模块依次连接的特征块匹配模块以及流速检测模块；
特征块精细定位模块包括剔除类水平像素子模块、特征块识别及粗定位子模块、特征块精确定位子模块，其中，
剔除类水平像素子模块用于剔除熔融流体轮廓中类水平分布的像素点，保留与水平成较大倾角的特征块像素点，从熔融流体轮廓线条中分离出特征块轮廓；
特征块识别及粗定位子模块用于计算特征块轮廓中两条相邻曲线之间的最小距离、上端点高度差以及夹角差，并基于最小距离、上端点高度差以及夹角差对熔融流体轮廓的特征块进行粗定位；
特征块精确定位子模块用于基于熔融流体轮廓对特征块的轮廓进行补全，并基于补全轮廓后的特征块的质心，对熔融流体轮廓的特征块进行精确定位；
特征块匹配模块，用于在相邻两帧之间利用特征块的特征进行相似性匹配，得到特征块在相邻两帧时间内移动的像素距离；
流速检测模块包括相机标定子模块和流速输出子模块，相机标定子模块用于对高速相机进行标定并求解特征块在世界坐标系移动距离，流速输出子模块用于计算并输出熔融流体的流速。
与现有技术相比，本发明的优点在于：
本发明提供的高温熔融流体流速检测方法及系统，通过采集高温高速熔融流体的视频流，将视频流分解成以时间为序的帧图像序列，并提取帧图像序列中的感兴趣熔融流体区域，提取感兴趣熔融流体区域的熔融流体轮廓，提取熔融流体轮廓的特征块，并基于特征块获取熔融流体的流速，特征块具体为高温熔融流体高速出流过程中出现的波纹或阴影，解决了现有对具有高温、高速、高光的熔融流体的流速检测精度不高的技术问题，通过利用非侵入式获取高温熔融流体出流的高帧率视频流，通过实时精确跟踪定位熔融流体上的小目标显著性特征块，从而实现对具有高温、高速、高光的熔融流体的流速检测过程。该方法及系统具有高精确性，强稳定性，长周期性，适用于高温或过高温的高速流动的流体，投资成本少等优点。
附图说明
图1为本发明实施例一的高温熔融流体流速检测方法的流程图；
图2为本发明实施例二的高温熔融流体流速检测方法的流程图；
图3为本发明实施例二的提取感兴趣熔融流体区域的熔融流体轮廓方法的流程图；
图4为本发明实施例二的类水平像素剔除模块的矩阵示意图；
图5为本发明实施例二的特征块和倒置特征块相邻曲线的状态图；
图6为本发明实施例二的特征块精细定位模块的程序框图；
图7为本发明的高温熔融流体流速检测系统的结构框图；
图8为本发明的熔融流体流速检测装置及现场示意图。
附图标记：
10、高速相机捕捉视频单元；20、视频采集单元；30、轮廓提取单元；40、熔融流体流速检测单元；101、高速相机；102、镜头防尘清扫模块；103、高速相机风冷模块；201、视频采集模块；202、视频处理模块；401、特征块精细定位模块；402、特征块匹配模块；403、流速检测模块。
具体实施方式
为了便于理解本发明，下文将结合说明书附图和较佳的实施例对本发明作更全面、细致地描述，但本发明的保护范围并不限于以下具体的实施例。
以下结合附图对本发明的实施例进行详细说明，但是本发明可以由权利要求限定和覆盖的多种不同方式实施。
实施例一
参照图1，本发明实施例一提供的高温熔融流体流速检测方法，包括：
步骤S101，采集高温高速熔融流体的视频流；
步骤S102，将视频流分解成以时间为序的帧图像序列，并提取帧图像序列中的感兴趣熔融流体区域；
步骤S103，提取感兴趣熔融流体区域的熔融流体轮廓；
步骤S104，提取熔融流体轮廓的特征块，并基于特征块获取熔融流体的流速，特征块具体为高温熔融流体高速出流过程中出现的波纹或阴影。
本发明实施例提供的高温熔融流体流速检测方法，通过采集高温高速熔融流体的视频流，将视频流分解成以时间为序的帧图像序列，并提取帧图像序列中的感兴趣熔融流体区域，提取感兴趣熔融流体区域的熔融流体轮廓，提取熔融流体轮廓的特征块，并基于特征块获取熔融流体的流速，特征块具体为高温熔融流体高速出流过程中出现的波纹或阴影，解决了现有对具有高温、高速、高光的熔融流体的流速检测精度不高的技术问题，通过利用非侵入式获取高温熔融流体出流的高帧率视频流，通过实时精确跟踪定位熔融流体上的小目标显著性特征块，从而实现对具有高温、高速、高光的熔融流体的流速检测过程。
具体地，由于现有技术不管是采用接触式还是非接触式检测高温熔融流体流速视，都会因检测对象的超高温性及环境的恶劣性严重影响检测的准确性，而本发明实施例较新颖地提出通过提取熔融流体轮廓的特征块，也即高温熔融流体高速出流过程中出现的波纹或阴影，并通过实时精确跟踪定位熔融流体上的小目标显著性特征块，从而实现对具有高温、高速、高光的熔融流体的流速检测过程。该方法具有高精确性，强稳定性，长周期性，适用于高温或过高温的高速流动的流体，投资成本少等优点。
实施例二
参照图2，本发明实施例二提供的高温熔融流体流速检测方法，包括：
步骤S201，采集高温高速熔融流体的视频流。
具体地，本发明实施例是通过高速相机采集高温高速熔融流体的视频流。
步骤S202，将视频流分解成以时间为序的帧图像序列，并提取帧图像序列中的感兴趣熔融流体区域。
具体地，本实施例通过高速相机采集到高温高速熔融流体对象的视频流信息后；将所采集的一段时间视频流分解成以时间为序的帧图像序列并提取帧图像序列中的感兴趣熔融流体区域；随后，对感兴趣区域进行图像增强操作并通过单像素轮廓提取模块提取出熔融流体特征块的轮廓，得到熔融流体的二值帧轮廓图像。
步骤S203，提取感兴趣熔融流体区域的熔融流体轮廓。
具体地，首先对帧图像组进行预处理，包含灰度化处理及图像增强处理。对于彩色图像，存在多种色彩空间，其中采用RGB数据处理最方便转换至灰度空间。对传输至FPGA的待处理 帧图像组信号首先经过RGB转灰度空间变换，得到一路灰度信号供后续处理，转换公式如下，
gray＝0.3×R+0.59×G+0.11×B      (1)
图像的边缘在图像的频域上对应于图像傅里叶变换的高频部分，而图像的背景区域则对应于低频部分，因此可利用频域高通滤波法使图像中的高频分量顺利通过，以增强图像的细节信息。图像增强程序是通过指数高通滤波器实现的，其传递函数为：
H(u,v)＝exp{-[D 0/D(u,v)] n}       (2)
其中，D(u,v)是距gray(x,y)傅里叶变换中心原点的距离，D 0为截止频率。图像gray(x,y)傅里叶变换后得到Gray(u,v)，与H(u,v)乘积后进行傅里叶反变换，得到高通滤波的图像，即，
Figure PCTCN2019096143-appb-000007
本发明实施例对帧图像组进行预处理后，提取感兴趣熔融流体区域的单像素轮廓。在经典轮廓检测算子中，Canny算子在检测轮廓方面，具有定位精度高、检测准确、抗干扰能力强等优点，但提取的轮廓为非单像素轮廓，为此，在Canny算子输出后加入形态学处理的骨架提取程序以细化轮廓。本发明实施例提取感兴趣熔融流体区域的单像素轮廓主要包括高斯滤波U521、梯度及梯度方向计算U522、非极大值抑制U523、双阈值分割U524和骨架提取U525五个步骤，如图3所示，首先用高斯滤波器平滑图像以减少噪声的干扰，然后用一阶偏导有限差分计算图像的梯度幅值和方向并对梯度幅值进行非极大值抑制，使用双阈值算法检测和连接轮廓，最后利用骨架提取算法细化轮廓。
步骤S204，提取熔融流体轮廓的特征块。
具体地，本发明实施例提取熔融流体轮廓的特征块包括：
步骤S2041，剔除熔融流体轮廓中类水平分布的像素点，保留与水平成预设倾角的特征块像素点，从熔融流体轮廓中分离出特征块轮廓。
具体地，由于本实施例检测的熔融流体的轮廓线条大致处于水平状态，而特征块与熔融流体的轮廓线条之间有较大的夹角，对此特点，本实施例通过剔除图像中类水平分布的像素点，保留与水平成较大倾角的特征块像素点，从熔融流体轮廓线条分离出特征块的轮廓。
二值化熔融流体图像E(x,y)，以图像E(x,y)中的像素点e(x,y)为中心，与其八邻域组成3×3的灰度值矩阵X，设定权重矩阵A，如图4，图4中(a)为灰度值矩阵X；(b)为权重矩阵A，搜寻目标像素点集合Ω，表达式为：
Figure PCTCN2019096143-appb-000008
得到图像的像素点灰度值b(x,y)被集合Ω决定，即，
Figure PCTCN2019096143-appb-000009
对图像进行离散小区域像素点去除，得到处理后的图像B(x,y)，熔融流体轮廓非特征块部分的类水平像素点基本消除，保留了完整的特征块轮廓，特征块顶部部分线条像素是类水平的，不可避免地被去除，特征块被分成了两条曲线。
步骤S2042，计算特征块轮廓中两条相邻曲线之间的最小距离、上端点高度差以及夹角差，并基于最小距离、上端点高度差以及夹角差对熔融流体轮廓的特征块进行粗定位。
要对特征块进行识别定位须对特征块进行特征分析，从而须将特征块的特征数字化。对于特征块被分裂成两条曲线，如图5显示了特征块及倒置特征块的两条曲线可能状态，特征块的两条相邻曲线之间有一定的特征关系：最小距离、上端点高度差、夹角差。其中最小距离表示相邻两曲线最小的像素距离，夹角差表示相邻两曲线的等效直线与水平的夹角，上端点高度差表示相邻两曲线最高两点的高度差，即纵坐标之差。
为计算夹角差特征关系，现假设曲线的等效直线为，
Y j＝tan(θ)X i+a        (6)
其中θ为等效直线与水平的夹角。X i,Y i分别为曲线上像素点i的横坐标和纵坐标，为确定等效直线的θ值，将实际值与等效直线的计算值的差分平方和最小作为优化判据，求取夹角θ值的表达式为，
Figure PCTCN2019096143-appb-000010
将图像中完整保留下来的特征块和倒置的特征块最高点及最低点的像素点灰度值置为零，使完整保留下来的特征块和倒置的特征块分裂成了两条曲线。经以上处理后的图像为B′(x,y)。特征块的两条曲线最小距离始终在一定的小阈值内，上端点也是近乎处于同一高度，即上端点高度差也处于小阈值内，同时倒置特征块和特征块的两曲线在夹角差上有较大的差别，对此，特征块识别的实现程序框图如图6所示，详细步骤如下：
step1:图像B′(x,y)中的有多个连通域，并对连通域从左至右编号为{r 1,r 2,…,r n}(S611)。
step2:取单个连通域r k(S612)，计算r k和r k+1之间的最小距离，表达式为，
d k＝min||r k(i)-r k+1(j)||       (8)
其中，r k(i)是连通域r k中的一个像素点位置，若满足d k＜λ，则继续下面步骤，否则转 到r k+1，执行步骤step2；
step3:若上端点高度差h k满足h k＜τ，则继续下面步骤，否则转到r k+1，执行步骤step2；
step4:利用公式(6)(7)计算连通域r k和r k+1之间的夹角差θ k-θ k+1，若满足θ k-θ k+1＜α，则标记连通域r k和r k+1为特征块曲线对(r k,r k+1)(S613),保持(r k,r k+1)的下部分平齐并转到r k+1，执行步骤step2，否则转到r k+1，执行步骤step2，直至遍历所有连通域后结束。
step5:将未被标记的连通域的像素点灰度值置为零(S614)。
识别特征块的两条曲线后得到特征块图像T(x,y)以及特征块曲线对集和{(r k,r k+1),…}，完成特征块的识别及粗定位过程。
步骤S2043，基于熔融流体轮廓对特征块的轮廓进行补全，并基于补全轮廓后的特征块的质心，对熔融流体轮廓的特征块进行精确定位。
上一个步骤中在损失了大量像素信息的图像B(x,y)中实现了特征块的粗定位功能，本步骤将在完整的轮廓图像E(x,y)上补全特征块轮廓及进一步筛选特征块。将图像E(x,y)和图像T(x,y)作差，图像T(x,y)对应到差分图像中，特征块曲线对(r k,r k+1)的两最高点之间在差分图像中搜寻连通域，若存在连通域能连通特征块曲线对，将该连通域填充到图像T(x,y)中，补全特征块的轮廓，若不存在，则剔除该特征块曲线对。
最后，利用质心公式计算特征块的质心(x t,y t)，对特征块精确定位，实现在高帧率视频中小目标特征块的快速精细定位，表达式为，
Figure PCTCN2019096143-appb-000011
这里N为连通域像素点的个数，(x n,y n)为连通域中一个像素点的坐标。
本发明实施例通过提取熔融流体轮廓的特征块，可以识别每一帧图像中的特征块并将特征块完整轮廓保留下来。且本实施例充分利用熔融流体的轮廓线条大致处于水平状态，且特征块与熔融流体的轮廓线条之间有较大的夹角的特点从熔融流体轮廓线条分离出特征块的轮廓，且分离准确、轮廓完整度高。
步骤S205，获取特征块的特征，并在相邻两帧之间利用特征块的特征进行相似性匹配，得到特征块在相邻两帧时间内移动的像素距离，特征包括大小特征、角度特征和位置特征。
具体地，本实施例为了得到特征块在相邻两帧时间内移动的像素距离，首先获取特征块的特征，并利用特征块的特征在相邻两帧之间进行相似性匹配。本实施例中特征块的特征分别是大小特征、角度特征和位置特征。大小特征表示特征块的正交轴向长度，即最大长度和宽度，角度特征表示特征块与水平方向的夹角，位置特征表示特征块质心的位置坐标。
为计算大小特征及角度特征，先计算特征块边界点的i+j阶矩，其定义为，
Figure PCTCN2019096143-appb-000012
这里M为特征块边界像素点的个数，特征块边界上的像素点表示为(x m,y m)。根据i+j阶矩可计算特征块等效椭圆的长半轴r 1和短半轴r 2，即特征块的正交轴向长度，表达式为，
Figure PCTCN2019096143-appb-000013
利用i+j阶矩还可得到特征块等效椭圆的与水平方向的夹角β，即特征块的角度特征，其表达式为，
Figure PCTCN2019096143-appb-000014
根据长半轴r 1、短半轴r 2、夹角β和质心(x t,y t)等特征块特征，即可确定特征块的位置形状和方向，利用以下搜寻算法确定前一帧图像T(x,y)与后一帧图像T′(x,y)中相似的特征块，该算法步骤如下：
step1：图像T(x,y)特征块质心的横坐标x t为起点，在图像T′(x,y)沿横坐标向后搜索特征块质心x′ t。
step2：在图像T′(x,y)中，将满足x′ t-x t＜l条件的质心标记，l为根据观察设定的阈值。
step3：短时间内，特征块的形状没有发生较大的变化，质心的高度只会微小波动。标记的质心纵坐标若存在满足|y′ t-y t|≤h′，表明两特征块可能相似，将不满足条件的特征块的标记去除，若不存在，去除所有标记，并跳转图像T(x,y)的下一个质心坐标，继续步骤step1。
step4：利用公式(11)(12)计算图像T′(x,y)被标记特征块的长半轴r 1′、短半轴r 2′和夹角β′，转化成量纲相同的三个参数(r 1′cosβ′,r 1′sinβ′,r′ 2)，特征块参数记为(x′ 1,x′ 2,x′ 3)，同理，图 像T(x,y)的特征块参数记为(x 1,x 2,x 3)。
step5：为了在线实时评判两个特征块的相似度，给出一个相似度系数ρ，即，
Figure PCTCN2019096143-appb-000015
ρ的取值范围为[-1,1]，ρ的值越大，相似性越好。图像T′(x,y)中匹配相似度ρ最大且满足ρ＞ξ的特征块则认定为匹配成功，从而得到特征块移动的水平像素距离S C，跳转图像T(x,y)的下一个质心坐标，继续步骤step1。其中S C表达式可写为，
S C＝x′ t-x t       (14)
本实施例通过利用特征块大小特征、角度特征和位置特征，从多角度对特征块定位结果在相邻帧之间进行相似性匹配，从而能精准地识别跟踪定位特征块，进一步精准地获取特征块在两相邻帧时间内移动的像素距离，大大提高了后续流速检测的精确性。
步骤S206，根据现场的安装参数对采集高温高速熔融流体的视频流的高速相机进行标定，建立图像坐标系与世界坐标系之间的关系。
步骤S207，求解特征块在世界坐标系中的移动距离，且计算移动距离的计算公式为：
Figure PCTCN2019096143-appb-000016
其中S W为特征块在世界坐标系中的移动距离，R为反应炉出口圆孔的直径，R c为反应炉出口在图像上的直径像素，S C为熔融流体上的特征块在相邻两帧时间内移动的像素距离。
步骤S208，基于移动距离以及速度公式获得高温高速熔融流体的流速。
本实施例通过图像坐标系与世界坐标系之间的关系，可以得到相邻两帧图像中多个特征块在真实世界坐标系移动的水平距离S Wj，此处j＝1,2,…,J，对S Wj取平均值，即，
Figure PCTCN2019096143-appb-000017
根据速度公式得到铁水流水平方向的流速V x，即，
Figure PCTCN2019096143-appb-000018
这里f是相机拍摄视频的帧率。根据现场出铁口的实际仰角φ，出铁口铁水流速表达为，
Figure PCTCN2019096143-appb-000019
从而最终完成整个工作流程。
本发明实施例提供的高温熔融流体流速检测方法，通过采集高温高速熔融流体的视频流，将视频流分解成以时间为序的帧图像序列，并提取帧图像序列中的感兴趣熔融流体区域，提取感兴趣熔融流体区域的熔融流体轮廓，提取熔融流体轮廓的特征块，并基于特征块获取熔融流体的流速，特征块具体为高温熔融流体高速出流过程中出现的波纹或阴影，解决了现有对具有高温、高速、高光的熔融流体的流速检测精度不高的技术问题，通过利用非侵入式获取高温熔融流体出流的高帧率视频流，通过实时精确跟踪定位熔融流体上的小目标显著性特征块，从而实现对具有高温、高速、高光的熔融流体的流速检测过程。该方法具有高精确性，强稳定性，长周期性，适用于高温或过高温的高速流动的流体，投资成本少等优点。
本发明的目的在于提供一种在熔融流体高速出流的轮廓图像中实现高帧率视频中小目标显著性特征块快速精细化定位的方法。本发明的目的在于提供一种利用熔融流体上特征块的特征在相邻两帧图像中实现显著性特征块瞬态匹配的方法。本发明的目的在于专利中提到的熔融流体流速检测算法，平行的移植到并行编程模型和多GPU通用计算架构上，极大的提高算法执行效率，满足了流速检测的实时性要求。
参照图7，本发明实施例提出的高温熔融流体流速检测系统，包括：高速相机捕捉视频单元10、与高速相机捕捉视频单元10依次连接的视频采集单元20、轮廓提取单元30以及熔融流体流速检测单元40，其中，
高速相机捕捉视频单元10，用于采集高温高速熔融流体的视频流；
视频采集单元20，用于将视频流分解成以时间为序的帧图像序列，并提取帧图像序列中的感兴趣熔融流体区域；
轮廓提取单元30，用于提取感兴趣熔融流体区域的熔融流体轮廓；
熔融流体流速检测单元40，用于提取熔融流体轮廓的特征块，并基于特征块获取熔融流体的流速，特征块具体为高温熔融流体高速出流过程中出现的波纹或阴影。
可选地，高速相机捕捉视频单元10包括用于采集高温高速熔融流体的视频流的高速相机101以及设置于高速相机101上的镜头防尘清扫模块102以及高速相机风冷模块103，其中，
镜头防尘清扫模块102，用于在高速相机101未工作时对高速相机101的镜头进行清扫处理；
高速相机风冷模块103，用于将高速相机机身完全覆盖，通过外接风机加快风冷装置与高速相机机身之间空气的流动速率达到散热冷却。
本实施例中镜头防尘清扫装置102在高速相机未工作时对高速相机的镜头进行清扫处理，清扫结束后，自动将保护镜头盖盖上以防止粉尘的掉落至镜头，直至高速相机开始工作；高 速相机风冷装置103是将高速相机机身完全覆盖，通过外接风机加快风冷装置与高速相机机身之间空气的流动速率达到散热冷却的目的。为满足测速需求，最主要是高速相机帧率和分辨率的参数选择，合适的帧率保证能跟踪高速流动熔融流体的特征块而不丢失，足够的分辨率在图像上能提供更多的特征块细节信息。其次，高速相机需安装于与熔融流体相距一定的位置，中间有一个金属挡板隔开以减弱热辐射，相机镜头通过挡板的窗口采集视频图像。
在冶炼现场，反应炉出口熔融流体出流的过程中，熔融流体向外散发强烈的热辐射，而且现场存在较多的粉尘，粉尘覆盖在高速相机101的镜头将造成成像出现黑影，对此必须在高速相机的镜头安装镜头冷却装置102和防尘清扫装置103，以保证高速相机在复杂恶劣的现场长期稳定运行的要求。
可选地，视频采集单元20包括视频采集模块201和视频处理模块202，其中视频采集模块201包括：
A/D模块，用于将高速相机捕捉视频单元10的视频源模拟信号转换成数字信号；
存储器模块，用于存储视频信号；
视频压缩模块，用于对视频信号进行压缩，并将已压缩的视频信号送至视频处理模块202；
视频处理模块202包括：
视频流分解模块，用于将视频采集模块201传递过来的压缩视频流分割成以时间为序的帧图像；
帧图像感兴趣区域提取模块，用于提取帧图像中感兴趣熔融流体区域，并将提取后的帧图像组传送至轮廓提取单元30。
本实施例中的视频采集模块201由视频采集卡将高速相机101捕捉视频单元的视频源模拟信号通过高速采集卡上的A/D转换模块转换成数字信号，然后送至板卡自带的存储器模块中，存储一段时间后再由视频采集卡上自带的视频压缩模块将大量的视频信号压缩，并将已压缩的视频信号送至视频处理模块202处理，与此同时，视频采集卡将当前的所采集的视频流在存储器开辟新的存储空间存储当前视频流，并释放前一段视频流的存储空间。
本实施例中的视频处理模块通过一个预载了视频流分解程序及帧图像ROI提取程序硬件逻辑电路的FPGA(Field-Programmable Gate Array)可编程硬件平台，将视频采集模块201传递过来的压缩视频流分割成以时间为序的帧图像的同时，提取出帧图像中感兴趣熔融流体区域，以减少后期处理的时间，然后将ROI提取后的帧图像组送至轮廓提取单元30，得到待处理的帧图像组，完成该单元的工作流程。
为提高装置集成度，本实施例的轮廓提取单元与视频采集单元在同一个FPGA可编程硬件平台实现。且轮廓提取单元由图像预处理模块和单像素轮廓提取模块构成。
可选地，熔融流体流速检测单元40包括特征块精细定位模块401、与特征块精细定位模块401依次连接的特征块匹配模块402以及流速检测模块403；
特征块精细定位模块401包括剔除类水平像素子模块、特征块识别及粗定位子模块、特征块精确定位子模块，其中，
剔除类水平像素子模块用于剔除熔融流体轮廓中类水平分布的像素点，保留与水平成较大倾角的特征块像素点，从熔融流体轮廓线条中分离出特征块轮廓；
特征块识别及粗定位子模块用于计算特征块轮廓中两条相邻曲线之间的最小距离、上端点高度差以及夹角差，并基于最小距离、上端点高度差以及夹角差对熔融流体轮廓的特征块进行粗定位；
特征块精确定位子模块用于基于熔融流体轮廓对特征块的轮廓进行补全，并基于补全轮廓后的特征块的质心，对熔融流体轮廓的特征块进行精确定位；
特征块匹配模块402，用于在相邻两帧之间利用特征块的特征进行相似性匹配，得到特征块在相邻两帧时间内移动的像素距离；
流速检测模块403包括相机标定子模块和流速输出子模块，相机标定子模块用于对高速相机进行标定并求解特征块在世界坐标系移动距离，流速输出子模块用于计算并输出熔融流体的流速。
熔融流体流速检测单元为本发明的关键单元，其主要目的是通过对轮廓中的特征块进行识别定位之后，分析每一帧的特征块并在相邻两帧图像中对特征块进行相似性匹配，完成检测过程。该单元是以多GPU(Graphics Processing Units)为硬件平台的架构图像处理软件系统，该硬件平台具有平行计算处理架构，实现对熔融流体流速的实时检测。该单元是由特征块精细定位模块401、特征块匹配模块402及流速检测模块403三个模块组成。
流速检测模块403是本发明实施例的最后结果输出部分，其功能是利用熔融流体流速检测单元的前两个模块处理完一个帧图像组之后整合处理，根据实际现场情况计算出熔融流体的流速。该模块由相机标定子模块和流速输出子模块两个功能子模块构成。
结合附图对本发明具体实施方案情况进一步说明，本发明应用于国内某2650m 3高炉上，在高炉上的三个出铁口的其中一处依照图8安装高速相机及其他装置。在相机选型上，最主要是帧率和分辨率的参数选择，合适的帧率能捕获铁水流的运动信息，足够的分辨率在图像上提供更多的细节信息。据现场工人估计，高炉出铁口铁水流速为5～6m/s。为满足流速检测需求，高速相机分辨率为1280*720，帧率为240。具体完成熔融流体流速的整个检测过程的实施方案步骤如下：
步骤1，根据高速相机的安装参数及现场数据，对相机进行标定，确定图像坐标系与世 界坐标系之间的关系；
步骤2，在高炉出铁的一个周期内，高速相机视频捕捉单元实时捕获铁水流出流视频。视频采集单元20实时采集一分钟时间的铁水流出流视频，压缩存储后送至视频处理单元，随后视频采集单元20重新采集当前的视频，重复该过程；
步骤3，视频处理单元将铁水流压缩后的视频分解成帧图像，并提取帧图像中感兴趣的铁水流区域，得到待处理的帧图像组；
步骤4，步骤3得到的帧图像组经灰度变换及图像增强处理后，完成了图像预处理过程(U51)，然后通过Canny轮廓检测算法及骨架提取处理完成的铁水流轮廓提取过程，得到单像素铁水流轮廓图像，以便后续的处理；
步骤5，对得到的二值铁水流轮廓图像进行类水平像素剔除处理，最后将目标特征块分裂成两条相邻曲线，针对于这两相邻的曲线，存在最小距离、夹角差和上端点高度差这三种特征，分析显著性特征块独有的形状特征后可完成特征块识别及粗定位过程，然后依据特征块精确定位模块将铁水流的特征块粗定位的结果可选地筛选，剔除识别错误的特征块，实现特征块的精细定位；
步骤6，利用特征块的大小、角度及位置三种特征，在相邻两帧图像上对特征块进行瞬态匹配过程，匹配相似度ρ＞0.6的特征块，由式(14)得到特征块在两帧时间内移动的水平像素距离，然后，根据步骤1相机标定获取图像坐标系与世界坐标系之间的关系，将像素距离转换成真实世界距离；
步骤7，以上步骤将一个帧图像组处理完之后，剔除所有距离数据中的异常值，取平均值，根据反应炉出口处的仰角，从而得到当前帧图像组所在时间段内铁水流出流的平均流速值。
本实施例的高温熔融流体流速检测系统的具体工作过程和工作原理可参照本实施例中的高温熔融流体流速检测方法的工作过程和工作原理。
以上仅为本发明的优选实施例而已，并不用于限制本发明，对于本领域的技术人员来说，本发明可以有各种更改和变化。凡在本发明的精神和原则之内，所作的任何修改、等同替换、改进等，均应包含在本发明的保护范围之内。

Claims (10)

  1. 一种高温熔融流体流速检测方法，其特征在于，所述方法包括：
    采集高温高速熔融流体的视频流；
    将所述视频流分解成以时间为序的帧图像序列，并提取所述帧图像序列中的感兴趣熔融流体区域；
    提取所述感兴趣熔融流体区域的熔融流体轮廓；
    提取所述熔融流体轮廓的特征块，并基于所述特征块获取熔融流体的流速，所述特征块具体为高温熔融流体高速出流过程中出现的波纹或阴影。
  2. 根据权利要求1所述的高温熔融流体流速检测方法，其特征在于，提取所述感兴趣熔融流体区域的熔融流体轮廓包括：
    对所述感兴趣熔融流体区域进行预处理，获得预处理图像；
    采用一阶偏导有限差分计算所述预处理图像的梯度幅值和方向，并对所述梯度幅值进行非极大值抑制；
    采用双阈值算法检测和连接所述感兴趣熔融流体区域的熔融流体轮廓，并利用骨架提取算法细化所述熔融流体轮廓。
  3. 根据权利要求1或2所述的高温熔融流体流速检测方法，其特征在于，提取所述熔融流体轮廓的特征块包括：
    剔除所述熔融流体轮廓中类水平分布的像素点，保留与水平成预设倾角的特征块像素点，从所述熔融流体轮廓中分离出特征块轮廓；
    计算所述特征块轮廓中两条相邻曲线之间的最小距离、上端点高度差以及夹角差，并基于所述最小距离、上端点高度差以及夹角差对所述熔融流体轮廓的特征块进行粗定位；
    基于所述熔融流体轮廓对所述特征块的轮廓进行补全，并基于补全轮廓后的所述特征块的质心，对所述熔融流体轮廓的特征块进行精确定位。
  4. 根据权利要求3所述的高温熔融流体流速检测方法，其特征在于，基于所述特征块获取熔融流体的流速包括：
    获取所述特征块的特征，所述特征包括大小特征、角度特征和位置特征；
    在相邻两帧之间利用所述特征块的特征进行相似性匹配，得到特征块在相邻两帧时间内移动的像素距离；
    根据现场的安装参数对采集高温高速熔融流体的视频流的高速相机进行标定，建立图像坐标系与世界坐标系之间的关系；
    求解所述特征块在世界坐标系中的移动距离，且计算所述移动距离的计算公式为：
    Figure PCTCN2019096143-appb-100001
    其中S W为所述特征块在世界坐标系中的移动距离，R为反应炉出口圆孔的直径，R c为反应炉出口在图像上的直径像素，S C为熔融流体上的特征块在相邻两帧时间内移动的像素距离；
    基于所述移动距离以及速度公式获得高温高速熔融流体的流速。
  5. 根据权利要求4所述的高温熔融流体流速检测方法，其特征在于，获取所述特征块的特征包括：
    计算所述特征块边界点的i+j阶矩，计算公式具体为：
    Figure PCTCN2019096143-appb-100002
    其中μ ij为所述特征块边界点的i+j阶矩，M为所述特征块边界像素点的个数，m＝0,2,…,M，(x m,y m)为所述特征块边界上第m个像素点，(x t,y t)为质心坐标；
    根据所述i+j阶矩计算所述特征块的等效椭圆的长半轴r 1和短半轴r 2，计算公式具体为：
    Figure PCTCN2019096143-appb-100003
    其中，r 1和r 2分别为所述特征块的等效椭圆的长半轴和短半轴，μ 2,0为i＝2，j＝0时μ ij的取值，μ 0,2为i＝0，j＝2时μ ij的取值，μ 1,1为i＝1，j＝1时μ ij的取值；
    基于所述i+j阶矩计算所述特征块的等效椭圆与水平方向的夹角，计算公式具体为：
    Figure PCTCN2019096143-appb-100004
    其中β为所述特征块的等效椭圆与水平方向的夹角；
    利用质心公式计算所述特征块的质心，具体为：
    Figure PCTCN2019096143-appb-100005
    其中(x t,y t)为所述特征块的质心，N为连通域像素点的个数，(x n,y n)为连通域中第n个像素点的坐标；
    根据所述长半轴r 1、短半轴r 2、夹角β和质心(x t,y t)，获得所述特征块的特征。
  6. 根据权利要求5所述的高温熔融流体流速检测方法，其特征在于，在相邻两帧之间利用所述特征块的特征进行相似性匹配，得到特征块在相邻两帧时间内移动的像素距离包括：
    step1：以前一帧图像T(x,y)的特征块质心的横坐标x t为起点，在后一帧图像T′(x,y)沿横坐标向后搜索特征块质心x′ t；
    step2：在图像T′(x,y)中，将满足x′ t-x t＜l条件的质心标记，l为根据观察设定的阈值；
    step3：判断标记的质心纵坐标是否满足|y′ t-y t|≤h′，其中h′为设定的阈值，y′ t和y t分别为后一帧图像特征块质心纵坐标和前一帧图像特征块质心纵坐标，并在判定为否时去除特征块的标记，并跳转图像T(x,y)的下一个质心坐标，继续步骤step1；
    step4：计算图像T′(x,y)被标记特征块的长半轴r 1′、短半轴r 2′和夹角β′，转化成量纲相同的三个参数(r 1′cosβ′,r 1′sinβ′,r′ 2)，特征块参数记为(x′ 1,x′ 2,x′ 3)，同理，图像T(x,y)的特征块参数记为(x 1,x 2,x 3)；
    step5：为了在线实时评判两个特征块的相似度，计算相似度系数，具体为：
    Figure PCTCN2019096143-appb-100006
    其中ρ为相似度系数，ρ的取值范围为[-1,1]，x k为图像T(x,y)的第k个特征块参数，x′ k为图像T′(x,y)的第k个特征块参数，图像T′(x,y)中匹配相似度ρ最大且满足ρ＞ξ的特征块则认定为匹配成功，从而得到特征块移动的水平像素距离S C，跳转图像T(x,y)的下一个质心坐标，继续步骤step1，其中S C的计算公式为：
    S C＝x′ t-x t。
  7. 一种高温熔融流体流速检测系统，其特征在于，所述系统包括高速相机捕捉视频单元(10)、与所述高速相机捕捉视频单元(10)依次连接的视频采集单元(20)、轮廓提取单元(30)以及熔融流体流速检测单元(40)，其中，
    所述高速相机捕捉视频单元(10)，用于采集高温高速熔融流体的视频流；
    所述视频采集单元(20)，用于将所述视频流分解成以时间为序的帧图像序列，并提取所述帧图像序列中的感兴趣熔融流体区域；
    所述轮廓提取单元(30)，用于提取所述感兴趣熔融流体区域的熔融流体轮廓；
    所述熔融流体流速检测单元(40)，用于提取所述熔融流体轮廓的特征块，并基于所述特征块获取熔融流体的流速，所述特征块具体为高温熔融流体高速出流过程中出现的波纹或阴影。
  8. 根据权利要求7所述的高温熔融流体流速检测系统，其特征在于，
    所述高速相机捕捉视频单元(10)包括用于采集高温高速熔融流体的视频流的高速相机(101)以及设置于所述高速相机(101)上的镜头防尘清扫模块(102)以及高速相机风冷模块(103)，其中，
    所述镜头防尘清扫模块(102)，用于在高速相机(101)未工作时对高速相机(101)的镜头进行清扫处理；
    所述高速相机风冷模块(103)，用于将高速相机机身完全覆盖，通过外接风机加快风冷装置与高速相机机身之间空气的流动速率达到散热冷却。
  9. 根据权利要求8所述的高温熔融流体流速检测系统，其特征在于，所述视频采集单元(20)包括视频采集模块(201)和视频处理模块(202)，其中所述视频采集模块(201)包括：
    A/D模块，用于将高速相机捕捉视频单元(10)的视频源模拟信号转换成数字信号；
    存储器模块，用于存储视频信号；
    视频压缩模块，用于对所述视频信号进行压缩，并将已压缩的视频信号送至所述视频处理模块(202)；
    所述视频处理模块(202)包括：
    视频流分解模块，用于将所述视频采集模块(201)传递过来的压缩视频流分割成以时间为序的帧图像；
    帧图像感兴趣区域提取模块，用于提取所述帧图像中感兴趣熔融流体区域，并将提取后的帧图像组传送至所述轮廓提取单元(30)。
  10. 根据权利要求9所述的高温熔融流体流速检测系统，其特征在于，所述熔融流体流速检测单元(40)包括特征块精细定位模块(401)、与所述特征块精细定位模块(401)依次连接的特征块匹配模块(402)以及流速检测模块(403)；
    所述特征块精细定位模块(401)包括剔除类水平像素子模块、特征块识别及粗定位子模块、特征块精确定位子模块，其中，
    所述剔除类水平像素子模块用于剔除所述熔融流体轮廓中类水平分布的像素点，保留与水平成较大倾角的特征块像素点，从所述熔融流体轮廓线条中分离出特征块轮廓；
    所述特征块识别及粗定位子模块用于计算所述特征块轮廓中两条相邻曲线之间的最小距离、上端点高度差以及夹角差，并基于所述最小距离、上端点高度差以及夹角差对所述熔融流体轮廓的特征块进行粗定位；
    所述特征块精确定位子模块用于基于所述熔融流体轮廓对所述特征块的轮廓进行补全，并基于补全轮廓后的所述特征块的质心，对所述熔融流体轮廓的特征块进行精确定位；
    所述特征块匹配模块(402)，用于在相邻两帧之间利用所述特征块的特征进行相似性匹配，得到特征块在相邻两帧时间内移动的像素距离；
    所述流速检测模块(403)包括相机标定子模块和流速输出子模块，所述相机标定子模块用于对高速相机(101)进行标定并求解特征块在世界坐标系移动距离，所述流速输出子模块用于计算并输出熔融流体的流速。
PCT/CN2019/096143 2018-11-16 2019-07-16 一种高温熔融流体流速检测方法及系统 Ceased WO2020098299A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US16/970,344 US11605169B2 (en) 2018-11-16 2019-07-16 Method and system for detecting flow velocity of high-temperature molten fluid

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201811365784.2 2018-11-16
CN201811365784.2A CN109543587B (zh) 2018-11-16 2018-11-16 一种高温熔融流体流速检测方法及系统

Publications (1)

Publication Number Publication Date
WO2020098299A1 true WO2020098299A1 (zh) 2020-05-22

Family

ID=65847999

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/096143 Ceased WO2020098299A1 (zh) 2018-11-16 2019-07-16 一种高温熔融流体流速检测方法及系统

Country Status (3)

Country Link
US (1) US11605169B2 (zh)
CN (1) CN109543587B (zh)
WO (1) WO2020098299A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112884806A (zh) * 2021-01-12 2021-06-01 昆明理工大学 一种基于结合块匹配和密集逆向搜索的视频测流方法及装置
CN115138208A (zh) * 2022-09-06 2022-10-04 山东超华环保智能装备有限公司 数据处理器及包含该处理器的一种废气智能处理系统
CN116580227A (zh) * 2023-04-23 2023-08-11 武汉理工大学 基于图像处理算法的易混淆有色金属破碎料识别分类方法

Families Citing this family (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109543587B (zh) * 2018-11-16 2020-04-24 中南大学 一种高温熔融流体流速检测方法及系统
JP7299400B2 (ja) * 2019-06-14 2023-06-27 株式会社神戸製鋼所 温度測定装置および温度測定方法
CN110487415B (zh) * 2019-08-19 2020-10-09 中南大学 熔融金属流体体数据检测装置、方法及系统
CN111413516B (zh) * 2020-03-11 2021-02-19 中南大学 一种基于极化特征的铁水流速检测装置
CN111445444B (zh) * 2020-03-11 2020-12-01 中南大学 一种基于极化特征的铁水流速检测方法
CN112270237B (zh) * 2020-10-22 2023-08-29 中冶南方工程技术有限公司 一种高炉煤气流状态识别方法、系统及存储介质
CN113030937B (zh) * 2021-02-07 2021-12-28 中南大学 一种基于太赫兹高速回波效应的流体流速测量方法
CN113188592B (zh) * 2021-04-12 2024-06-25 攸太科技(台州)有限公司 一种尿液流速流量识别及取样方法
CN113434090B (zh) * 2021-06-30 2023-03-28 同济大学 一种用于高速视频测量的海量数据异步存储方法
CN114547914B (zh) * 2022-03-30 2022-10-25 中国石油大学(北京) 确定流体与岩石相互反应过程的方法、装置及电子设备
CN115131308B (zh) * 2022-06-27 2025-05-16 大庆恒驰电气有限公司 智能流量计算系统
CN115346172B (zh) * 2022-08-16 2023-04-21 哈尔滨市科佳通用机电股份有限公司 一种钩提杆复位弹簧丢失和折断检测方法及系统
CN115561480A (zh) * 2022-09-30 2023-01-03 福建万福信息技术有限公司 一种基于深度学习的排水口流速检测方法和系统
CN115546561B (zh) * 2022-10-25 2024-07-02 中山大学 一种基于水体覆盖特征识别的山间溪流量测验方法
CN115730512B (zh) * 2022-11-03 2023-11-17 宝钢工程技术集团有限公司 一种用于连铸结晶器铜板温度场计算的方法
CN117372966B (zh) * 2023-12-06 2024-03-01 陕西龙跃锐星科技有限公司 基于视频监控的转台状态监测方法
CN117809230A (zh) * 2024-02-29 2024-04-02 四川省水利科学研究院 一种基于图像识别的水流流速识别方法及相关产品
CN120431100B (zh) * 2025-07-08 2025-10-03 矿冶科技集团有限公司 基于泡沫图像的矿物浮选细碎泡流速在线检测方法及装置

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102103753A (zh) * 2009-12-22 2011-06-22 三星电子株式会社 使用实时相机运动估计检测和跟踪运动对象的方法和终端
KR20110131678A (ko) * 2010-05-31 2011-12-07 호서대학교 산학협력단 지능형 운송 시스템의 도로 및 차선 검출 시스템 및 방법
CN104198752A (zh) * 2014-08-18 2014-12-10 浙江大学 基于机器视觉的高温钢坯运动状态的多率检测方法
CN105488816A (zh) * 2015-11-27 2016-04-13 中南大学 一种基于三维视觉信息的矿物浮选泡沫流速在线检测装置及方法
CN107093187A (zh) * 2017-03-31 2017-08-25 上海拓攻机器人有限公司 一种无人机飞行速度的测量方法及装置
CN109543587A (zh) * 2018-11-16 2019-03-29 中南大学 一种高温熔融流体流速检测方法及系统

Family Cites Families (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6293156B1 (en) * 1999-01-22 2001-09-25 Panametrics, Inc. Coherent multi-path flow measurement system
US6941005B2 (en) * 2002-11-01 2005-09-06 Coulter International Corp. Monitoring and control of droplet sorting
CN101566639A (zh) * 2009-06-05 2009-10-28 天津大学 一种基于电磁层析成像的钢水流速测量方法和装置
CN102337369B (zh) * 2011-10-27 2014-02-05 北京首钢国际工程技术有限公司 高风温旋流喷射扰动熔融还原和预还原联合装置及方法
CN103035011B (zh) * 2012-12-06 2016-01-13 河海大学 一种基于目标特征的运动矢量估计方法
CN103480813B (zh) * 2013-09-30 2015-11-25 南京钢铁股份有限公司 一种连铸结晶器高温钢液流速测量装置及测量方法
CN103624234B (zh) * 2013-10-25 2016-09-28 北京航科发动机控制系统科技有限公司 基于图像处理的离心感应压铸机工作过程自动监测方法
CN104036522A (zh) * 2014-04-16 2014-09-10 嘉兴博海信息科技有限公司 一种基于视频中运动目标识别的流水速度监测方法
CN104131126B (zh) * 2014-08-07 2016-06-29 中冶南方工程技术有限公司 基于模糊模型的高炉熔渣流量检测方法
CN105478737A (zh) * 2015-12-15 2016-04-13 湖南红宇耐磨新材料股份有限公司 一种保温炉浇注平台自动浇注方法和系统
CN105701842A (zh) * 2016-01-08 2016-06-22 天津大学 基于链码轮廓特征的液膜波动速度测量方法

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102103753A (zh) * 2009-12-22 2011-06-22 三星电子株式会社 使用实时相机运动估计检测和跟踪运动对象的方法和终端
KR20110131678A (ko) * 2010-05-31 2011-12-07 호서대학교 산학협력단 지능형 운송 시스템의 도로 및 차선 검출 시스템 및 방법
CN104198752A (zh) * 2014-08-18 2014-12-10 浙江大学 基于机器视觉的高温钢坯运动状态的多率检测方法
CN105488816A (zh) * 2015-11-27 2016-04-13 中南大学 一种基于三维视觉信息的矿物浮选泡沫流速在线检测装置及方法
CN107093187A (zh) * 2017-03-31 2017-08-25 上海拓攻机器人有限公司 一种无人机飞行速度的测量方法及装置
CN109543587A (zh) * 2018-11-16 2019-03-29 中南大学 一种高温熔融流体流速检测方法及系统

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112884806A (zh) * 2021-01-12 2021-06-01 昆明理工大学 一种基于结合块匹配和密集逆向搜索的视频测流方法及装置
CN112884806B (zh) * 2021-01-12 2022-09-02 昆明理工大学 一种基于结合块匹配和密集逆向搜索的视频测流方法及装置
CN115138208A (zh) * 2022-09-06 2022-10-04 山东超华环保智能装备有限公司 数据处理器及包含该处理器的一种废气智能处理系统
CN116580227A (zh) * 2023-04-23 2023-08-11 武汉理工大学 基于图像处理算法的易混淆有色金属破碎料识别分类方法
CN116580227B (zh) * 2023-04-23 2025-10-10 武汉理工大学 基于图像处理算法的易混淆有色金属破碎料识别分类方法

Also Published As

Publication number Publication date
US11605169B2 (en) 2023-03-14
CN109543587B (zh) 2020-04-24
US20210090270A1 (en) 2021-03-25
CN109543587A (zh) 2019-03-29

Similar Documents

Publication Publication Date Title
WO2020098299A1 (zh) 一种高温熔融流体流速检测方法及系统
Song et al. Real-time tunnel crack analysis system via deep learning
CN109541254B (zh) 一种高温熔融流体流速检测装置
CN111445444B (zh) 一种基于极化特征的铁水流速检测方法
CN112785628B (zh) 一种基于全景视角检测跟踪的轨迹预测方法及系统
CN103571994B (zh) 转炉红外钢渣探测方法
CN110218824B (zh) 一种基于图像识别的高炉渣离心粒化粒径检测系统及方法
CN1897015A (zh) 基于机器视觉的车辆检测和跟踪方法及系统
CN107862675A (zh) 一种用于电铲斗齿缺失的实时视觉检测方法
CN110136172A (zh) 一种矿工下井前防护设备佩戴的检测方法
CN115830501A (zh) 一种基于视觉的高炉铁水流渣铁识别方法及系统
CN108182682A (zh) 一种基于计算机图像处理和模式识别的检测方法
Fu et al. Research on image-based detection and recognition technologies for cracks on rail surface
CN114663672A (zh) 一种输电线路塔钢构件腐蚀检测方法及系统
CN113160132A (zh) 焊缝缺陷图像的检测处理方法及系统
CN106290379A (zh) 基于面扫描相机的钢轨表面缺陷检测装置及方法
CN110751669A (zh) 一种新型cbocp在线红外转炉出钢钢流自动检测与跟踪方法及系统
CN108364306B (zh) 一种高速周期运动的视觉实时检测方法
Jia et al. Detection model of drainage pipe defect based on improved YOLOv5
CN111539395A (zh) 一种基于光流法和支持向量机的挖掘机实时检测方法
CN102542239B (zh) 一种烧结机尾红断面有效图像自动捕捉方法
CN118365902A (zh) 基于多任务学习与背景解耦的孔探损伤计数方法
CN108441889B (zh) 一种阳极导杆钢爪附着物检测方法及装置
CN106636510A (zh) 一种基于机器视觉的自动换钎方法
Tang et al. A method for automatically locating defects in CCTV inspection data of sewer pipes

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19884635

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19884635

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 21.01.2022)

122 Ep: pct application non-entry in european phase

Ref document number: 19884635

Country of ref document: EP

Kind code of ref document: A1