CN117190904A - Intelligent identification and remote temperature measurement system and method for sand core of casting sand of aircraft casing - Google Patents

Abstract

The invention designs an intelligent recognition and remote temperature measurement system and method for sand cores of casting sand of an aircraft case, wherein, firstly, a field visible light image of the sand mold is collected through a remote temperature measurement system, and an image training data set of the sand mold of the casting sand mold of the aircraft case is established; training an aircraft case casting sand mould network identification model; mapping the identified sand core image of the casting sand of the aircraft casing into a thermal imaging image through the coordinate relation between the thermal imaging and the visible light image; extracting temperature data of a casting sand core of a target aircraft casing in the thermal imaging image by an image processing method, and finally outputting a temperature data result; the system accurately identifies the sand core image of the aircraft case casting and acquires the target temperature data in real time under a complex production environment, realizes unmanned detection of the temperature data of the target sand core all the day, and improves the temperature measurement efficiency and detection accuracy of the production link.

Description

Aviation casing casting sand core intelligent identification and remote temperature measurement system and method

Technical field

The invention relates to the field of aviation casing casting sand core casting, and in particular to a system and method for intelligent identification and remote temperature measurement of aviation casing casting sand cores.

Background technique

In the machinery manufacturing and processing equipment industry, the production of castings plays a very important role, and temperature is an important factor in the casting production process. In the process of sand core casting of aviation magnesium aluminum alloy casing, for the use of resin material (material is resin sand, the resin material is generally PEP-SET, the resin sand is hammered into shape, the density is about 1.6, and the surface roughness is generally Ra6.3-Ra12.5) sand molds and sand cores. During the casting process, if the production temperature is not properly controlled, the sand mold will easily produce defects such as moisture absorption, deformation, and lack of flesh. These defects will cause the alloy liquid to During the pouring process, air entrainment, inclusions, fleshiness and loss of important features are formed, which reduces the mechanical properties of the castings produced by pouring, and in severe cases causes the scrapping of the castings. Therefore, accurate temperature control is required in the aviation magnesium-aluminum alloy casing. Sand core castings have a significant impact on the casting production process.

In the field of casting, PEP-SET material resin produced by Ashland Company of the United States is widely used in aviation aluminum-magnesium alloy sand casting. With the rapid development of the aviation industry, aluminum-magnesium alloy aircraft made of PEP-SET resin Cassette castings also face many production problems, such as temperature monitoring problems during the casting process of aluminum-magnesium alloy sand molds using PEP-SET resin materials, especially the reconstruction of the target shape and worker occlusion during the sand mold closing process in the casting process. The problem causes low recognition rate.

Existing temperature measurement methods are mainly divided into contact temperature measurement and non-contact temperature measurement. In order to avoid damaging the integrity of the casting and reduce the impact of the temperature sensor arrangement on the temperature field distribution, the non-contact temperature measurement method is used to monitor the temperature of the casting. The traditional non-contact handheld thermometer requires workers to hold the temperature measurement. The operation process is cumbersome and can only measure the local temperature of the sand mold. The measurement efficiency is low and the temperature cannot be measured 24 hours a day.

Currently, target recognition technology has a wide range of applications in real life, especially in fields such as driverless cars, video inspection, industrial inspection, and aerospace. Target recognition technology uses computer vision technology and image processing algorithms to identify castings. Accurately detects the temperature of target castings and can adapt to castings of different shapes and sizes. Therefore, combining target recognition and non-contact temperature measurement is very necessary in the casting field.

Contents of the invention

In order to solve the problems of low identification accuracy of sand mold cores and casting defects caused by temperature changes during the aviation casing casting process, the present invention provides an intelligent identification and remote temperature measurement system and method for the sand cores of aviation casing casting.

Aviation casing casting sand core intelligent identification and remote temperature measurement system, including: aviation casing casting sand core image acquisition module, temperature and humidity data acquisition module, aviation casing casting sand core identification module, ring light source, computer and display ;

Among them, the computer preprocesses the on-site sand mold visible light image collected by the image acquisition module through the aviation casing casting sand core identification module to obtain the characteristic image, and inputs the characteristic image into the sand core identification model based on deep learning for identification. The identified sand core boundary box information is mapped to the thermal imaging image through the coordinate relationship between the visible light image and the thermal imaging image. Image processing is performed on the sand core boundary box mapped in the thermal imaging image to obtain the thermal imaging image. The target sand core image is combined with the thermal imaging temperature data to extract the temperature data of the target sand core image and record the temperature data;

The aviation casing casting sand core image acquisition module collects visible light images and thermal imaging images through a thermal imaging bispectral camera, and transmits the visible light images and thermal imaging images to the computer in the form of electrical signals; the bispectral camera is installed on A certain distance away from the wall of the target sand mold working area to reduce the interference of on-site operations on camera image collection;

The temperature and humidity data acquisition module collects on-site temperature and humidity data through temperature and humidity sensors, and transmits the temperature and humidity data of the on-site environment to the computer. The temperature and humidity sensors are arranged in parallel around the sand mold and sand core in an array form, and communicate with the computer. Serial port connection to reduce the impact of transmission distance on data transmission performance;

The aviation casing casting sand core identification module is used to identify the target aviation casing casting sand core; using the YOLOv5 network model; the aviation casing casting sand core identification module constructs a data set of aviation casing casting images, and The collected image data is divided into training set, verification set, and test set;

The ring light source is used for lighting the sand core of aviation casing castings on site;

The computer is used to run the system program and recognize and process images;

The display is used to display the collected image information and temperature data.

The intelligent identification and remote temperature measurement method of aviation casing casting sand cores is implemented based on the above intelligent identification and remote temperature measurement system of aviation casing casting sand cores, including the following steps:

Step 1: Before identifying the target aviation casing casting sand core, turn on the ring light source; use the thermal imaging dual-spectrum camera of the image acquisition module to collect visible light images and thermal imaging images of the casting site; and transmit them to the computer; through the temperature The humidity sensor collects temperature and humidity data at the casting site and transmits it to the computer;

Step 2: The computer performs image preprocessing on the visible light image, including using grayscale, image enhancement and Gaussian filtering methods to enhance image features;

Step 3: Establish a deep learning recognition network model for aviation casing casting sand cores. Input the characteristic image into the deep learning recognition network model for aviation casing casting sand cores. Recognize the image. If the target sand mold is recognized, then in the visible light The identification frame of the target sand mold is displayed in the image. Save the parameters of the identification frame and perform step 4. If not, perform step 6;

The establishment of the deep learning recognition network model of the aviation casing casting sand core includes the following steps:

Step S1: The computer controls the thermal imaging bispectral camera, and the camera collects visible light images of the casting process of the target aviation casing sand core at a set collection frequency;

Step S2: Repeat step S1 multiple times to collect a large number of image samples;

Step S3: The computer performs image preprocessing on the visible light image of the aviation casing casting sand core, including using grayscale, image enhancement and Gaussian filtering methods to enhance image features;

Step S4: Augment the image data, and increase the number of images by rotating, scaling, adding salt and pepper noise, adding Gaussian noise, and image shading to the preprocessed visible light images, thereby improving the performance of the network model. Generalization;

Step S5: The computer uses the labelimg software to annotate the target image boundary box of the target aviation casing casting sand core in the augmented visible light image, and obtains the annotated image and image annotation data;

Step S6: The computer randomly uses 70% of the annotated images and corresponding image annotation data as a training set, 20% as a verification set, and 10% as a test set;

Step S7: Input the image training set into the YOLOv5 network model, train the sand core recognition model, and output the training error and training weight data to achieve the purpose of training the network;

Step S8: Repeat the above steps to create a training set, a verification set and a test set for image samples of the existing target aviation casing casting sand core, and perform model training; the loss function is calculated for each iteration, and Update parameter values to minimize the value of the loss function until the model converges;

Step S9: Save the weight parameters of the trained aviation casing casting sand core model, and use the test set images for testing to achieve the training results of the model;

Step 4: The computer outputs the identification frame parameters of the target sand mold in the visible light image, including the pixel coordinates of the center point of the bounding box (u _rgb , v _rgb ), the length h _rgb and the width w _rgb of the bounding box; through the comparison between the visible light image and the thermal imaging image The coordinate mapping relationship formula is used to obtain the bounding box data of the target sand mold in the thermal imaging image. The bounding box of the target sand mold uses the point (u _ir , v _ir ) as the center point pixel coordinate (u _ir , v _ir ) of the bounding box, and h _ir and w _ir are used as the length and width of the bounding box; the coordinate mapping relationship formula and the bounding box length and width relationship formula are as follows:

Among them, K _rgb and K _ir are the internal parameter matrices of the visible light camera and the thermal imaging camera built in the thermal imaging bispectrum camera respectively; E _rgb and E _ir are the external parameter matrices of the visible light camera and the thermal imaging camera respectively; Z _rgb and Z _ir is the scale factor; p _rgb is the coordinate under the visible light image, p _rgb = [u _rgb ,v _rgb ,1] ^T ; p _ir is the coordinate under the thermal imaging image, p _ir = [u _ir ,v _ir ,1] ^T ; a, b are the scaling ratios in the x-direction and y-direction between the visible light image and the thermal imaging image;

Perform image processing on the target bounding box of the thermal imaging image to obtain the target sand core image. Combined with the temperature data of the thermal imaging image itself, the temperature data of the target sand core image is accurately extracted, and the average temperature value is calculated:

Where T _avg is the average temperature of the target; N is the number of effective pixels in the identification area; _Ti is the thermal imaging temperature of the effective pixels; the image processing is specifically: image grayscale, histogram equalization for thermal imaging Perform image processing on the bounding box area to obtain the target sand core image;

Step 5: Determine whether the average temperature T _avg of the aviation casing casting sand core is greater than the normal temperature value preset in the computer system. If it is greater, it is abnormal temperature data. If not, it is normal temperature data;

Step 6: The computer controls the display to display the collected image and temperature data of the aviation casing casting sand core. At the same time, a temperature data table is generated and stored in the local database.

The invention has beneficial technical effects:

The intelligent identification and temperature measurement method of the aviation casing casting sand core of the present invention adopts non-contact temperature measurement. Compared with the traditional manual temperature collection method, it increases the stability of target temperature collection, reduces labor costs, and uses thermal imaging. The camera collects the temperature of the target casting sand mold and accurately identifies the sand core of the target aviation casing casting by using machine vision technology to further reduce the error of manual temperature measurement and improve the identification efficiency.

Visual inspection technology is used to achieve accurate and uninterrupted real-time monitoring. Compared with traditional template matching recognition methods, the application of target recognition technology can help machines accurately identify and locate the position and shape of parts to be cast, improving industrial automation. degree and production efficiency, ensuring casting quality and reducing manual errors.

The network model of YOLOv5 based on deep learning is trained on a large number of sand core images of aluminum-magnesium alloy aviation casing castings to build a network model of sand cores for aluminum-magnesium alloy aviation casing castings, which greatly improves the target recognition accuracy. It also has high robustness to complex light irradiation at the casting site, can achieve rapid recognition of targets and accurate recognition of blocked images, and improves the automation of casting production.

Description of the drawings

Figure 1 is a flow chart of the sand core identification and temperature measurement system for aviation casing casting sand cores according to the embodiment of the present invention.

Figure 2 is a schematic diagram of the sand core identification and temperature measurement system for aviation casing casting sand cores according to the embodiment of the present invention.

Detailed ways

The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

The invention provides a system and method for intelligent identification and remote temperature measurement of aviation casing casting sand cores.

Aviation casing casting sand core intelligent identification and remote temperature measurement system, as shown in Figure 2, includes: aviation casing casting sand core image acquisition module, temperature and humidity data acquisition module, aviation casing casting sand core identification module , ring lights, computers and monitors;

Among them, the computer preprocesses the on-site sand mold visible light image collected by the image acquisition module through the aviation casing casting sand core identification module to obtain the characteristic image, and inputs the characteristic image into the sand core identification model based on deep learning for identification. The identified sand core boundary box information is mapped to the thermal imaging image through the coordinate relationship between the visible light image and the thermal imaging image. Image processing is performed on the sand core boundary box mapped in the thermal imaging image to obtain the thermal imaging image. The target sand core image is combined with the thermal imaging temperature data to extract the temperature data of the target sand core image and record the temperature data;

The aviation casing casting sand core image acquisition module collects visible light images and thermal imaging images through a thermal imaging dual spectrum camera, and transmits the visible light images and thermal imaging images to the computer in the form of electrical signals through Ethernet; the dual spectrum The camera is installed on a wall about 8.3 meters away from the target sand molding operation area and 4.2 meters above the ground to reduce the interference of on-site operations on the images collected by the camera;

The temperature and humidity data acquisition module collects on-site temperature and humidity data through temperature and humidity sensors, and transmits the temperature and humidity data of the on-site environment to the computer through the 485 bus. The temperature and humidity sensors are arranged in parallel around the sand mold and sand core in the form of an array. 12VDC power supply, connected to the computer serial port through a 120-meter-long four-core shielded cable to reduce the impact of transmission distance on data transmission performance;

The aviation casing casting sand core identification module is used to identify the target aviation casing casting sand core; using the YOLOv5 network model; the aviation casing casting sand core identification module constructs a data set of aviation casing casting images, and The collected image data is divided into training set, verification set, and test set;

The ring light source is used for lighting the sand core of aviation casing castings on site;

The computer is used to run the system program and recognize and process images;

The display is used to display the collected image information and temperature data.

The intelligent identification and remote temperature measurement method of aviation casing casting sand cores is implemented based on the above intelligent identification and remote temperature measurement system of aviation casing casting sand cores. As shown in Figure 1, it includes the following steps:

Step 1: Before identifying the target aviation casing casting sand core, turn on the ring light source; use the thermal imaging dual-spectrum camera of the image acquisition module to collect visible light images and thermal imaging images of the casting site; and connect them to the 485 bus through the Ethernet Transmit to the computer; collect the temperature and humidity data of the casting site through the temperature and humidity sensor, and transmit it to the computer through the 485 bus;

Step 2: The computer performs image preprocessing on the visible light image, including using grayscale, image enhancement and Gaussian filtering methods to enhance image features;

Step 3: Establish a deep learning recognition network model for aviation casing casting sand cores. Input the characteristic image into the deep learning recognition network model for aviation casing casting sand cores. Recognize the image. If the target sand mold is recognized, then in the visible light The identification frame of the target sand mold is displayed in the image and step 4 is performed. If not, step 6 is performed;

The establishment of the deep learning recognition network model of the aviation casing casting sand core includes the following steps:

Step S1: The computer controls the thermal imaging bispectral camera through the built-in program, and the camera collects visible light images of the casting process of the sand core of the target aviation casing at a frequency of 1 minute;

Step S2: Repeat step S1 multiple times to collect a large number of image samples;

Step S3: The computer performs image preprocessing on the visible light image of the aviation casing casting sand core, including using grayscale, image enhancement and Gaussian filtering methods to enhance image features;

Step S4: Augment the image data, and increase the number of images by rotating, scaling, adding salt and pepper noise, adding Gaussian noise, and image shading to the preprocessed visible light images, thereby improving the performance of the network model. Generalization;

Step S5: The computer uses the labelimg software to annotate the target image boundary box of the target aviation casing casting sand core in the augmented visible light image, and obtains the annotated image and image annotation data;

Step S6: The computer randomly uses 70% of the annotated images and corresponding image annotation data as a training set, 20% as a verification set, and 10% as a test set;

Step S7: Input the image training set into the YOLOv5 network model, train the sand core recognition model, and output the training error and training weight data to achieve the purpose of training the network;

Step S8: Repeat the above steps to create a training set, a verification set and a test set for image samples of the existing target aviation casing casting sand core, and perform model training; the loss function is calculated for each iteration, and Update parameter values to minimize the value of the loss function until the model converges; at the same time, in order to avoid overfitting, the number of iterations is set to 500.

Step S9: Save the weight parameters of the trained aviation casing casting sand core model, and use the test set images for testing to achieve the training results of the model;

Step 4: The computer outputs the identification frame parameters of the target sand mold in the visible light image, including the pixel coordinates of the center point of the bounding box (u _rgb , v _rgb ), the length h _rgb and the width w _rgb of the bounding box; through the comparison between the visible light image and the thermal imaging image The coordinate mapping relationship formula is used to obtain the bounding box data of the target sand mold in the thermal imaging image. The bounding box of the target sand mold uses the point (u _ir , v _ir ) as the center point pixel coordinate (u _ir , v _ir ) of the bounding box, and h _ir and w _ir are used as the length and width of the bounding box; the coordinate mapping relationship formula and the bounding box length and width relationship formula are as follows:

Among them, K _rgb and K _ir are the internal parameter matrices of the visible light camera and the thermal imaging camera respectively; E _rgb and E _ir are the external parameter matrices of the visible light camera and the thermal imaging camera respectively; Z _rgb and Z _ir are the scale factors; p _rgb is the visible light The coordinates under the image, p _rgb = [u _rgb ,v _rgb ,1] ^T ; p _ir is the coordinate under the thermal imaging image, p _ir = [u _ir ,v _ir ,1] ^T ; a, b are the visible light image and The scaling ratio in the x-direction and y-direction between thermal imaging images;

Perform image processing on the target bounding box of the thermal imaging image to obtain the target sand core image. Combined with the temperature data of the thermal imaging image itself, the temperature data of the target sand core image is accurately extracted, and the average temperature value is calculated:

Where T _avg is the average temperature of the target; N is the number of effective pixels in the identification area; _Ti is the thermal imaging temperature of the effective pixels; the image processing is specifically: image grayscale, histogram equalization for thermal imaging Perform image processing on the bounding box area to obtain the target sand core image;

Step 5: Determine whether the average temperature T _avg of the aviation casing casting sand core is greater than the normal temperature value preset in the computer system. If it is greater, it is abnormal temperature data. If not, it is normal temperature data;

Step 6: The computer controls the display to display the collected image and temperature data of the aviation casing casting sand core. At the same time, a temperature data table is generated and stored in the local database.

This invention establishes an aviation casing sand mold recognition network model based on deep learning to identify the target aviation casing casting sand mold, uses an infrared camera to extract the temperature of the target aviation casing casting sand mold, and at the same time, by measuring the on-site environmental humidity, reduces the impact of humidity on the target temperature. influence, improving the accuracy of temperature measurement. It is used in the mass production mode of magnesium-aluminum alloy casing sand molds at sand molds and sand core molding core manufacturing sites to monitor the integrity of the appearance structure of the sand core and the average temperature of the sand core.

Claims (5)

1. Intelligent identification and remote temperature measurement system of aircraft receiver foundry sand psammitolite, its characterized in that includes: the device comprises an image acquisition module of sand and sand cores of the aircraft casing castings, a temperature and humidity data acquisition module, an identification module of sand and sand cores of the aircraft casing castings, an annular light source, a computer and a display;

the computer preprocesses the field sand mold visible light image acquired by the image acquisition module through the aircraft casing casting sand core identification module to obtain a characteristic image, inputs the characteristic image into the sand mold core identification model based on deep learning for identification, maps the identified sand mold core boundary frame information onto the thermal imaging image through the coordinate relationship between the visible light image and the thermal imaging image, performs image processing on the sand mold core boundary frame mapped in the thermal imaging image to obtain a target sand core image in the thermal imaging image, extracts temperature data of the target sand core image in combination with thermal imaging temperature data, and records the temperature data.

2. The intelligent recognition and remote temperature measurement system for the sand core of the casting of the aircraft casing according to claim 1, wherein the image acquisition module for the sand core of the casting of the aircraft casing acquires visible light images and thermal imaging images through a thermal imaging dual-spectrum camera and transmits the visible light images and the thermal imaging images to a computer in the form of electric signals; the double-spectrum camera is arranged at a position which is a distance from a target sand mold operation area to a wall so as to reduce interference of field operation on images acquired by the camera;

the temperature and humidity data acquisition module acquires field temperature and humidity data through a temperature and humidity sensor and transmits the temperature and humidity data of the field environment to a computer, and the temperature and humidity sensor is arranged around a sand mold and a sand core in parallel in an array form and is connected with a serial port of the computer so as to reduce the influence of transmission distance on data transmission performance;

the aircraft case casting sand-sand core identification module is used for identifying the sand-sand core of the target aircraft case casting; adopting a YOLOv5 network model; the aircraft casing casting sand core identification module constructs a data set of an aircraft casing casting image, and the acquired image data is divided into a training set, a verification set and a test set;

the annular light source is used for polishing the sand core of the cast of the field aviation case;

the computer is used for executing the system program and identifying and processing the image;

the display is used for displaying the acquired image information and the temperature data.

3. The intelligent identification and remote temperature measurement method for the sand core of the casting sand of the aircraft casing is realized based on the intelligent identification and remote temperature measurement system for the sand core of the casting sand of the aircraft casing according to the above claim 1, and is characterized by comprising the following steps:

step 1: before the sand core of the casting sand of the target aviation case is identified, an annular light source is started; the method comprises the steps of utilizing a thermal imaging dual-spectrum camera of an image acquisition module to acquire visible light images and thermal imaging images on a casting site; and transmitting it to the computer; acquiring temperature and humidity data of a casting site through a temperature and humidity sensor, and transmitting the temperature and humidity data to a computer;

step 2: the computer carries out image preprocessing on the visible light image, which comprises the steps of enhancing the image characteristics by using methods of graying, image enhancement and Gaussian filtering;

step 3: establishing an aircraft casing casting sand core deep learning identification network model, inputting a characteristic image into the aircraft casing casting sand core deep learning identification network model, identifying the image, if a target sand mold is identified, displaying an identification frame of the target sand mold in a visible light image, storing parameters of the identification frame, and executing the step 4, if not, executing the step 6;

step 4: the computer outputs the identification frame parameters of the target sand mold in the visible light image, including the center point pixel coordinates (u _rgb ,v _rgb ) Length h of bounding box _rgb Sum width w _rgb The method comprises the steps of carrying out a first treatment on the surface of the Obtaining the boundary frame data of the target sand mold in the thermal imaging image through a coordinate mapping relation formula of the visible light image and the thermal imaging image, wherein the boundary frame of the target sand mold is formed by points (u _ir ,v _ir ) Center point pixel coordinates (u) _ir ,v _ir ) In h _ir And w is equal to _ir As bounding box length and width;

performing image processing on a target boundary frame of the thermal imaging image to obtain a target sand core image, accurately extracting temperature data of the target sand core image by combining temperature data of the thermal imaging image, and calculating an average temperature value;

step 5: judging average temperature T of sand core of casting sand of aircraft casing _avg Whether the temperature value is larger than a normal temperature value preset in the computer system or not, if so, the temperature value is abnormal temperature data, and if not, the temperature value is normal temperature data;

step 6: the computer controls the display to display the collected image and temperature data of the casting sand core of the aircraft casing, and meanwhile, a temperature data table is generated and stored in a local database.

4. The intelligent recognition and remote temperature measurement method for the sand core of the casting sand of the aircraft casing according to claim 2, wherein the establishing of the deep learning recognition network model for the sand core of the casting sand of the aircraft casing in the step 3 comprises the following steps:

step S1, controlling a thermal imaging double-spectrum camera by a computer, and acquiring visible light images of a casting process of a sand core of a target aviation receiver by the camera at a set acquisition frequency;

s2, repeating the step S1 for a plurality of times to collect a large number of image samples;

s3, performing image preprocessing on the visible light image of the sand core cast by the aircraft casing by the computer, wherein the image preprocessing comprises the steps of enhancing the image characteristics by using a method of graying, image enhancement and Gaussian filtering;

step S4, the image data is amplified, and the number of images is increased by respectively performing rotation, scaling, addition of salt and pepper noise, addition of Gaussian noise and image shading processing on the preprocessed visible light images, so that the generalization capability of a network model is improved;

s5, labeling a target image boundary box of the casting sand core of the target aviation casing in the amplified visible light image by using a computer through labelimg software, and obtaining a labeled image and image labeling data;

step S6, the computer randomly takes 70% of the marked images and the corresponding image marked data as a training set, 20% as a verification set and 10% as a test set;

s7, inputting a training set of the image into a network model of YOLOv5, training a sand core identification model, and outputting training errors and training weight data to achieve the aim of training a network;

s8, repeating the steps, namely repeating the steps for manufacturing a training set, a verification set and a test set for image samples of the sand cores cast by the existing target aircraft case for multiple times, and performing model training; calculating a loss function in each iteration, and updating parameter values to minimize the value of the loss function until the model converges;

and S9, saving weight parameters of the trained aircraft case casting sand mould core model, and testing by using the test set image to achieve a training result of the model.

5. The intelligent identification and remote temperature measurement method for the sand core cast by the aircraft casing according to claim 2, wherein the coordinate mapping relation formula and the bounding box length-width relation formula in the step 4 are as follows:

wherein K is _rgb ，K _ir The internal reference matrixes of a visible light camera and a thermal imaging camera which are arranged in the thermal imaging dual-spectrum camera are respectively; e (E) _rgb ，E _ir The external parameter matrixes are respectively a visible light camera and a thermal imaging camera; z is Z _rgb And Z is _ir Is a scale factor; p is p _rgb Is the coordinates under the visible light image, p _rgb ＝[u _rgb ,v _rgb ,1] ^T ；p _ir For coordinates under the thermographic image, p _ir ＝[u _ir ,v _ir ,1] ^T The method comprises the steps of carrying out a first treatment on the surface of the a, b is the scaling in the x-direction and y-direction between the visible light image and the thermographic image;

performing image processing on a target boundary frame of the thermal imaging image to obtain a target sand core image, accurately extracting temperature data of the target sand core image by combining temperature data of the thermal imaging image, and calculating an average temperature value:

wherein T is _avg An average temperature that is a target; n is the number of effective pixel points of the identification area; t (T) _i A thermal imaging temperature for the effective pixel point; the image processing specifically comprises the following steps: and carrying out image processing on the thermal imaging boundary box area by image graying and histogram equalization to obtain the target sand core image.