Temperature monitoring method for master alloy turning based on infrared image

Abstract

The invention relates to the technical field of image processing, and in particular to a method for monitoring temperature of a turning process of an intermediate alloy based on an infrared image. The method comprises the following steps: collecting an infrared image during turning process of the intermediate alloy, selecting any pixel point of an alloy region of any infrared image, calculating the high temperature degree of the pixel point according to the deviation of the #imgabs0# channel value of the pixel point and the sum of the differences between the pixel point and its neighboring pixel points in each channel, calculating the first significant characteristic value of the pixel point according to the sum of the differences between the #imgabs1# channel mean value in a local range of the pixel point and the #imgabs2# channel mean value in a local range corresponding to the pixel point in an alloy region of other infrared images, and the high temperature degree, executing a significant detection algorithm based on the first significant characteristic values ​​of all pixel points in the alloy region, and identifying the high temperature region of the intermediate alloy. The method reduces the influence of noise through a user-defined first significant characteristic value, obtains an accurate significant detection result, and ensures the accuracy of temperature monitoring.

Description

Technical Field

[0001] The invention relates to the technical field of image processing, and in particular to a method for monitoring temperature of a master alloy turning process based on infrared images. Background Art

[0002] In modern manufacturing, temperature plays a vital role in processing quality and tool durability. In particular, temperature control is particularly important in the turning process of intermediate alloys. If the temperature is too high during turning, the intermediate alloy will oxidize, thereby significantly reducing its performance and product quality. Therefore, real-time monitoring and control of the temperature of alloy parts during processing is the key to improving product quality and production efficiency. For example, the Chinese patent application document with the existing publication number CN114646396A discloses an online monitoring method for internal field temperature during milling, including: Step 1: Data acquisition: Use an infrared thermal imager to monitor the tool and tool in real time. Step 1: Take a picture of the part to obtain the tool external field temperature thermal map; Step 2: Data processing: Extract the temperature of the tool external field surface point; Step 3: Heat transfer analysis: Establish a heat conduction forward problem model for milling; Step 4: Virtual simulation: Simulate the heat transfer process of the milling cutter in milling, apply different sizes of heat fluxes in the tool internal field, solve the tool internal temperature field, and obtain the temperature rise sensitivity coefficient of each point inside the tool at different temperatures; Step 5: Heat transfer inverse solution: Construct a heat transfer inverse problem model, regard the heat transfer inverse solution problem as an optimization problem, and solve the heat flux size of the tool internal field at each moment; Step 6: Solve the internal field temperature, and realize temperature monitoring through the above steps.

[0003] In order to further improve the accuracy of temperature abnormality area detection and accurately locate the temperature abnormality area of ​​the alloy part, the saliency detection CA (Context-Aware) algorithm, also known as the Goferman algorithm, can be used to identify the temperature abnormality area by collecting infrared images and analyzing the color characteristics and spatial position relationship of the pixel points. For example, the Chinese patent application document with the publication number CN118096738A discloses a new energy photovoltaic module fault detection method and system, firstly obtaining the initial image of the photovoltaic module, and determining the color distance of the two component block areas according to the gray value distribution difference between the two component block areas, the abnormal probability value of the component block area, and the background gray value; according to the difference in the abnormal probability value of the two component block areas and the initial spatial distance, and the difference in the color distance between the component block area and all the component block areas within the preset neighborhood range, the adjusted spatial distance of the two component block areas is obtained; according to the component saliency map, the fault detection result of the photovoltaic module is obtained. By constructing the single-scale saliency value as an indicator, this scheme can effectively reflect the difference between the weak hot spot area and the normal area, improve the saliency detection effect of the component saliency map, and obtain the abnormal area of ​​the photovoltaic module.

[0004] However, in practical applications, complex processing environments may lead to noise interference during the infrared image acquisition process of alloy parts. This is mainly caused by electromagnetic interference or radiation from electrical equipment in the surrounding environment. These noises appear as random pixel value fluctuations or dot-like bright spots in the infrared images of alloy parts, which will directly affect the accuracy of the subsequent significance detection results, making the final identified temperature abnormality area inaccurate. Summary of the invention

[0005] In order to solve the problem that environmental noise affects the accuracy of the significance detection result of the infrared image of the alloy part, resulting in the inability to accurately identify the temperature abnormal area of ​​the alloy part, the present invention proposes a method for monitoring the temperature of the master alloy turning process based on infrared images, comprising:

[0006] Collect infrared images of the master alloy turning process for a period of time, segment the alloy area of ​​each infrared image, select any pixel point of the alloy area of ​​an infrared image, and The high temperature degree of the pixel is calculated based on the deviation of the channel value and the sum of the differences between the pixel and its neighboring pixels in each channel, wherein the high temperature degree is positively correlated with the deviation and negatively correlated with the sum of the differences;

[0007] According to the local range of the pixel point The channel mean and the local range corresponding to the pixel point in the alloy area of ​​other infrared images The sum of the differences of the channel means and the high temperature degree are used to calculate a first significant eigenvalue of the pixel point, wherein the first significant eigenvalue is negatively correlated with the sum of the differences and positively correlated with the high temperature degree;

[0008] A significance detection algorithm is executed based on the first significance characteristic value of all pixels in the alloy area of ​​the infrared image to obtain the second significance characteristic value of all pixels in the alloy area and the significance area in the alloy area, and the temperature estimation value of the alloy area is evaluated by the area size of the significance area and the size of the second significance characteristic value, and whether there is temperature anomaly in the alloy area is monitored according to the size of the temperature estimation value.

[0009] The above technical solution can obtain time series data of temperature changes of alloy parts during processing by analyzing the infrared image and multiple other infrared images that are adjacent to it in time sequence, which is helpful to capture the trend and anomaly of temperature change. Segmenting the image into specific alloy areas can reduce unnecessary background interference and ensure that subsequent analysis focuses only on the area of ​​interest. When calculating the high temperature degree of the pixel point, the temperature information of the pixel point itself and the relationship with the surrounding pixels are considered, which can more accurately reflect the temperature state of a single pixel point. Pixel points with large deviations usually represent higher temperatures, and pixels with small differences from neighboring pixels indicate that the temperature of the point is relatively uniform, and it is less likely to be a high temperature abnormal point caused by noise. When calculating the first significant eigenvalue of the pixel point, the pixel point is usually more likely to be a high temperature abnormal point caused by noise. By comparing the temperature changes of the same pixel at different times, areas with sudden or continuously high temperatures can be identified. This cross-frame comparison helps detect real temperature anomalies rather than short-term fluctuations caused by random noise. Executing a significance detection algorithm based on the first significant eigenvalue of a custom pixel can accurately identify high-temperature areas that are more significant than the surrounding environment, reducing the impact of noise factors and helping to accurately locate potential temperature anomaly areas of alloy parts. By evaluating the temperature estimate of the alloy area through the area size of the significant area and the size of the second significant eigenvalue, the degree of abnormal temperature can be quantified, which helps to take relevant cooling measures in a timely manner and avoid product quality problems caused by temperature abnormalities.

[0010] Further, the method for evaluating the temperature estimation value of the alloy region by the area size of the significant region and the size of the second significant characteristic value is:

[0011] The ratio of the total number of pixels of the significant region in the alloy region of the infrared image to the total number of pixels of the alloy region is used as the area ratio of the significant region in the alloy region;

[0012] The value obtained by normalizing the product of the accumulated value of the second significant characteristic values ​​of all the pixels in the alloy region and the area ratio is used as the temperature estimation value of the alloy region.

[0013] The above technical scheme can quantitatively reflect the relative size of the high-temperature area in the entire alloy area by calculating the area ratio of the significant area, which helps to evaluate its influence on the overall temperature distribution. The second significant eigenvalues ​​of all pixels are accumulated and combined with the area ratio, which can comprehensively reflect the temperature intensity and range of the high-temperature area, and help to more accurately estimate the actual temperature level of the alloy area. The normalization process makes the temperature estimation value more comparable, which is convenient for longitudinal and transverse comparative analysis of the temperature anomaly of the significant area under different times and conditions.

[0014] Furthermore, the method for monitoring whether the alloy region has temperature anomaly according to the magnitude of the temperature estimation value is:

[0015] A temperature threshold is preset. If the estimated temperature value of the alloy area is greater than or equal to the temperature threshold, there is a temperature anomaly in the alloy area, and a cooling operation is performed; if the estimated temperature value of the alloy area is less than the temperature threshold, there is no temperature anomaly in the alloy area.

[0016] Furthermore, the method for obtaining the second significant characteristic value of all pixel points in the alloy region is:

[0017] The number of scales for each pixel in the alloy region when executing the saliency detection algorithm is set to , is a preset value, and each scale corresponds to a range of different sizes;

[0018] Calculate, at each scale, the average of the differences between the first significant eigenvalues ​​of the pixel point and other pixels at the scale as the first significant eigenvalue of the pixel point at the scale;

[0019] An average value of the first significant feature value of the pixel at all scales is calculated as the second significant feature value of the pixel.

[0020] The above technical solution can integrate information at different scales by calculating the average value of significant eigenvalues ​​at all scales, improve the stability and reliability of significant eigenvalues, and help to more accurately evaluate the degree of temperature anomaly in significant areas.

[0021] Furthermore, the method for obtaining the significant region in the alloy region is:

[0022] A threshold value of the second significant characteristic value is preset. If the second significant characteristic value of a pixel point in the alloy area is greater than the threshold value of the second significant characteristic value, the pixel point belongs to the significant area. Otherwise, the pixel point does not belong to the significant area. All pixel points belonging to the significant area constitute the significant area in the alloy area.

[0023] Furthermore, the first significant eigenvalue satisfies the following relationship:

[0024]

[0025] In the formula, For the The first infrared image of the alloy area The first significant eigenvalue of pixels, For the The first infrared image of the alloy area The high temperature of each pixel, For the The number of other infrared images besides the infrared image, is the serial number of other infrared images, is the natural exponential function, For the The first infrared image of the alloy area All pixels in the local area of ​​​​the pixel The mean of the channel values, For the The first infrared image of the alloy area All pixels in the local area of ​​the pixel are in The corresponding pixel points in the infrared image The mean of the channel values.

[0026] The above technical solution calculates the local range of the same pixel point in different infrared images. The mean difference of channel values ​​can effectively capture the trend of temperature changes over time, which helps to identify abnormal temperature areas. The natural exponential function can also be used to amplify the significant characteristic values ​​of pixels with large temperature differences, thereby highlighting abnormally high temperature areas. The high temperature degree and the pixel points in the local range of multiple consecutive infrared images can be combined to obtain the abnormal temperature of the pixels. The mean difference of channel values ​​can reduce the impact of noise and improve the stability and reliability of significant eigenvalues.

[0027] Furthermore, the high temperature degree of the pixel point satisfies the following relationship:

[0028]

[0029] In the formula, For the The first infrared image of the alloy area The high temperature of each pixel, For the The first infrared image of the alloy area Pixels The deviation of the channel value is Pixels Channel values ​​and all pixels in the alloy area The difference between the mean values ​​of the channel values, is the number of channels, is the channel number, is the number of neighboring pixels of the pixel, is the sequence number of the neighborhood pixel, For the The first infrared image of the alloy area The pixel Channel values, For the The first infrared image of the alloy area The pixel The first neighboring pixel channel values, and , For the The first infrared image of the alloy area Pixels Channel value, For the All pixels of the alloy area in the infrared image The mean of the channel values, is the natural exponential function, is the normalization function.

[0030] The above technical solution comprehensively considers the temperature information of the pixel itself and the difference between the pixel and the surrounding pixels, which helps to more accurately identify and evaluate high temperature abnormal areas. Through nonlinear conversion, it can reduce the impact of noise and improve the accuracy of high temperature degree calculation.

[0031] Furthermore, the local range of the pixel point is centered on the pixel point. The size range of pixels, is the default value.

[0032] Furthermore, the The other infrared images besides the first infrared image are The continuous infrared images before the acquisition time Infrared image, is the default value.

[0033] Furthermore, the method used to segment the alloy region of each infrared image is semantic segmentation.

[0034] The present invention has the following effects:

[0035] The present invention can accurately capture the trend and anomaly of temperature change by continuously collecting infrared images and calculating the high temperature degree of pixel points, effectively solving the problem of background noise interference, and executing the significance detection algorithm according to the customized high temperature significance feature value of each pixel point to obtain accurate significance detection results. It can accurately locate the high temperature area, and accurately monitor the temperature of the cutting process of the alloy part according to the temperature estimation value of the high temperature area, which is beneficial for the operator to make timely adjustments during the processing and ensure the quality and stability of the alloy processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0037] Figure 1 It is a schematic flow chart of the method of the present invention. DETAILED DESCRIPTION

[0038] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0039] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0040] Reference Figure 1 The method for monitoring the temperature of the master alloy turning process based on infrared images includes steps S1 to S3, which are as follows:

[0041] S1: Collect infrared images of the master alloy turning process over a period of time, segment the alloy area of ​​each infrared image, and evaluate the high temperature level of any pixel in the alloy area of ​​any infrared image.

[0042] In this step, an infrared thermal imager is used to collect infrared images of the intermediate alloy during turning processing. The collection is continuous for a period of time. In this embodiment, the continuous collection time is set to 30 minutes, and the interval between each collection is 5 seconds. After that, semantic segmentation is performed on each collected infrared image to obtain the alloy area and background area in each infrared image. The background area is an irrelevant area. The subsequent temperature monitoring is mainly performed on the alloy area in the infrared image.

[0043] In order to prevent the noise interference in the infrared image from affecting the accuracy of the subsequent saliency detection results, this step first Channel values ​​and all pixels The difference between the channel values ​​and the difference between the channel values ​​of the pixel and its neighboring pixels are used to evaluate the high temperature degree of each pixel in each infrared image. The higher the high temperature degree, the more likely the pixel is to belong to the high temperature area in the infrared image. The lower the high temperature degree, the less likely the pixel is to belong to the high temperature area and the more likely it is to be a noise point.

[0044] In infrared images, the Pixels with high channel values ​​are usually located in higher temperature areas. Therefore, if a pixel If the channel value is high, then the pixel may belong to the high temperature area. The channel values ​​of these three channels are very different from the pixels in their neighborhood. The smaller the similarity between this pixel and its neighborhood pixels, the more likely it is noise, and the less likely it is that the pixel belongs to a higher temperature area. This is because noise usually appears as isolated outliers.

[0045] First If the infrared image is the infrared image used at the current analysis moment, the high temperature degree of each pixel in the infrared image is calculated, and the high temperature degree of the pixel satisfies the following relationship:

[0046]

[0047] In the formula, For the The first infrared image of the alloy area The high temperature of each pixel, For the The first infrared image of the alloy area Pixels The deviation of the channel value is Pixels Channel values ​​and all pixels in the alloy area The difference between the mean values ​​of the channel values, is the number of channels, since These three channels, therefore The value of is 3, is the channel number, is the number of neighboring pixels of the pixel, is the sequence number of the neighborhood pixel, For the The first infrared image of the alloy area The pixel Channel values, For the The first infrared image of the alloy area The pixel The first neighboring pixel channel values, and , For the The first infrared image of the alloy area Pixels Channel value, For the All pixels of the alloy area in the infrared image The mean of the channel values.

[0048] In this formula, Indicates The first infrared image of the alloy area Pixels The channel value and all pixels in the entire alloy area The difference in the mean of the channel values, The larger the value is, the higher the temperature of the pixel is than the average temperature of the surrounding pixels, and the more likely the pixel is to belong to the high temperature area. and is a positive correlation, otherwise, The smaller the value is, the lower the temperature of the pixel is than the average temperature of the surrounding pixels, and the less likely the pixel is to belong to the high temperature area.

[0049] In this formula, It reflects the relationship between the pixel and its neighboring pixels. The sum of the differences in the three channels. The larger the value, the greater the difference between the pixel and its neighboring pixels in each channel. The pixel is more likely to be a noise pixel and less likely to be a pixel in a high temperature area. The negative exponential function reflects the sum of the values. This means that the larger the sum of the differences, the smaller the high temperature of the pixel, and the two are negatively correlated. Conversely, the smaller the sum of the differences, the more similar the pixel is to its neighboring pixels, the less likely it is to be a noise pixel, and the more likely it is to be a pixel in the high temperature area, so its high temperature is greater.

[0050] Therefore, comprehensive and Get high temperature, The larger the value, the The larger, and The smaller it is, the more likely the pixel is to belong to the high temperature area in the infrared image and the less likely it is to be a noise point. The smaller the pixel The smaller, and The larger the value is, the less likely the pixel is to belong to the high temperature area in the infrared image, and the more likely it is to be a noise point. This step can make a preliminary assessment of the noise level of the pixel. The high temperature level of the pixel can preliminarily reflect the possibility of whether the pixel belongs to the high temperature area of ​​the infrared image or a noise point.

[0051] S2: Calculate the first significant eigenvalue of the pixel based on the high temperature degree of the pixel, the R channel mean value in the local range of the pixel, and the R channel mean value in the corresponding local range of the alloy region of other infrared images.

[0052] If we only use one infrared image to calculate the high temperature of a pixel and make a preliminary judgment on whether the pixel belongs to a high temperature area pixel or a noise pixel, the following defects will occur:

[0053] Difficulty in distinguishing noise points from real high temperature areas: Noise points in a single infrared image may form local bright spots or bright spots, which are difficult to distinguish from real high temperature areas. If the noise points happen to form a clustered area, then it looks like a real high temperature area.

[0054] Lack of time series information: A single infrared image can only provide temperature distribution information at a certain moment and cannot reflect the temperature change trend over time. This means that it is impossible to determine whether a local bright spot or bright spot is caused by temperature anomaly or noise.

[0055] It is impossible to rule out the factor of chance: a single infrared image may just capture an accidental event, such as momentary electromagnetic interference, which will cause an erroneous bright spot to appear in the image. This randomness may lead to misjudgment of the bright spot as a pixel point of a real high-temperature area.

[0056] Therefore, in order to overcome the above shortcomings, this step is for the above-mentioned first infrared images, and obtain the temporally adjacent Further analysis of the infrared image shows It is a preset value, which is set to 10 in this embodiment.

[0057] By analyzing multiple consecutive infrared images, the trend of temperature changes can be observed. During the turning process, the temperature change on the alloy surface usually changes slowly rather than suddenly. If a certain area always appears as a bright spot or bright spot in multiple infrared images, it is more likely to be a real high temperature area. On the contrary, if a bright spot or bright spot only appears in one or a few infrared images, it is more likely to be caused by noise. By analyzing the temperature changes in a certain area in multiple infrared images, statistics such as the mean and standard deviation of the temperature in the area can be calculated. These statistics help determine whether the temperature change is statistically significant, so as to more accurately determine whether the area is a real high temperature area. By analyzing multiple consecutive infrared images, the trend of temperature changes over time can be identified. If the temperature of a certain area gradually increases or decreases, it is more likely to be a real high temperature area.

[0058] In summary, the use of multiple infrared image analysis can more accurately distinguish real high temperature areas from noise because it provides time series information, which helps to confirm the trend of temperature change and reduce the impact of accidental factors. In contrast, single infrared image analysis may have difficulty distinguishing real high temperature from noise because it lacks time series information and statistical support.

[0059] Therefore, in order to improve the detection accuracy, this step proposes a method that combines the information of multiple infrared images in a continuous time and the high temperature degree of each pixel to calculate the first significant feature value of each pixel. The specific process is as follows:

[0060] First, for the above The feature analysis is performed on each pixel of the infrared image to obtain the high temperature degree of each pixel.

[0061] Then, set the local range: take each pixel as the center and select a The range of pixel sizes is used as the local range of the pixel. is a preset value. In this embodiment, .

[0062] In step S1, Take the infrared image as an example and obtain the The 10 infrared images before the first infrared image are used as reference images to form a reference image set. The first infrared image of the alloy area The local range of pixels.

[0063] In each reference image in the reference image set, find the image that matches the one in the first The area corresponding to the local range of the pixel points is calculated based on the The channel mean is consistent with the above All pixels in the local area of ​​​​the pixel The difference between the channel means, and the above The high temperature degree of each pixel is calculated as mentioned above. The first significant feature value of the pixel point is used to accurately evaluate the above The possibility that a pixel point belongs to a high temperature area, the larger the first significant eigenvalue, the higher the probability that the The pixel point is more likely to belong to the high temperature area, rather than being caused by electronic noise. On the contrary, if the first significant eigenvalue is smaller, the above The pixels are more likely to be caused by electronic noise.

[0064] The first significant eigenvalue satisfies the following relationship:

[0065]

[0066] In the formula, For the The first infrared image of the alloy area The first significant eigenvalue of pixels, For the The first infrared image of the alloy area The high temperature of each pixel, For the The number of other infrared images besides the infrared image, is the serial number of other infrared images, is the natural exponential function, For the The first infrared image of the alloy area All pixels in the local area of ​​​​the pixel The mean of the channel values, For the The first infrared image of the alloy area All pixels in the local area of ​​the pixel are in The corresponding pixel points in the infrared image The mean of the channel values.

[0067] This formula calculates the first significant eigenvalue of a pixel taking into account and These two factors, It is positively correlated with the first significant eigenvalue. It is negatively correlated with the first significant eigenvalue. The higher the value, the more likely it is that the pixel belongs to the high temperature area, and the lower the possibility that the pixel is a noise point. Represents the local range of all pixels in the pixel The cumulative difference of the mean of the channel values ​​in multiple consecutive infrared images. The smaller the value, the more likely it is that the pixel point has similar characteristics to all the pixels in its local range in the multiple consecutive infrared images, and the more likely it is that the pixel point is a real pixel point in the high-temperature area, and the less likely it is that the bright spot is generated by electronic noise.

[0068] Therefore, when The bigger, The smaller it is, the larger the first significant eigenvalue of the pixel point is, indicating that the The first infrared image of the alloy area The pixel is not only The probability that the pixel belongs to the high temperature area is high, and all the pixels in the local range of the pixel in the continuous infrared images are The mean change of the channel value is small, indicating that the pixel is not caused by accidental noise, which further confirms that the pixel belongs to the high temperature area in the infrared image. This design can better distinguish whether the pixel belongs to the real high temperature area or the bright spot area caused by electronic noise, and then accurately distinguish the real high temperature area and the bright spot area caused by electronic noise in the infrared image according to the area to which each pixel belongs.

[0069] In summary, this step aims to analyze multiple infrared images in a continuous time, combining the high temperature degree of each pixel and the local range of all pixels in the pixel. The change in the mean of the channel values ​​can be used to more accurately identify high-temperature areas and eliminate the influence of electronic noise. This method utilizes time series data and statistical analysis to improve the accuracy and reliability of detection.

[0070] S3: Execute a significance detection algorithm based on the first significance eigenvalue to obtain the second significance eigenvalue of all pixels in the alloy area and the significance area in the alloy area, evaluate the temperature estimation value of the significance area through the second significance eigenvalue, and monitor whether there is temperature anomaly in the alloy area according to the size of the temperature estimation value.

[0071] Salient-Object-Detection is an important research direction in the field of computer vision. It is mainly used to automatically identify and extract the most salient objects or regions from images or videos. Salient-Object-Detection is widely used in many practical applications, such as target tracking, image segmentation, video summary generation, etc. In the manufacturing industry, especially in the turning process of intermediate alloys, saliency detection can be used to monitor the temperature changes in the processing process in real time. By detecting the high-temperature area in the infrared image, the surface temperature of the alloy cylinder can be estimated to ensure the processing quality.

[0072] The traditional saliency detection CA (Context-Aware) algorithm is an efficient saliency detection method. This detection method mainly combines the two factors of color distance and spatial distance to calculate the saliency value of the pixel. Color distance refers to the difference between the colors of two pixels in an image. It is usually expressed by the distance in the color space, such as the distance in the RGB color space or the HSV color space. The color distance is used to distinguish the regions of different colors in the image and is an important basis for saliency detection. Spatial distance refers to the distance between two pixels in the image in space. The Euclidean distance between the pixel coordinates is usually used. The spatial distance is used to maintain the continuity between adjacent pixels and helps to maintain the integrity of the saliency region. The saliency value refers to the saliency degree of each pixel in the image. The saliency value is usually composed of the weighted sum of the color distance and the spatial distance. The weight can be adjusted according to the specific situation. The saliency region refers to the region with a higher saliency value in the image. The saliency region is determined by setting a threshold. In the turning process of master alloys, the saliency region corresponds to the high temperature region and can be used to estimate the surface temperature of the alloy cylinder.

[0073] Advantages of the saliency detection CA algorithm: High contrast: It can highlight salient objects or areas in the image and improve the contrast; Precise positioning: It can accurately locate salient objects or areas in the image; Strong adaptability: It is suitable for various types of images, including natural images and infrared images; High computational efficiency: The algorithm has low computational complexity and is suitable for real-time detection.

[0074] Problems with the saliency detection CA algorithm: The traditional saliency detection CA algorithm calculates the saliency feature value of each pixel based on color distance and spatial distance, and then screens out the saliency area based on the comparison between the saliency feature value of the pixel and the threshold size. However, the presence of noise will affect the accuracy of saliency detection, that is, the detected saliency area is not accurate enough, and the high-temperature area on the surface of the alloy part cannot be accurately located, affecting the accuracy of temperature monitoring.

[0075] Noise can cause the colors of some pixels in the image to deviate from their true color values. When the algorithm relies on color distance to calculate saliency, noise can lead to misjudgment of saliency. For example, in infrared images, noise can cause some non-high temperature areas to look like high temperature areas, or conversely, make the real high temperature areas less salient. Noise can affect saliency calculations based on neighborhood comparison. If pixels in a high temperature area are mistakenly assigned to different locations due to noise, the saliency of the area may be underestimated. A non-high temperature pixel that is originally adjacent to a high temperature area may be mistakenly considered to be part of the high temperature area due to noise. These situations will affect the accuracy of saliency detection.

[0076] Therefore, in order to reduce the impact of noise on the detection results of the saliency detection CA algorithm, this step uses a customized first saliency feature value to replace the saliency feature value calculated based on color distance and spatial distance in the traditional saliency detection CA algorithm, and then performs saliency detection to obtain the accurate saliency area of ​​the alloy area of ​​the infrared image, that is, the high temperature area of ​​the alloy area of ​​the infrared image, so as to accurately realize temperature monitoring.

[0077] In this step, the saliency value of each pixel is calculated by the saliency detection CA algorithm to detect the high temperature area (saliency area) in the infrared image. The specific steps are as follows:

[0078] Determine the number of scales: Determine the Infrared image of The number of scales of pixels in saliency detection is 5. In saliency detection, different scales can capture features of different sizes, which is crucial for identifying salient areas in images. At smaller scales, the algorithm can better identify small details or small objects in the image, which helps detect local changes in the image. For example, in infrared images, small scales can help identify smaller high-temperature areas. At larger scales, the algorithm can better identify larger areas in the image. In infrared images, larger scales help identify larger high-temperature areas.

[0079] Scale expansion: When expanding between adjacent scales, the area after each expansion is 3 pixels longer than the area before expansion. For example, pixels as the center, assuming the initial scale is , we get an initial region, which becomes , we get the first expansion area, and after the second expansion, it becomes , get the second expansion area, and then go to the fifth scale to get the fifth expansion area. Calculate the above The significant feature value of each pixel at each expansion scale can be captured by gradually increasing the size of the expansion area. The significant feature values ​​of pixels from small to large.

[0080] Calculation of significant eigenvalues: For the above For each pixel point, at each scale, the average of the differences between the first significant feature value of the pixel point and the first significant feature values ​​of other pixels in the expanded area of ​​the scale is calculated, and the average is used as the significant feature value of the pixel point at the scale.

[0081] Calculation of the second significant eigenvalue: The significant feature values ​​of a pixel at all scales are averaged, and the average is used as the second significant feature value of the pixel. By calculating the significant feature values ​​of the pixel at different scales and taking the average value, the influence of noise can be reduced and the stability of saliency detection can be improved.

[0082] Significant region screening: The threshold of the second significant feature value is preset to 0.8. The alloy area of ​​the infrared image If the second significant eigenvalue of a pixel is greater than 0.8, the pixel belongs to the significant area; otherwise, the pixel does not belong to the significant area, thereby screening out all the pixels belonging to the significant area of ​​the alloy area of ​​the infrared image, and these pixels constitute the significant area in the alloy area.

[0083] Through the above steps, the second significant eigenvalue of each pixel point in the alloy area of ​​each infrared image and the significant area in the alloy area, that is, the high temperature area, can be obtained. In this step, it is necessary to estimate the temperature value of the alloy area of ​​the infrared image based on the above two indicators. The larger the area of ​​the significant area of ​​the alloy area of ​​each infrared image, the more high temperature areas there are in the alloy area, and the larger the temperature estimation value of the alloy area. The larger the second significant eigenvalue of the pixel point in the alloy area of ​​each infrared image, the higher the temperature of the pixel point in the alloy area, and the larger the temperature estimation value of the alloy area.

[0084] According to the above logic, the normalized temperature estimation value of the alloy area in each infrared image of the master alloy turning process is calculated. Based on the temperature estimation value, it is judged whether there is temperature abnormality in the alloy area, which can improve the accuracy of temperature monitoring and help to ensure the quality control of the processing process. The specific estimation method is:

[0085] Based on the above The ratio of the total number of pixels in all significant regions in the alloy region of the infrared image to the total number of pixels in the alloy region is expressed as the area ratio of all significant regions in the alloy region:

[0086]

[0087] In the formula, For the The total area of ​​all significant regions in the alloy region of the infrared image is The proportion of the alloy area in the infrared image, For the The total area of ​​all significant regions in the alloy region of the infrared image (expressed by the total number of pixels in all significant regions), For the The area of ​​the alloy region in an infrared image (expressed by the total number of pixels in the alloy region).

[0088] The value obtained by normalizing the product of the accumulated value of the second significant characteristic value of all pixels in the alloy region and the area ratio is used as the temperature estimation value of the alloy region, which is expressed by the formula:

[0089]

[0090] In the formula, Indicates The estimated temperature of the alloy region in the infrared image, For the The accumulated value of the second significant eigenvalues ​​of all pixels in the alloy area of ​​the infrared image, For the The total area of ​​all significant regions in the alloy region of the infrared image is The proportion of the alloy area in the infrared image, is the normalization function.

[0091] The normalized temperature estimate above is Reflects the The temperature estimation of the alloy area of ​​the infrared image is a comprehensive consideration of the area ratio of the significant area and the cumulative value of the second significant eigenvalues ​​of all pixels. The area ratio of the significant area indicates the total area ratio of the significant area (i.e., the high-temperature area) in the alloy area. The larger the area ratio, the higher the proportion of the high-temperature area in the alloy area, indicating that the overall temperature of the alloy area is higher. The larger the cumulative value of the second significant eigenvalues ​​of all pixels in the alloy area, the more likely the pixels in the alloy area are to belong to the high-temperature area. Through normalization processing, the calculation results can be mapped to a fixed Within the range, the normalized values ​​are more convenient for comparing and analyzing the temperature estimation of alloy areas in different infrared images.

[0092] Traditional temperature numerical values ​​are suitable for directly measuring the temperature of a single pixel, which may be difficult to compare directly due to the different pixel value ranges of different infrared images. The normalized temperature estimate is more suitable for evaluating the temperature condition of the entire alloy area, especially when it is necessary to comprehensively evaluate the area and temperature intensity of the significant area. The normalized temperature estimate can better reflect the overall temperature condition of the alloy area by comprehensively considering the area ratio of the significant area and the cumulative value of the second significant eigenvalues ​​of all pixels. Compared with traditional temperature numerical values, it has better comparability and applicability, and is more suitable for temperature anomaly detection and quality control.

[0093] Among them, according to the above The method of monitoring whether there is temperature anomaly in the alloy area is as follows:

[0094] The preset temperature threshold is 0.9. If the temperature estimation value of the alloy area in the infrared image is greater than or equal to the temperature threshold, the temperature of the alloy area is too high, resulting in temperature anomaly, and a cooling operation needs to be performed. Otherwise, there is no temperature anomaly in the alloy area.

[0095] For each moment in the turning process, the corresponding infrared image can be obtained. The same analysis and processing process of each infrared image is performed to obtain the temperature estimation value of the alloy area of ​​the infrared image corresponding to that moment. According to the size of the temperature estimation value, it is judged whether there is temperature anomaly in the alloy area at that moment, so as to realize the temperature monitoring of the intermediate alloy turning process.

[0096] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0097] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. A method for monitoring temperature of master alloy turning based on infrared images, characterized in that: include: Collect infrared images of the master alloy turning process for a period of time, segment the alloy area of ​​each infrared image, select any pixel point of the alloy area of ​​an infrared image, and The high temperature degree of the pixel is calculated based on the deviation of the channel value and the sum of the differences between the pixel and its neighboring pixels in each channel, wherein the high temperature degree is positively correlated with the deviation and negatively correlated with the sum of the differences; According to the local range of the pixel point The channel mean and the local range corresponding to the pixel point in the alloy area of ​​other infrared images The sum of the differences in the channel means and the high temperature degree are used to calculate the first significant eigenvalue of the pixel point, satisfying the following relationship: ； In the formula, For the The first infrared image of the alloy area The first significant eigenvalue of pixels, For the The first infrared image of the alloy area The high temperature of each pixel, For the The number of other infrared images besides the infrared image, is the serial number of other infrared images, is the natural exponential function, For the The first infrared image of the alloy area All pixels in the local area of ​​​​the pixel The mean of the channel values, For the The first infrared image of the alloy area All pixels in the local area of ​​the pixel are in The corresponding pixel points in the infrared image The mean of the channel values; A significance detection algorithm is executed based on the first significance characteristic value of all pixels in the alloy area of ​​the infrared image to obtain the second significance characteristic value of all pixels in the alloy area and the significance area in the alloy area, and the temperature estimation value of the alloy area is evaluated by the area size of the significance area and the size of the second significance characteristic value, and whether there is temperature anomaly in the alloy area is monitored according to the size of the temperature estimation value.

2. The method for monitoring temperature of master alloy turning based on infrared image according to claim 1, characterized in that: The method for evaluating the temperature estimation value of the alloy region by the area size of the significant region and the size of the second significant characteristic value is: The ratio of the total number of pixels of the significant region in the alloy region of the infrared image to the total number of pixels of the alloy region is used as the area ratio of the significant region in the alloy region; The value obtained by normalizing the product of the accumulated value of the second significant characteristic values ​​of all the pixels in the alloy region and the area ratio is used as the temperature estimation value of the alloy region.

3. The method for monitoring temperature of master alloy turning based on infrared image according to claim 2, characterized in that: The method for monitoring whether the alloy region has temperature anomaly according to the size of the temperature estimation value is: A temperature threshold is preset. If the estimated temperature value of the alloy area is greater than or equal to the temperature threshold, there is a temperature anomaly in the alloy area, and a cooling operation is performed; if the estimated temperature value of the alloy area is less than the temperature threshold, there is no temperature anomaly in the alloy area.

4. The method for monitoring temperature of master alloy turning based on infrared image according to claim 2, characterized in that: The method for obtaining the second significant characteristic value of all pixel points in the alloy area is: The number of scales for each pixel in the alloy region when executing the saliency detection algorithm is set to , is a preset value, and each scale corresponds to a range of different sizes; Calculate, at each scale, the average of the differences between the first significant eigenvalues ​​of the pixel point and other pixels at the scale as the first significant eigenvalue of the pixel point at the scale; An average value of the first significant feature value of the pixel at all scales is calculated as the second significant feature value of the pixel.

5. The method for monitoring temperature of master alloy turning based on infrared image according to claim 2, characterized in that: The method for obtaining the significant region in the alloy region is as follows: A threshold value of the second significant characteristic value is preset. If the second significant characteristic value of a pixel point in the alloy area is greater than the threshold value of the second significant characteristic value, the pixel point belongs to the significant area. Otherwise, the pixel point does not belong to the significant area. All pixel points belonging to the significant area constitute the significant area in the alloy area.

6. The method for monitoring temperature of master alloy turning based on infrared image according to claim 4, characterized in that: The high temperature degree of the pixel point satisfies the following relationship: ； In the formula, For the The first infrared image of the alloy area The high temperature of each pixel, For the The first infrared image of the alloy area Pixels The deviation of the channel value is Pixels Channel values ​​and all pixels in the alloy area The difference between the mean values ​​of the channel values, is the number of channels, is the channel number, is the number of neighboring pixels of the pixel, is the sequence number of the neighborhood pixel, For the The first infrared image of the alloy area The pixel Channel values, For the The first infrared image of the alloy area The pixel The first neighboring pixel channel values, and , For the The first infrared image of the alloy area Pixels Channel value, For the All pixels of the alloy area in the infrared image The mean of the channel values, is the natural exponential function, is the normalization function.

7. The method for monitoring temperature of master alloy turning based on infrared image according to claim 4, characterized in that: The local range of the pixel point is centered on the pixel point. The size range of pixels, is the default value.

8. The method for monitoring temperature of master alloy turning based on infrared image according to claim 4, characterized in that: The said The other infrared images besides the first infrared image are The continuous infrared images before the acquisition time Infrared image, is the default value.

9. The method for monitoring temperature of master alloy turning based on infrared image according to claim 1, characterized in that: The method used to segment the alloy region of each infrared image is semantic segmentation.

Images