A method, system and device for improving the recognition rate of infrared thermal imaging for personnel

Abstract

The present invention belongs to the technical field of image processing, and particularly relates to a method, system and device for improving the recognition rate of infrared thermal imaging of personnel, aiming to solve the problem of low recognition rate of infrared thermal imaging of personnel affected by humanoid heat-emitting objects in the prior art. The present invention includes: collecting video images of mine working conditions, and extracting high-level features and low-level features of each image; establishing Gaussian pyramids and Laplacian pyramids of high-level features, and fusing them with corresponding low-level features; calculating the feature space contrast based on the pixel value gradient of the fused features, and taking the connected domain with a contrast greater than the set threshold as the candidate target domain; generating a motion trajectory based on the position of the centroid of each candidate target domain in the video image and the image frame sequence; taking the candidate target domain corresponding to the irregular motion trajectory as the personnel recognition result. The present invention can eliminate the misdetection of personnel caused by the heat generation of humanoid devices, and can effectively improve the recognition rate of infrared thermal imaging of personnel.

Description

Background Art

[0002] With the continuous improvement of the mechanization level of coal mining and excavation equipment, there are more and more equipment in the heading face. During the production operation process, the equipment shuttles back and forth and alternates cyclically, bringing great potential safety hazards to the personnel in the heading face. Especially during production, the dust concentration is high, the visibility is dim, the noise is loud, and there are large visual blind spots during the movement of the equipment, making it extremely easy to occur accidents of squeezing and colliding between the equipment and the personnel. And due to the large number of operators and the poor self-discipline of individual employees, safety accidents often occur due to inadequate supervision.

[0003] With the progress and development of technology, using modern monitoring equipment to replace traditional manual monitoring has become a characteristic of the modern coal mining industry. Among them, target detection is an important part of coal mine safety detection. Most of the existing technologies use the following methods for target detection: First, a background model is obtained by statistical methods, and the background model is updated in real time to adapt to the changes in light and the scene itself. Morphological methods and the detection of the connected domain area are used for post-processing to eliminate the influence of noise and background disturbance, and shadows are detected in the HSV chromaticity space to obtain accurate moving targets.

[0004] However, during coal mining and excavation production, the dust concentration is high and the visibility is dim. Through the thermal imager, it can be recognized without being interfered by dust and lighting conditions, and it can clearly image under the conditions of high dust and low illuminance, which is convenient for image analysis.

[0005] Since the thermal imager collects the heat of the human body and converts it into an image, however, some equipment will also generate heat during normal operation, such as cable heating, water pump operation, motor heating, etc. Using an infrared thermal imager for target detection may misjudge some humanoid heat-generating objects as humans, resulting in a decrease in the personnel recognition accuracy rate and an increase in the false alarm rate, making it difficult to ensure the safety of personnel and equipment in coal mining and excavation operations. Summary of the Invention

[0006] The present invention provides a method, system and device for improving the recognition rate of infrared thermal imaging personnel to solve at least one of the above technical problems existing in the prior art.

[0007] The present invention is implemented by adopting the following technical solutions: A method for improving the recognition rate of infrared thermal imaging personnel includes the following steps:

[0008] S10: Collect video images of the mine working conditions through a thermal imager, and extract the high-level features and low-level features of each image in the video images through a feature extraction network;

[0009] S20: Establish a Gaussian pyramid and a Laplacian pyramid of the high-level features, and fuse them with the corresponding low-level features in the feature channel dimension to obtain fused features;

[0010] S30: Calculate the spatial contrast of the fusion feature based on the pixel value gradient of the fusion feature, and use the connected region with a contrast greater than the set threshold as the candidate target region;

[0011] S40: Obtain the positions of the centroids of each candidate target region in the video image respectively, and generate the motion trajectories of the centroids of each candidate target region based on the sequence of image frames;

[0012] S50: Use the candidate target region corresponding to the irregular motion trajectory as the personnel recognition result.

[0013] Preferably, before extracting the high-level features and low-level features of each image in the video image in step S10, there is also a step of infrared image enhancement, including:

[0014] Perform high-frequency extraction on each image in the video image and photometric fusion of the corresponding visible light image of each image respectively through a set of multiple Gaussian filters to obtain a multi-scale enhanced image;

[0015] Perform weighted fusion on each scale enhanced image of the multi-scale enhanced image to obtain the enhanced image corresponding to each image in the video image.

[0016] Preferably, the multi-scale enhanced image is:

[0017]

[0018] where h k and y k are the high-frequency component and the corresponding enhanced image obtained by high-frequency extraction and photometric fusion at the k-th scale, k represents the number of scales, W k is the Gaussian operator corresponding to the k-th scale, x k is the feature of the visible light image corresponding to the infrared image at the k-th scale, and a is a hyperparameter preset to be lower than the average value of the image pixels, is the set photometric fusion factor.

[0019] Preferably, the enhanced image corresponding to each image in the video image is:

[0020]

[0021] where Y is the enhanced image corresponding to an image in the video image, and y t is the enhanced image obtained by high-frequency extraction and photometric fusion at the t-th scale.

[0022] Preferably, in step S30, based on the pixel value gradient of the fusion feature, the method for calculating the spatial contrast of the fusion feature is:

[0023] Extract the target region and the background region of the fusion feature, and calculate the information entropy of the target region and the background region respectively;

[0024] Take the absolute value of the difference in information entropy within a set width range at the adjacent part of the target region and the background region as the spatial contrast of the fusion feature.

[0025] Preferably, the method for calculating the information entropy of the target region and the background region respectively is as follows:

[0026] Divide the gray values of the target region and the background region into a set number of discrete gray levels;

[0027] Calculate the proportion of the number of pixel points in each gray level in the pixel points of the entire fusion feature respectively, and obtain the pixel probability of each gray level;

[0028] Based on the pixel probability of each gray level, calculate the information entropy of the target region and the background region respectively.

[0029] Preferably, the calculation method of the information entropy is as follows:

[0030]

[0031] Where H is the information entropy, M is the number of gray levels, and p(u) is the pixel probability of the u-th gray level.

[0032] The present invention also provides a system for improving the recognition rate of infrared thermal imaging of personnel, including:

[0033] A data acquisition and feature extraction module, configured to collect video images of mine working conditions through a thermal imager, and extract high-level features and low-level features of each image in the video images through a feature extraction network;

[0034] A feature fusion module, configured to establish a Gaussian pyramid and a Laplacian pyramid of the high-level features, and fuse them with the corresponding low-level features in the feature channel dimension to obtain a fusion feature;

[0035] A target candidate region screening module, configured to calculate the spatial contrast of the fusion feature based on the pixel value gradient of the fusion feature, and take the connected region with a contrast greater than a set threshold as a candidate target region;

[0036] A motion trajectory extraction module, configured to respectively obtain the positions of the centroids of each candidate target region in the video image, and generate the motion trajectories of the centroids of each candidate target region based on the sequence of image frames;

[0037] The target recognition module is configured to use the candidate target domain corresponding to the irregular motion trajectory as the personnel recognition result.

[0038] The present invention also provides an electronic device, including: at least one processor; and a memory communicatively connected to at least one of the processors; wherein, the memory stores instructions executable by the processor, and the instructions are used to be executed by the processor to implement a method for improving the infrared thermal imaging personnel recognition rate.

[0039] The present invention also provides a computer-readable storage medium, characterized in that: the computer-readable storage medium stores computer instructions, and the computer instructions are used to be executed by the computer to implement a method for improving the infrared thermal imaging personnel recognition rate.

[0040] Compared with the prior art, the beneficial effects of the present invention are:

[0041] (1) For the method for improving the infrared thermal imaging personnel recognition rate of the present invention, by using a thermal imager as the acquisition device for the environmental information of the underground personnel safety protection system and taking the image features of the thermal imager as the input of the deep learning model, the environmental adaptability of the underground image recognition system can be improved, ensuring that the system can clearly image in harsh environments with high dust and low illuminance, thereby improving the reliability of the safety protection system.

[0042] (2) For the method for improving the infrared thermal imaging personnel recognition rate of the present invention, by using the irregular motion characteristics of the human body target and the non-motion characteristics of the humanoid heat source target, it is determined whether the heat source target is a human body target through the historical motion trajectory, reducing the interference of the humanoid heat source target in the underground on the model recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0044] Figure 1 It is a schematic flowchart of a method for improving the infrared thermal imaging personnel recognition rate of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0045] With reference to the accompanying drawings in the embodiments of the present invention, the technical solutions in the embodiments of the present invention are clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other implementation manners obtained by those of ordinary skill in the art without creative efforts fall within the scope protected by the present invention.

[0046] It should be noted that the structures, ratios, sizes, etc. shown in the drawings of this specification are only used to cooperate with the content disclosed in the specification for those who are familiar with this technology to understand and read, and are not used to limit the limited conditions under which the present invention can be implemented. Therefore, they do not have technical substance significance. Any modification of the structure, change of the proportional relationship, or adjustment of the size, without affecting the effects that the present invention can produce and the purposes that can be achieved, should fall within the scope covered by the technical content disclosed in the present invention. It should be noted that in this specification, relational terms such as first and second are only used to distinguish one entity from several other entities, and do not necessarily require or imply any actual relationship or order between these entities.

[0047] A method for improving the recognition rate of infrared thermal imaging of personnel according to the present invention includes the following steps:

[0048] S10: Collect video images of mine working conditions through a thermal imager, and extract high-level features and low-level features of each image in the video images through a feature extraction network;

[0049] S20: Establish a Gaussian pyramid and a Laplacian pyramid of the high-level features, and fuse them with the corresponding low-level features in the feature channel dimension to obtain fused features;

[0050] S30: Calculate the spatial contrast of the fused features based on the pixel value gradient of the fused features, and use the connected domain with a contrast greater than the set threshold as the candidate target domain;

[0051] S40: Respectively obtain the positions of the centroids of each candidate target domain in the video images, and generate the motion trajectories of the centroids of each candidate target domain based on the sequence of image frames;

[0052] S50: Use the candidate target domain corresponding to the irregular motion trajectory as the personnel recognition result.

[0053] For a clearer description of a method for improving the recognition rate of infrared thermal imaging of personnel according to the present invention, the following combines Figure 1 Details of each step in the embodiments of the present invention are elaborated.

[0054] Step S10, collect video images of mine working conditions through a thermal imager, and extract high-level features and low-level features of each image in the video images through a feature extraction network.

[0055] In one example of the present invention, for each frame of the video image of the mine working conditions collected by the thermal imager, the temperature data corresponding to each pixel point of the image is obtained, and a temperature matrix corresponding to the image is generated according to the corresponding relationship of the pixel positions of the image. The size of the temperature matrix is W*TH, where W is the pixel width of the image corresponding to the temperature matrix, and TH is the pixel height of the image corresponding to the temperature matrix. The value of each element in the temperature matrix is the actual temperature value of the corresponding pixel point in the infrared image of the thermal imaging.

[0056] Before extracting the high-level features and low-level features of each image in the video image in step S10, there is also a step of infrared image enhancement, including:

[0057] Performing high-frequency extraction of each image in the video image and photometric fusion of the corresponding visible light image of each image through a set of multiple Gaussian filters to obtain a multi-scale enhanced image;

[0058] Performing weighted fusion on each scale enhanced image of the multi-scale enhanced image to obtain an enhanced image corresponding to each image in the video image.

[0059] The multi-scale enhanced image is:

[0060]

[0061] where h k and y k are the high-frequency component and the corresponding enhanced image obtained by high-frequency extraction and photometric fusion at the k-th scale, k represents the number of scales, W k is the Gaussian operator corresponding to the k-th scale, x k is the feature of the visible light image corresponding to the infrared image at the k-th scale, and a is a preset hyperparameter lower than the average value of the image pixels, is the set photometric fusion factor.

[0062] Performing weighted fusion on each scale enhanced image of the multi-scale enhanced image to obtain an enhanced image corresponding to each image in the video image:

[0063]

[0064] where Y is the enhanced image corresponding to an image in the video image, and y t is the enhanced image obtained by high-frequency extraction and photometric fusion at the t-th scale.

[0065] In step S20, a Gaussian pyramid and a Laplacian pyramid of the high-level features are established and fused with the corresponding low-level features in the feature channel dimension to obtain fused features.

[0066] Step S30: Calculate the spatial contrast of the fusion feature based on the pixel value gradient of the fusion feature, and use the connected region with a contrast greater than the set threshold as the candidate target region.

[0067] In step S30, the method for calculating the spatial contrast of the fusion feature based on the pixel value gradient of the fusion feature is as follows:

[0068] Extract the target region and the background region of the fusion feature, and calculate the information entropy of the target region and the background region respectively;

[0069] Take the absolute value of the difference in information entropy within a set width range at the adjacent part of the target region and the background region as the spatial contrast of the fusion feature.

[0070] The method for calculating the information entropy of the target region and the background region respectively is as follows:

[0071] Divide the gray values of the target region and the background region into a set number of discrete gray levels;

[0072] Calculate the proportion of the number of pixel points in each gray level in the pixel points of the entire fusion feature respectively, and obtain the pixel probability of each gray level;

[0073] Based on the pixel probability of each gray level, calculate the information entropy of the target region and the background region respectively.

[0074] The calculation method of the information entropy is as follows:

[0075]

[0076] where H is the information entropy, M is the number of gray levels, and p(u) is the pixel probability of the u-th gray level.

[0077] Take the absolute value of the difference in information entropy within a set width range at the adjacent part of the target region and the background region as the spatial contrast of the fusion feature.

[0078] Step S40: Obtain the position of the centroid of each candidate target region in the video image respectively, and generate the motion trajectory of the centroid of each candidate target region based on the sequence of image frames.

[0079] Before obtaining the motion trajectory of the centroid of the target region, there is also a step of preliminary screening of the target region, and the method is as follows:

[0080] Obtain the temperature value in the temperature matrix corresponding to each pixel of the target region, and perform SVM data classification on the temperature data corresponding to each target region respectively:

[0081] If the temperature data conforms to the temperature range of the human body target, it is initially judged as a human body, and its corresponding target domain meets the requirements; if the temperature data conforms to the temperature range of objects such as motors and cables, it is initially judged as a non-human body, and its corresponding target domain does not meet the requirements.

[0082] Clear the target domains that do not meet the requirements to obtain the candidate target domains after preliminary screening.

[0083] Using the temperature data of each pixel point in the image collected by the infrared thermal imager, reconstruct the temperature data to form a temperature matrix. Input the temperature image formed by the infrared camera into the improved YOLO V5l algorithm to obtain the human body. Since the temperature of each object is different, through the temperature matrix, it can be judged whether the contour of the heating object collected by the thermal imager is a human body, thereby improving the human body target recognition rate of the system. The specific implementation process is as follows:

[0084] Form a temperature matrix with the temperature data collected by the infrared thermal imager according to the position of each pixel in the infrared image; the size of the temperature matrix is W*TH, and the value of each element in the temperature matrix represents the actual temperature value of the corresponding pixel point in the thermal imaging infrared image. (Among them, there are W rows of pixel points in the picture, and each row has TH pixels)

[0085] Collect the thermal imaging data of the tunneling face by using the infrared thermal imager, and label the temperature matrix data corresponding to the human body target and heating objects such as motors, cables, and water pumps in the collected infrared image. Use the SVM algorithm to classify them to obtain data sets such as the temperature range of the human body target in the underground tunneling face based on the thermal imager and the heating temperature range during motor operation.

[0086] Take the consecutive single-frame infrared images obtained from the thermal imaging video collected by the thermal imager as the input of the improved YOLOV5L algorithm, and finally obtain the coordinates of the detection boxes labeled as human bodies.

[0087] Perform SVM data classification on the temperature data within the marked detection box. If the temperature data conforms to the temperature range of the human body target, it is judged as a human body; if the temperature data conforms to the temperature range of objects such as motors and cables, it is judged as a non-human body.

[0088] Among them, the specific implementation process of SVM is as follows:

[0089] Use the thermal imager to collect the scene data on site, including the collection of on-site temperature data and the corresponding collection pictures; the collection includes data of different human body targets and different non-human body targets of the same camera at different distances and different ambient temperatures.

[0090] Extract the temperature matrix W*TH of the sample data collected on-site (where there are W rows of pixel points in the image, and each row has TH pixels), and manually label the target samples with the image data of the corresponding on-site infrared image. The samples are divided into two types: humanoid targets and human targets.

[0091] Input the sample set s = {(x i ’, y i ), i = 1, 2, 3 ······ n} composed of the feature vectors of the samples into the PCA-SVM training machine. x i ’ is a data vector (with a very high dimension); y i is a constraint condition.

[0092] Perform dimensionality reduction on the data: Use the principal component analysis method (PCA) to perform dimensionality reduction on the data, reduce the impact of invalid and incorrect data on modeling, and improve the accuracy of modeling and the operation speed of the model.

[0093] The sample set after dimensionality reduction is: s = {(x i , y i ), i = 1, 2, 3 ······ n}, x i ∈ R.

[0094] Determine the SVM vector machine expression using the following steps:

[0095] Since the sample set is linearly inseparable, use the Gaussian kernel function to map the data to a high-dimensional space:

[0096]

[0097] where: σ 2 is the variance, e is the natural logarithm, K is the Gaussian kernel function; x i , x j are both vectors in the i-th dimension and the j-th dimension in the h-dimensional space after dimensionality reduction;

[0098] Assume that on the h-dimensional space, all samples can be linearly separable by the optimal classification plane.

[0099] To allow the machine to have some misclassified points, usually add a slack variable ζ > 0 to the constraint condition and add a constant C as a penalty factor. The optimal classification plane of the sample set can be solved by the following formula:

[0100]

[0101] where ω is the normal vector of the classification hyperplane of the linear classification, which determines the direction of the hyperplane, b is the displacement term that determines the distance between the classification hyperplane and the origin, the classification hyperplane can be determined by ω and b, ξ iis the relaxation vector in the i-th dimension, h is the dimension of the space, that is, the number of samples.

[0102] For the solution to the above problem, the SVM regression expression y(x) can be obtained:

[0103]

[0104] where y i is the constraint condition, and b0 is the classification threshold;

[0105] Through KKT, the constraint on λ i is:

[0106]

[0107] where λ i , λ j are the Lagrange product factors corresponding to different support vectors in the h-dimensional space.

[0108] The constraint condition of the above formula is:

[0109] y i (ω T X i +b)-1≥0, i = 1, 2, …, h

[0110] ζ > 0

[0111]

[0112] 0 ≤ λ i ≤ C

[0113] ω T is the transposed vector of the classification hyperplane normal vector for linear classification;

[0114] The parameters σ of the Gaussian kernel function and the penalty factor C are determined by the following method:

[0115] (1) Randomly divide the samples into m parts, 1 < m < n; select m - 1 as the test samples and n - m + 1 as the training data;

[0116] (2) Given the ranges of σ and C values, and determine the initial values of σ 2 and C;

[0117] (3) Substitute the data σ and C into the obtained SVM classification function to get the category of the test data, and judge whether it is classified correctly;

[0118] (4) Increase or decrease the value of σ using the dichotomy method according to the classification correctness, and traverse the σ within the interval 2 value, 2For the value, select the σ with the highest classification accuracy as the parameter of the final Gaussian kernel function.

[0119] Step S50, use the candidate target domain corresponding to the irregular motion trajectory as the personnel recognition result.

[0120] In the above embodiments, although the various steps are described in the above sequential order, those skilled in the art can understand that in order to achieve the effects of this embodiment, different steps do not have to be executed in such an order. They can be executed simultaneously (in parallel) or in a reversed order, and these simple changes are all within the protection scope of the present invention.

[0121] A system for improving the infrared thermal imaging personnel recognition rate according to the second embodiment of the present invention includes:

[0122] A data acquisition and feature extraction module, configured to collect mine working condition video images through a thermal imager, and extract high-level features and low-level features of each image in the video images through a feature extraction network;

[0123] A feature fusion module, configured to establish a Gaussian pyramid and a Laplacian pyramid of the high-level features, and fuse them with the corresponding low-level features in the feature channel dimension to obtain fused features;

[0124] A target candidate domain screening module, configured to calculate the spatial contrast of the fused features based on the pixel value gradient of the fused features, and use the connected domain with a contrast greater than a set threshold as the candidate target domain;

[0125] A motion trajectory extraction module, configured to respectively obtain the positions of the centroids of each candidate target domain in the video images, and generate the motion trajectories of the centroids of each candidate target domain based on the sequence of image frames;

[0126] A target recognition module, configured to use the candidate target domain corresponding to the irregular motion trajectory as the personnel recognition result.

[0127] It should be noted that the system for improving the infrared thermal imaging personnel recognition rate provided in the above embodiments is only illustrated by the above division of each functional module. In actual applications, the above functions can be allocated to different functional modules according to needs, that is, the modules or steps in the embodiments of the present invention can be further decomposed or combined. For example, the modules in the above embodiments can be combined into one module, or further split into multiple sub-modules to complete all or part of the functions described above. For the names of the modules and steps involved in the embodiments of the present invention, they are only used to distinguish each module or step, and are not regarded as an improper limitation of the present invention.

[0128] An electronic device according to a third embodiment of the present invention includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein, the memory stores instructions executable by the processor, and the instructions are used to be executed by the processor to implement the above method for improving the recognition rate of infrared thermal imaging of personnel.

[0129] A computer-readable storage medium according to a fourth embodiment of the present invention stores computer instructions, and the computer instructions are used to be executed by the computer to implement the above method for improving the recognition rate of infrared thermal imaging of personnel.

[0130] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes and related descriptions of the above-described storage device and processing device can refer to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0131] Those skilled in the art should be able to realize that the modules and method steps of each example described in combination with the embodiments disclosed in this article can be implemented by electronic hardware, computer software, or a combination of the two. The programs corresponding to the software modules and method steps can be placed in random access memory (RAM), memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disks, removable disks, CD-ROMs, or any other form of storage medium well-known in the art. To clearly illustrate the interchangeability of electronic hardware and software, the components and steps of each example have been generally described according to functions in the above description. Whether these functions are executed in the form of electronic hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0132] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention should be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A method for improving the recognition rate of infrared thermal imaging personnel, characterized in that: The following steps are involved: S10: Collecting video images of mine working conditions through a thermal imager, and extracting high-level features and low-level features of each image in the video images through a feature extraction network; S20: Establishing the Gaussian pyramid and the Laplacian pyramid of the high-level features, and fusing them with the corresponding low-level features in the feature channel dimension to obtain fused features; S30: Based on the pixel value gradient of the fused feature, the spatial contrast of the fused feature is calculated, and the connected domain with a contrast greater than a set threshold is taken as a candidate target domain; the method for calculating the spatial contrast of the fused feature based on the pixel value gradient of the fused feature is: extracting the target area and the background area of ​​the fused feature, and calculating the information entropy of the target area and the background area respectively; taking the absolute value of the difference in the information entropy within the set width range at the junction of the target area and the background area as the spatial contrast of the fused feature; S40: respectively obtaining the position of the centroid of each candidate target domain in the video image, and generating a motion trajectory of the centroid of each candidate target domain based on a sequence of image frames; S50: taking the candidate target domain corresponding to the irregular motion trajectory as a person recognition result.

2. The method for improving the recognition rate of infrared thermal imaging personnel according to claim 1, characterized in that: Before extracting the high-level features and low-level features of each image in the video image in step S10, an infrared image enhancement step is also provided, including: By using multiple Gaussian filters, high-frequency extraction of each image in the video image and photometric fusion of the visible light image corresponding to each image are performed to obtain a multi-scale enhanced image; Each scale enhanced image of the multi-scale enhanced images is weightedly fused to obtain an enhanced image corresponding to each image in the video image.

3. The method for improving the recognition rate of infrared thermal imaging personnel according to claim 2, characterized in that: The multi-scale enhanced image is: Among them, h k and k is the high-frequency component and the corresponding enhanced image obtained by high-frequency extraction and photometric fusion at the kth scale, k represents the number of scales, W k is the Gaussian operator corresponding to the kth scale, x k is the feature of the visible light image corresponding to the infrared image of the kth scale, a is a preset hyperparameter lower than the average value of the image pixels, is the set photometric fusion factor.

4. The method for improving the recognition rate of infrared thermal imaging personnel according to claim 3, characterized in that: The enhanced image corresponding to each image in the video image is: Among them, Y is the enhanced image corresponding to an image in the video image, y t It is the enhanced image obtained by high-frequency extraction and photometric fusion at the t-th scale.

5. The method for improving the recognition rate of infrared thermal imaging personnel according to claim 4, characterized in that: The method of respectively calculating the information entropy of the target area and the background area is: Dividing the grayscale values ​​of the target area and the background area into a set number of discrete grayscale levels; Calculate the proportion of the number of pixels in each gray level in the entire fusion feature to obtain the pixel probability of each gray level; Based on the pixel probability of each of the gray levels, the information entropy of the target area and the background area is calculated respectively.

6. The method for improving the recognition rate of infrared thermal imaging personnel according to claim 5, characterized in that: The information entropy is calculated as follows: Among them, H is the information entropy, M is the number of gray levels, and p(u) is the pixel probability of the u-th gray level.

7. A system for improving the recognition rate of infrared thermal imaging personnel, used to implement the method for improving the recognition rate of infrared thermal imaging personnel as claimed in any one of claims 1 to 6, characterized in that: include: A data acquisition and feature extraction module, configured to acquire video images of mine working conditions through a thermal imager, and extract high-level features and low-level features of each image in the video images through a feature extraction network; A feature fusion module is configured to establish a Gaussian pyramid and a Laplacian pyramid of the high-level features, and fuse them with the corresponding low-level features in the feature channel dimension to obtain a fused feature; a target candidate domain screening module, configured to calculate the spatial contrast of the fused feature based on the pixel value gradient of the fused feature, and take the connected domain with a contrast greater than a set threshold as a candidate target domain; A motion trajectory extraction module is configured to respectively obtain the position of the centroid of each candidate target domain in the video image, and generate a motion trajectory of the centroid of each candidate target domain based on a sequence of image frames; The target recognition module is configured to use the candidate target domain corresponding to the irregular motion trajectory as a person recognition result.

8. An electronic device, characterized in that: include: at least one processor; and a memory communicatively coupled to at least one of the processors; The memory stores instructions executable by the processor, and the instructions are used to be executed by the processor to implement a method for improving the infrared thermal imaging personnel recognition rate as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to be executed by the computer to implement the method for improving the infrared thermal imaging personnel recognition rate as described in any one of claims 1 to 6.

Images