AI training image enhancement method and system based on thermal imaging life detection

Abstract

The invention discloses an AI training image enhancement method and system based on thermal imaging life detection. Collected images are divided according to seasons; selecting a plurality of to-be-matted images containing different human body forms and M background images of typical environments from each season category; cutting out a human body in the image to be cut out, then splicing a plurality of cut-out human body targets into N foreground images which contain a plurality of human body targets and are provided with transparent base images, and then performing multi-angle rotation to obtain 6N foreground images; human body targets in the obtained 6N foreground images are labeled; and carrying out superposition synthesis on the obtained 6N foreground images and annotations and M background images of typical environments to generate M * 6N synthetic pictures and annotations of different human body targets and backgrounds, and then adding the synthetic pictures and annotations into an AI data set to complete AI training image enhancement. The diversity of the training data is increased, and the generalization ability of the AI model is improved; and by utilizing the existing image resources, the cost and difficulty of extra data acquisition are reduced.

Description

Technical Field

[0001] The present invention belongs to the field of AI technology, and in particular relates to an AI training image enhancement method and system based on thermal imaging life detection. Background Art

[0002] In the wild environment, traditional search and rescue operations often face great challenges due to factors such as complex terrain, changeable climate and possible dangerous creatures. These factors not only make the search and rescue inefficient, but also may bring risks to the rescuers. However, with the development of technology, the application of new flying platforms such as drones has brought new solutions to search and rescue work.

[0003] As an efficient aerial platform, drones can carry a variety of sensors and equipment, among which thermal imaging optoelectronic pods are important tools for life detection. This equipment can capture infrared radiation emitted by human bodies on the ground, thereby quickly locating human targets in complex environments. Compared with traditional optical cameras, thermal imaging technology is not limited by lighting conditions and can work at night or in bad weather conditions, greatly improving the flexibility and efficiency of search and rescue operations.

[0004] In order to further improve the application effect of drones in life search, artificial intelligence (AI) technology has been introduced into the data processing link. Through AI algorithms such as deep learning, the system can automatically identify human targets from a large number of thermal imaging images. This process involves multiple steps such as feature extraction, pattern recognition and decision analysis of images. By learning different types of human morphological images, the AI ​​model can gradually improve its accuracy and generalization ability of human detection.

[0005] However, in practical applications, it is difficult to fully collect images of various human forms in different environments, which limits the diversity of training data for AI models. To meet this challenge, researchers have tried to use image enhancement technology to expand the training data set. The commonly used AI Copy-Paste image enhancement cannot rotate the target, and has limited improvement in the generalization ability of human detection.

[0006] Therefore, in order to further improve the generalization ability of AI models, more advanced image enhancement technologies need to be developed. For example, deep learning technologies such as generative adversarial networks (GANs) can be used to generate images with new environments and new human forms. These technologies can simulate human forms and environmental changes in real scenes, thereby providing AI models with richer and more diverse training data.

[0007] In short, by combining drones, thermal imaging technology and AI algorithms, the efficiency and accuracy of searching for life in the wild can be effectively improved. In the future, with the continuous advancement of related technologies, this solution will become more complete and efficient. Summary of the invention

[0008] The technical problem to be solved by the present invention is to provide an AI training image enhancement method and system based on thermal imaging life detection in response to the deficiencies in the above-mentioned prior art. In the existing collected images, human targets are cut out, rotated at multiple angles, and then superimposed on different environmental background images to generate new images, which are used to solve the technical problem of insufficient detection generalization ability caused by insufficient comprehensive images in AI training.

[0009] The present invention adopts the following technical solutions: An AI training image enhancement method based on thermal imaging life detection includes the following steps: S1, divide all the collected images into spring and autumn, winter and summer according to the season when the images were collected; S2, selecting a plurality of preparation images containing different human body shapes and M background images of typical environments from each season category obtained in step S1; S3, cutting out the human body in the prepared cutout image obtained in step S2, and then splicing the cutout multiple human targets into N foreground images with a transparent base map containing multiple human targets, and then performing multi-angle rotation to obtain 6N foreground images; S4, marking the human targets in the 6N foreground images obtained in step S3; S5. Superimpose and synthesize the 6N foreground images and annotations obtained in step S4 with the M background images of typical environments obtained in step S2 to generate M*6N composite images and annotations of different human targets and backgrounds, and then add them to the AI ​​data set to complete AI training image enhancement.

[0010] Preferably, in step S1, temperatures below 13°C are classified as winter, temperatures between 13°C and 26°C are classified as spring and autumn, and temperatures above 26°C are classified as summer.

[0011] Preferably, in step S2, the human target in the background image is deleted.

[0012] Preferably, step S3 is specifically: Determine the cutout method, cut out the human body part, save the cutout human body part as a foreground image with a transparent background, and then splice it into N foreground images with transparent background images containing multiple human targets, rotate the N foreground images by 0°, 15°, 30°, 45°, 60° and 75° respectively to generate 6N foreground images.

[0013] Preferably, in step S4, a labeling tool is used to carefully label the human targets in the 6N foreground images.

[0014] Preferably, during the labeling process, a unique label is assigned to each detected target.

[0015] Preferably, in step S5, the 6N foreground images are superimposed on the M background images of typical environments by using the AlphaBlend method to generate M*6N composite images of different human targets and backgrounds, and at the same time generate their respective corresponding annotation files.

[0016] In a second aspect, an embodiment of the present invention provides an AI training image enhancement system based on thermal imaging life detection, comprising: A division module divides all acquired images into spring and autumn, winter and summer according to the season when the images were acquired; A selection module selects multiple cutout images containing different human body shapes and M background images of typical environments from each season category; The stitching module cuts out the human body in the prepared image, and then stitches the cut-out human body targets into N foreground images with a transparent background containing multiple human body targets, and then performs multi-angle rotation to obtain 6N foreground images; The labeling module labels the human targets in the obtained 6N foreground images; The enhancement module superimposes and synthesizes the obtained 6N foreground images and annotations with M background images of typical environments to generate M*6N synthetic images and annotations of different human targets and backgrounds, and then adds them to the AI ​​dataset to complete AI training image enhancement.

[0017] In a third aspect, a computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned AI training image enhancement method based on thermal imaging life detection when executing the computer program.

[0018] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, including a computer program, which, when executed by a processor, implements the steps of the above-mentioned AI training image enhancement method based on thermal imaging life detection.

[0019] Compared with the prior art, the present invention has at least the following beneficial effects: A method for AI training image enhancement based on thermal imaging life detection can ensure that the AI ​​training data set contains image samples under various environmental conditions by classifying images according to seasons, thereby improving the adaptability and accuracy of the model in practical applications. In addition, this classification method also helps to synthesize data more targetedly during the image enhancement process, further improving the quality and diversity of training data; by carefully selecting spare images containing different human forms and background images representing typical environments, and ensuring that the background images do not contain human targets, high-quality basic materials can be provided for AI training image enhancement methods. Such preparation not only increases the diversity of the data set, but also enhances the robustness and accuracy of the AI ​​model in the face of complex and changeable practical application scenarios; by accurately cutting out human parts and creating foreground images with transparent backgrounds, high-quality data sets can be prepared for subsequent image synthesis and AI training; by carefully annotating human targets in N foreground images, valuable learning resources can be provided for AI models; although this process is challenging, through reasonable methods and tool selection, as well as rigorous workflows, high-quality completion of the annotation work can be ensured, thereby improving the performance and application value of the AI ​​model.

[0020] Furthermore, the human body in the prepared image is cut out using software, and multiple human targets cut out are spliced ​​into N foreground images with transparent base images containing multiple human targets, so as to facilitate superposition and synthesis with multiple background images.

[0021] Furthermore, the N foreground images are rotated by 0 / 15 / 30 / 45 / 60 / 75 degrees respectively to generate 6N foreground images, which expands the target morphology and enhances the generalization ability of AI. Since the mosaic enhancement will be performed vertically and horizontally in the later stage, only a 90-degree internal rotation is required here.

[0022] Furthermore, 6N foreground images and annotations are superimposed and synthesized with M background images to generate M*6N synthetic images and annotations of different human targets and backgrounds, which can quickly generate a large number of data sets for training.

[0023] It can be understood that the beneficial effects of the second to fourth aspects mentioned above can be found in the relevant description of the first aspect mentioned above, and will not be repeated here.

[0024] In summary, the present invention cuts out human targets from existing collected images, rotates them at multiple angles, and then superimposes them on different environmental background images to generate new images, which are used to solve the problem of insufficient detection generalization ability caused by incomplete images in AI training.

[0025] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 This is a schematic diagram of the cutout software interface; Figure 2 Generate an interface map for the foreground image; Figure 3 Rotate the interface image for the foreground image; Figure 4 Labeling interfaces for the human body; Figure 5 To enhance the data effect diagram; Figure 6 It is a schematic diagram of the process of the present invention; Figure 7 A schematic diagram of a computer device provided by an embodiment of the present invention; Figure 8 The block diagram is a chip provided according to an embodiment of the present invention. DETAILED DESCRIPTION

[0027] 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 ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0028] In the description of the present invention, it should be understood that the terms “include” and “comprises” indicate the presence of described features, wholes, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or collections thereof.

[0029] It should also be understood that the terms used in the present specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0030] It should be further understood that the term "and / or" used in the present specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes these combinations. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in the present invention generally indicates that the associated objects are in an "or" relationship.

[0031] It should be understood that, although the terms first, second, third, etc. may be used to describe preset ranges, etc. in the embodiments of the present invention, these preset ranges should not be limited to these terms. These terms are only used to distinguish preset ranges from each other. For example, without departing from the scope of the embodiments of the present invention, the first preset range may also be referred to as the second preset range, and similarly, the second preset range may also be referred to as the first preset range.

[0032] The word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrases "if it is determined" or "if (stated condition or event) is detected" may be interpreted as "when it is determined" or "in response to determining" or "when detecting (stated condition or event)" or "in response to detecting (stated condition or event)", depending on the context.

[0033] Various structural schematic diagrams of the embodiments disclosed in the present invention are shown in the accompanying drawings. These figures are not drawn to scale, and some details are magnified and some details may be omitted for the purpose of clear expression. The shapes of various regions and layers shown in the figures and the relative sizes and positional relationships therebetween are only exemplary, and may deviate in practice due to manufacturing tolerances or technical limitations, and those skilled in the art may additionally design regions / layers with different shapes, sizes, and relative positions according to actual needs.

[0034] In the field of life detection, the combination of thermal imaging technology and AI has brought breakthrough progress in practical applications. The present invention provides an AI training image enhancement method based on thermal imaging life detection. By extracting human targets from existing acquired images and superimposing them on background images of different environments, new images are generated. This can effectively improve the quality and diversity of training data, thereby improving the performance of AI models.

[0035] See also Figure 6 The present invention provides an AI training image enhancement method based on thermal imaging life detection, comprising the following steps: S1. Classification All acquired images were divided into three categories according to the season when the images were acquired: spring and autumn, winter, and summer.

[0036] Season classification plays an important role in AI training image enhancement methods based on thermal imaging life detection. This is because the ambient temperature and the temperature difference between the human body and the environment vary in different seasons, and these factors directly affect the image features captured by the thermal imaging camera. Therefore, classifying images according to seasons helps to better simulate and reproduce various situations in the real environment, thereby improving the generalization ability and detection accuracy of the AI ​​model.

[0037] Specifically, the main basis for seasonal classification is the ambient temperature, which determines the temperature difference between the human body and the environment. In thermal imaging images, the greater the temperature difference, the higher the contrast between the human target and the background, and the easier it is to detect. On the contrary, if the temperature difference is small, the human target may be mixed with the background and difficult to distinguish. The following is the specific temperature range division: Winter (below 13°C): In winter, the ambient temperature is low and the temperature difference between the human body and the environment is large. In this case, the human target in the thermal imaging image usually presents a higher brightness, which is in sharp contrast with the background. Therefore, classifying the environment below 13°C as winter helps to collect human images with high contrast, which is very valuable for training AI models; Spring and autumn (13~26℃): The ambient temperature in these two seasons is moderate, and the temperature difference between the human body and the environment is relatively small. Under such conditions, the contrast between the human target and the background in the thermal imaging image is moderate, neither too obvious nor completely integrated into the background. Therefore, dividing the environment between 13℃ and 26℃ into spring and autumn helps to collect images of the human body with medium contrast, which is very important for improving the detection ability of AI models in common environments; Summer (above 26°C): The ambient temperature is high in summer, and the temperature difference between the human body and the environment is small. Under such conditions, the contrast between the human target and the background in the thermal imaging image is low, and the human target may be difficult to distinguish from the background. Therefore, classifying the environment above 26°C as summer helps to collect human images with low contrast, which is crucial for training the robustness of AI models in extreme environments.

[0038] By classifying images according to seasons, we ensure that the AI ​​training data set contains image samples under various environmental conditions, thereby improving the adaptability and accuracy of the model in practical applications. In addition, it helps to synthesize data more targetedly during the image enhancement process, further improving the quality and diversity of training data.

[0039] S2. Selection In the AI ​​training image enhancement method based on thermal imaging life detection, selecting the right images is crucial for subsequent data enhancement and the performance of the AI ​​model. This process involves selecting specific images from each season category that cover different human forms and represent typical environmental backgrounds. The specific steps are as follows: S201, select the cutout images containing different human body shapes In each season classification, multiple images with different human forms need to be selected as spare images. These images should cover as many forms of the human body as possible in different postures, movements, and clothing. For example, they should include a variety of human postures such as standing, walking, running, sitting, and lying, and take into account the influence of different body shapes and clothing materials. This selection strategy is to ensure that the AI ​​model can recognize and adapt to various forms and states that the human body may appear in the real world.

[0040] S202: Select a background image representing a typical environment In addition to the selection of spare images, you also need to select M background images that can represent typical environments in each season category. These background images should be able to reflect the common environmental features of the season, such as snow scenes in winter, green trees in summer, and shadows under direct sunlight. The purpose of selecting these background images is to create realistic and representative synthetic images in the subsequent image synthesis steps, thereby improving the AI ​​model's adaptability to various real-world scenarios.

[0041] Background image processing requirements When selecting background images, it is important to ensure that these images do not contain any human targets. This is because in the image synthesis stage, the goal is to add cut-out human images on the background to create new training samples. If the background image itself already contains human targets, the synthesized image will not conform to the expected training data format and may mislead the learning process of the AI ​​model. Therefore, for those background images that contain human targets, they need to be processed first (such as erasing or replacing these portraits) to ensure that they are suitable for subsequent image enhancement steps.

[0042] By carefully selecting spare images containing different human forms and background images representing typical environments, and ensuring that the background images do not contain human targets, we provide high-quality basic materials for AI training image enhancement methods. Such preparation not only increases the diversity of the data set, but also enhances the robustness and accuracy of the AI ​​model in the face of complex and changing practical application scenarios.

[0043] S3. Cutout S301, accurately cut out the human body part See also Figure 1 ,The goal of image cutting is to separate the human target from the background. In ,thermal imaging images, the human target usually has a significant temperature contrast ,with the background, which helps to keep the boundary of the human body clear and accurate ,during the image cutting process.

[0044] Commonly used cutout methods are as follows: Use the Magic Wand Tool: The Magic Wand Tool can automatically select areas based on color similarity, which is very suitable for situations where there is a large temperature difference between the human body and the background in thermal imaging images.

[0045] Manual outlining with the pen tool: For situations where the temperature difference with the background is small or there is a complex background, you can use the pen tool to manually draw a path along the outline of the human body to achieve precise cutouts.

[0046] Background subtraction: This method involves taking or acquiring an image of the background without the human target and then subtracting the two images through image processing software to extract the human target.

[0047] S302. Create a foreground image with a transparent background See also Figure 2 Once the human subjects have been successfully cut out from the prepared image, the next step is to save these cut out human subjects as foreground images with transparent backgrounds. This usually involves exporting the cut out parts to an image format that supports transparency, such as PNG. The advantage of transparent backgrounds is that when these foreground images are composited with different background images, they can blend naturally into the new background without leaving any traces of artificial processing.

[0048] S303, Rotation See also Figure 3 , rotate the N foreground images by 0 degrees / 15 degrees / 30 degrees / 45 degrees / 60 degrees / 75 degrees respectively to generate 6N foreground images.

[0049] If a certain image contains multiple human targets, it is recommended to cut out these human targets separately and stitch them into a new foreground image as needed. This not only increases the diversity of the dataset, but also creates synthetic images containing different numbers and postures of human targets, thereby improving the recognition ability of the AI ​​model in various scenarios. During the stitching process, pay attention to maintaining the natural spacing and relative positions between human targets to avoid producing synthetic images that do not conform to the actual situation.

[0050] S4. Annotation See also Figure 4 , use labeling tools (such as labelimg) to carefully label the human targets in the above 6N foreground images. During the labeling process, it is necessary to assign a unique label to each detection target for subsequent model training.

[0051] Selection of annotation tools There are many annotation tools available on the market, such as labelimg, LabelMe, VGG ImageAnnotator (VIA), etc. These tools usually support multiple annotation methods, such as rectangular boxes, polygons, circles, etc., allowing users to assign labels to objects in the image according to their needs. Considering the characteristics of thermal imaging images and the diversity of human targets, choosing a powerful and easy-to-use annotation tool is crucial to improving annotation efficiency.

[0052] When using annotation tools (such as labelimg) to annotate human targets in foreground images, pay attention to the following points: Accuracy: Make sure the annotation boundary matches the edge of the human object as closely as possible. For non-rectangular human poses, use polygonal annotation tools to more accurately depict the human body contour.

[0053] Consistency: Maintaining consistency during the annotation process is critical to training high-quality AI models. This includes aspects such as the shape of the annotations, the naming of the labels, and the level of detail in the annotations.

[0054] Uniqueness: Each detected target is assigned a unique label. This helps to accurately identify and count different human targets during subsequent model training.

[0055] Detailed records: Record detailed information for each annotation, such as coordinate location, size, label name, etc. This information is crucial for training and verifying AI models.

[0056] S5. Synthesis Use image synthesis software to superimpose the annotated 6N foreground images and M background images. Each foreground image and each background image should be synthesized into at least one new image, so that a total of M*6N different synthetic images are generated. At the same time, the synthesis software should maintain the correctness of the annotation information to ensure that each synthetic image has a corresponding annotation file.

[0057] Finally, all synthetic images and annotations of the three categories of spring, autumn, winter, and summer were generated. These newly added images and annotations were added to the AI ​​training dataset to achieve data enhancement.

[0058] See also Figure 5 , which is part of the AI ​​training data set finally generated. It can be seen from the figure that the obtained data set has achieved good training data enhancement.

[0059] When synthesizing images, it is necessary to consider the consistency of lighting direction and shadow distribution. Improper lighting and shadow processing will make the synthesized image look unnatural and even affect the learning effect of the AI ​​model.

[0060] Image processing software such as Adobe Photoshop provides powerful image synthesis functions that can help users better control lighting, shadows and perspective effects, making the synthesized images look more natural and realistic.

[0061] In order to make the target in the foreground image blend naturally into the background image, it may be necessary to scale or rotate the foreground image. These operations need to be performed carefully to avoid changing the true proportion of the target object or causing visual incongruity.

[0062] For image synthesis tasks that require batch processing, you can develop or use ready-made scripts (such as Python scripts combined with the image processing library Pillow) to automate tasks such as resizing, rotation, and annotation information updating to improve work efficiency and accuracy.

[0063] When synthesizing images, it is necessary to ensure that the original annotation information is also accurately transferred to the new synthesized image. This involves the precise conversion and adjustment of the annotation coordinates to ensure that the annotation box or polygon completely corresponds to the human target in the synthesized image.

[0064] Establish a strict quality control process to check each synthesized image one by one to ensure that the lighting, shadows, size and annotation information of each image are correct. This step is time-consuming but crucial to ensuring the quality of the final AI model.

[0065] After the synthesis is completed, the following methods are used to evaluate the impact of the newly added synthetic images on the performance of the AI ​​model, as follows: Cross-validation: Use a portion of the original images and synthetic images for model training, and the other portion for model testing. The effect of data augmentation is evaluated by comparing the performance of the model on the original dataset and the augmented dataset.

[0066] Performance indicator analysis: Pay attention to changes in performance indicators such as accuracy, recall, and F1 score, especially when detecting targets in different postures and lighting conditions.

[0067] The method of the present invention can not only significantly increase the quantity and diversity of training data, but also effectively improve the detection ability and generalization performance of the AI ​​model when facing different environments and human body forms.

[0068] In yet another embodiment of the present invention, an AI training image enhancement system based on thermal imaging life detection is provided. The system can be used to implement the above-mentioned AI training image enhancement method based on thermal imaging life detection. Specifically, the AI ​​training image enhancement system based on thermal imaging life detection includes a division module, a selection module, a splicing module, a labeling module and an enhancement module.

[0069] Among them, the division module divides all the collected images into spring and autumn, winter and summer according to the season when the images were collected; A selection module selects multiple cutout images containing different human body shapes and M background images of typical environments from each season category; A splicing module cuts out the human body in the prepared cutout image, and then splices the cutout human body targets into N foreground images with transparent background images containing multiple human body targets; A labeling module labels the human targets in the N obtained foreground images; The enhancement module superimposes and synthesizes the N foreground images and annotations obtained with M background images of typical environments to generate M*N composite images and annotations of different human targets and backgrounds, and then adds them to the AI ​​dataset to complete AI training image enhancement.

[0070] In another embodiment of the present invention, a terminal device is provided, which includes a processor and a memory, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is suitable for implementing one or more instructions, and is specifically suitable for loading and executing one or more instructions to implement the corresponding method flow or corresponding function; the processor described in the embodiment of the present invention can be used for the operation of the AI ​​training image enhancement method based on thermal imaging life detection, including: According to the season when the images were collected, all the collected images were divided into spring and autumn, winter and summer; from each season category, multiple spare images containing different human body shapes and M background images of typical environments were selected; the human body in the spare images was cut out, and then the multiple human targets were spliced ​​into N foreground images with transparent background images containing multiple human targets, and then multi-angle rotation was performed to obtain 6N foreground images; the human targets in the obtained 6N foreground images were labeled; the obtained 6N foreground images and annotations were superimposed and synthesized with M background images of typical environments to generate M*6N composite images with different human targets and backgrounds and annotations, and then added to the AI ​​dataset to complete AI training image enhancement.

[0071] See also Figure 7 , the terminal device is a computer device. The computer device 60 of this embodiment includes: a processor 61, a memory 62, and a computer program 63 stored in the memory 62 and executable on the processor 61. When the computer program 63 is executed by the processor 61, the method for calculating the fluid composition in the reservoir transformation wellbore in the embodiment is implemented. To avoid repetition, it is not described one by one here. Alternatively, when the computer program 63 is executed by the processor 61, the functions of each model / unit in the AI ​​training image enhancement system based on thermal imaging life detection in the embodiment are implemented. To avoid repetition, it is not described one by one here.

[0072] The computer device 60 may be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server. The computer device 60 may include, but is not limited to, a processor 61 and a memory 62. Those skilled in the art will appreciate that Figure 7 This is only an example of the computer device 60 and does not constitute a limitation of the computer device 60. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the computer device may also include input and output devices, network access devices, buses, etc.

[0073] The processor 61 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0074] The memory 62 may be an internal storage unit of the computer device 60, such as a hard disk or memory of the computer device 60. The memory 62 may also be an external storage device of the computer device 60, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc., equipped on the computer device 60.

[0075] Furthermore, the memory 62 may include both an internal storage unit of the computer device 60 and an external storage device. The memory 62 is used to store computer programs and other programs and data required by the computer device. The memory 62 may also be used to temporarily store data that has been output or is to be output.

[0076] See also Figure 8 , the terminal device is a chip, and the chip 600 of this embodiment includes a processor 622, which may be one or more, and a memory 632 for storing a computer program executable by the processor 622. The computer program stored in the memory 632 may include one or more modules, each corresponding to a set of instructions. In addition, the processor 622 may be configured to execute the computer program to perform the above-mentioned AI training image enhancement method based on thermal imaging life detection.

[0077] In addition, the chip 600 may further include a power supply component 626 and a communication component 650, wherein the power supply component 626 may be configured to perform power management of the chip 600, and the communication component 650 may be configured to implement communication, for example, wired or wireless communication, of the chip 600. In addition, the chip 600 may further include an input / output interface 658. The chip 600 may operate based on an operating system stored in the memory 632.

[0078] In another embodiment of the present invention, the present invention also provides a storage medium, specifically a computer-readable storage medium, which is a memory device in a terminal device for storing programs and data. It can be understood that the computer-readable storage medium here can include both the built-in storage medium in the terminal device and the extended storage medium supported by the terminal device. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and these instructions can be one or more computer programs. It should be noted that the computer-readable storage medium here can be a high-speed RAM memory, or a non-volatile memory (Non-VolatileMemory), such as at least one disk memory.

[0079] One or more instructions stored in a computer-readable storage medium may be loaded and executed by a processor to implement the corresponding steps of the AI ​​training image enhancement method based on thermal imaging life detection in the above embodiment; one or more instructions in the computer-readable storage medium are loaded and executed by the processor as follows: According to the season when the images were collected, all the collected images were divided into spring and autumn, winter and summer; from each season category, multiple spare images containing different human body shapes and M background images of typical environments were selected; the human body in the spare images was cut out, and then the multiple human targets were spliced ​​into N foreground images with transparent background images containing multiple human targets, and then multi-angle rotation was performed to obtain 6N foreground images; the human targets in the obtained 6N foreground images were labeled; the obtained 6N foreground images and annotations were superimposed and synthesized with M background images of typical environments to generate M*6N composite images with different human targets and backgrounds and annotations, and then added to the AI ​​dataset to complete AI training image enhancement.

[0080] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present invention described and shown in the drawings here can usually be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0081] Example 1 S1. Image acquisition First, a large number of thermal imaging images are taken under different environmental conditions, including indoor and outdoor scenes, daytime and nighttime, sunny and rainy and snowy weather, etc. These images will serve as the basic data set for subsequent processing.

[0082] S2. Human target cutout Image segmentation technology is used to accurately cut out human targets from existing thermal imaging images. This step requires dealing with various postures, movements, and occlusions to ensure that the cut-out human targets are as complete and accurate as possible.

[0083] S3. Background image collection Obtain a large number of background images from various sources, which should cover different scenes and environmental conditions, such as city streets, forest grasslands, building interiors, etc. These background images will be used to synthesize new thermal imaging images.

[0084] S4, Image superposition synthesis The N foreground images are rotated by 0 / 15 / 30 / 45 / 60 / 75 degrees respectively to generate 6N foreground images, which are then superimposed on different background images to generate new composite images. In this process, the following factors need to be considered: Size and scale: Ensure that the size and scale of human targets in new images match the actual scene.

[0085] Illumination and Contrast: Adjust the illumination and contrast of the synthetic image to match the style of the real thermal image.

[0086] Occlusion and Shadow: Consider the occlusion and shadow problems that may occur in actual scenes to make the synthesized images more realistic.

[0087] S5. Data Annotation For the generated synthetic images, necessary data annotation work is performed, such as drawing the human body bounding box, key point detection, etc. This annotation information will be used to train and verify the AI ​​model.

[0088] In order to verify the effectiveness of the image enhancement method proposed in this paper, an experiment was conducted using a public thermal imaging life detection dataset. The experimental results show that after using the enhanced dataset to train the AI ​​model, its detection accuracy and recall rate in various scenarios have been improved. In addition, this method also effectively improves the robustness of the model when facing unknown environments and complex backgrounds.

[0089] Example 2 Classification: The thermal images (640*512 png format) collected from July 2023 to August 2024 are divided into three categories according to the ambient temperature (below 13℃: winter; 13~26℃: spring and autumn; above 26℃: summer). Image enhancement processing is performed on each category respectively. The following processing takes the summer category as an example.

[0090] choose: In the summer images, select some images with typical human forms (standing, sitting, squatting, walking, running, lying, prone and various sports postures of men, women, old and young) as the spare images; select M typical environmental backgrounds (squares, roads, wastelands, beaches, woods, etc. at different temperatures and time periods) as the background images sbg01.png~sbgm.png. No human targets should appear in the background images. If there are any, they will be smeared out with the surrounding background; Cutout: Use the "Hi Format Cutout Master" software to import the images to be cut out in turn, use the logo text cutout option to cut out the human target, and then use manual optimization to remove the redundant parts and only keep the required human target (see Figure 1 ), save the cutout result file; Use Paint 3D software to create a fully transparent 640*512 file trans.png.

[0091] Paint 3D opens trans.png, clicks Stickers, and adds the cutout result file (see Figure 2 ) (You can add multiple cutout result files and place them in non-overlapping positions on the screen) and save them as s1.png.

[0092] Continue the above operation until all the clipping result files are used up, generating sn.png (n=1~N) with a total of N foreground images.

[0093] Rotation: Open trans.png in Paint 3D, click Stickers, add the foreground image sn.png, and rotate it by 0 / 15 / 30 / 45 / 60 / 75 degrees to generate 6N foreground images, namely sn_0.png / sn_15.png / sn_30.png / sn_45.png / sn_60.png / sn_75.png. Figure 3 shown.

[0094] Annotation: Open labelimg software and label sn_0.png / sn_15.png / sn_30.png / sn_45.png / sn_60.png / sn_75.png (see Figure 4 ), generate 6N annotation files sn_0.xml / sn_15.xml / sn_30.xml / sn_45.xml / sn_60.xml / sn_75.xml.

[0095] synthesis: Write software to superimpose 6N foreground images and M background images using AlphaBlend to generate M*6N composite images, and at the same time generate the corresponding annotation files.

[0096] Specific process: Take the foreground image s1_0.png and the background image sgb1.png, superimpose them using AlphaBlend to generate the composite image s1_0_sbg1.png; copy s1_0.xml and rename it to s1_0_sbg1.xml to complete the first training enhancement data. Follow the above method to complete M*6N training enhancement data in sequence, such as Figure 4 shown.

[0097] After the summer class is finished, the spring, autumn and winter classes are completed, and all the training enhancement data is achieved. All the enhancement data is added to the AI ​​training set to improve the generalization ability of the AI ​​model after training.

[0098] In summary, the present invention provides an AI training image enhancement method and system based on thermal imaging life detection, which increases the diversity of training data and helps to improve the generalization ability of the AI ​​model. It utilizes existing image resources to reduce the cost and difficulty of additional data collection, and provides a feasible solution to the problems of insufficient training data and insufficient generalization ability.

[0099] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0100] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0101] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in the present invention can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0102] In the embodiments provided by the present invention, it should be understood that the disclosed devices / terminals and methods can be implemented in other ways. For example, the device / terminal embodiments described above are only schematic. For example, the division of the modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0103] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0104] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0105] If the integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0106] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices, and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0107] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0108] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0109] The above contents are only for explaining the technical idea of ​​the present invention and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical idea proposed by the present invention shall fall within the protection scope of the claims of the present invention.

Claims

1. An AI training image enhancement method based on thermal imaging life detection, characterized in that: The following steps are involved: S1, divide all the collected images into spring and autumn, winter and summer according to the season when the images were collected; S2, selecting a plurality of preparation images containing different human body shapes and M background images of typical environments from each season category obtained in step S1; S3, cutting out the human body in the prepared cutout image obtained in step S2, and then splicing the cutout multiple human targets into N foreground images with a transparent base map containing multiple human targets, and then performing multi-angle rotation to obtain 6N foreground images; S4, marking the human targets in the 6N foreground images obtained in step S3; S5. Superimpose and synthesize the 6N foreground images and annotations obtained in step S4 with the M background images of typical environments obtained in step S2 to generate M*6N composite images and annotations of different human targets and backgrounds, and then add them to the AI ​​data set to complete AI training image enhancement.

2. The AI ​​training image enhancement method based on thermal imaging life detection according to claim 1 is characterized in that: In step S1, temperatures below 13°C are classified as winter, temperatures between 13°C and 26°C are classified as spring and autumn, and temperatures above 26°C are classified as summer.

3. The AI ​​training image enhancement method based on thermal imaging life detection according to claim 1 is characterized in that: In step S2, the human target in the background image is deleted.

4. The AI ​​training image enhancement method based on thermal imaging life detection according to claim 1 is characterized in that: Step S3 is specifically as follows: Determine the cutout method, cut out the human body part, save the cutout human body part as a foreground image with a transparent background, and then splice it into N foreground images with transparent background images containing multiple human targets, rotate the N foreground images by 0°, 15°, 30°, 45°, 60° and 75° respectively to generate 6N foreground images.

5. The AI ​​training image enhancement method based on thermal imaging life detection according to claim 1 is characterized in that: In step S4, the human targets in the 6N foreground images are carefully labeled using a labeling tool.

6. The AI ​​training image enhancement method based on thermal imaging life detection according to claim 5 is characterized in that: During the labeling process, a unique label is assigned to each detected object.

7. The AI ​​training image enhancement method based on thermal imaging life detection according to claim 1 is characterized in that: In step S5, the 6N foreground images are superimposed on the M background images of typical environments by using the AlphaBlend method to generate M*6N composite images of different human targets and backgrounds, and at the same time generate their respective corresponding annotation files.

8. An AI training image enhancement system based on thermal imaging life detection, characterized in that: include: A division module divides all acquired images into spring and autumn, winter and summer according to the season when the images were acquired; A selection module selects multiple cutout images containing different human body shapes and M background images of typical environments from each season category; The stitching module cuts out the human body in the prepared image, and then stitches the cut-out human body targets into N foreground images with a transparent background containing multiple human body targets, and then performs multi-angle rotation to obtain 6N foreground images; The labeling module labels the human targets in the obtained 6N foreground images; The enhancement module superimposes and synthesizes the obtained 6N foreground images and annotations with M background images of typical environments to generate M*6N synthetic images and annotations of different human targets and backgrounds, and then adds them to the AI ​​dataset to complete AI training image enhancement.

9. A computer-readable storage medium storing one or more programs, characterized in that: The one or more programs include instructions, which, when executed by a computing device, cause the computing device to perform the method of any one of claims 1 to 7.

10. A computing device, characterized in that include: One or more processors, a memory and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include steps for executing the method according to any one of claims 1 to 7.

Images