Infrared target detection method and system based on deep learning

Abstract

The invention provides an infrared target detection method and system based on deep learning, and relates to the technical field of deep learning and infrared images, and the method comprises the steps: obtaining an original infrared image and environment data corresponding to the original infrared image; marking a target object appearing in the original infrared image to obtain target marking information; training a deep learning model for target detection based on the original infrared image, the environment data and the target labeling information to obtain a target detection model; and inputting the preprocessed to-be-identified infrared image and the corresponding environment data into a target detection model to obtain an infrared target detection result. By implementing the method, the target detection model can be helped to learn the feature modes of the infrared detection target in different environments, the influence of environmental data on infrared detection is considered, the accuracy and efficiency of infrared target detection are improved, and the method is more suitable for efficient infrared target recognition in a complex environment.

Description

Technical Field

[0001] The present application relates to the field of deep learning and infrared image technology, and in particular to an infrared target detection method and system based on deep learning. Background Art

[0002] With the advancement of computer vision technology, especially driven by deep learning algorithms, infrared image target detection technology has made significant progress. Due to its unique imaging characteristics, infrared images can play an important role at night or in environments with obstructed vision, so they are widely used in military reconnaissance, environmental monitoring, and traffic safety.

[0003] Traditional infrared target detection technology mainly relies on manually designed feature extraction and rule matching to identify targets in images. These methods usually include edge detection, shape matching, and motion detection techniques to identify and track target objects.

[0004] However, traditional infrared detection methods perform poorly when dealing with environmental changes (such as changes in lighting and the influence of climatic conditions). Especially in complex environmental conditions, such as haze, rain and snow, or extreme lighting conditions, traditional feature extraction techniques are difficult to accurately capture target features, resulting in a significant decrease in the accuracy of target detection. Summary of the invention

[0005] The present application provides an infrared target detection method and system based on deep learning, which are used to improve the target detection accuracy in infrared images.

[0006] In a first aspect, the present application provides an infrared target detection method based on deep learning, which is applied to an infrared target detection system. The method comprises: Acquire an original infrared image and environmental data corresponding to the original infrared image, the environmental data including geographic location coordinates, temperature and humidity, and features of markers in the image, the features of the markers including the location and category of the markers; Annotate the target object appearing in the original infrared image to obtain target annotation information; Training a deep learning model for target detection based on the original infrared image, the environmental data, and the target annotation information to obtain a target detection model; The preprocessed infrared image to be identified and the environmental data corresponding to the infrared image are input into the target detection model to obtain the infrared target detection result. The preprocessing includes image enhancement, color optimization and texture optimization.

[0007] Through the above embodiments, the infrared target detection system uses the original infrared image marked with the target object and its corresponding environmental data to train the target detection model, and uses the trained model to perform target detection on the preprocessed infrared image. This method can help the target detection model learn the characteristic patterns of infrared detection targets in different environments, consider the influence of environmental data on infrared detection, improve the accuracy and efficiency of infrared target detection, and is more suitable for efficient infrared target recognition in complex environments.

[0008] In some embodiments, the step of acquiring the original infrared image and the environmental data corresponding to the original infrared image specifically includes: Obtaining the original infrared image, geographic location coordinates, and temperature and humidity sent by an infrared detection device, where the infrared detection device is equipped with a positioning system, a temperature sensor, and a humidity sensor; The original infrared image is recognized based on an image recognition model to obtain the features of the markers in the original infrared image.

[0009] Through the above embodiments, the infrared target detection system extracts features from the original infrared image through the image recognition model, thereby improving the efficiency of extracting marker features.

[0010] In some embodiments, before the step of identifying the original infrared image based on the image recognition model to obtain the marker features in the original infrared image, the step further includes: Annotate the original infrared image sample to determine the location and category of the marker in the original infrared image sample; The convolutional neural network model is trained using the labeled original infrared image samples to obtain an image recognition model, which is used to identify the features of the markers in the original infrared image.

[0011] Through the above embodiments, the infrared target detection system can build an image recognition model based on the original infrared images acquired historically before acquiring the landmark features of the original infrared images, which is used to perform landmark features on the original images in the training data set when subsequently constructing the target detection model, thereby improving the efficiency of constructing the training data set.

[0012] In some embodiments, the step of training a deep learning model for target detection based on the original infrared image, the environmental data, and the target annotation information to obtain a target detection model specifically includes: Constructing a training data set based on the original infrared image, the environmental data, and the target annotation information, wherein the training data set includes a training set, a test set, and a validation set; The deep learning model is trained using the training data set to obtain a target detection model, which is used to identify the target object in the original infrared image based on the input original infrared image and environmental data.

[0013] Through the above embodiments, the infrared target detection system constructs a training data set including original infrared images and their corresponding environmental data, which is used to help the target detection model learn the characteristic patterns of infrared detection targets in different environments, thereby improving the adaptability and accuracy of the target detection model in detecting targets in original infrared images in different environments.

[0014] In some embodiments, the step of training the deep learning model using the training data set to obtain the target detection model specifically includes: Divide the collection time of the environmental data into multiple time periods; Dividing the training data set according to the multiple time periods to obtain multiple sub-training data sets; The multiple sub-training data sets are used to train the deep learning model respectively to obtain the target detection model corresponding to each time period.

[0015] Through the above embodiments, the infrared target detection system divides the training data set into multiple sub-training data sets in different time periods and performs model training, so that different target detection models can adapt to the characteristic patterns of infrared detection targets in different time environments (such as winter, summer, etc.), making the target detection model more targeted.

[0016] In some embodiments, after the step of respectively training the deep learning model using the multiple sub-training data sets to obtain the target detection model corresponding to each time period, the method further includes: Determine the acquisition time corresponding to the infrared image to be identified, the acquisition time of the infrared image and the environmental data corresponding to the infrared image being the same; The infrared image to be identified is identified using a target detection model that matches the acquisition time to obtain an infrared target detection result.

[0017] Through the above embodiments, the infrared target detection system can more accurately handle environmental factors related to specific times, such as different conditions at night and during the day, by selecting a model that matches the image acquisition time for infrared detection, thereby further improving the accuracy of detection.

[0018] In some embodiments, after the step of inputting the preprocessed infrared image to be identified and the environmental data corresponding to the infrared image into the target detection model to obtain the infrared target detection result, the method further includes: Annotate the infrared images with recognition errors, and combine the environmental data corresponding to the infrared images to expand the incremental training set; After detecting that the incremental training set exceeds a preset capacity threshold, the object detection model is incrementally trained using the incremental training set.

[0019] Through the above embodiments, the infrared target detection system marks the images with recognition errors and uses these data to perform incremental training of the model, continuously optimizes and adjusts the model, thereby improving the real-time performance of the target detection model and the accuracy of infrared detection.

[0020] In a second aspect, the present application provides an infrared target detection system, the infrared target detection system comprising: one or more processors and a memory; The memory is coupled to the one or more processors, and the memory is used to store computer program code, which includes computer instructions. The one or more processors call the computer instructions so that the infrared target detection system can implement an infrared target detection method based on deep learning provided in the above embodiment, which will not be repeated here.

[0021] In a third aspect, the present application provides a computer-readable storage medium comprising instructions. When the instructions are executed on an infrared target detection system, the infrared target detection system can implement an infrared target detection method based on deep learning provided in the above-mentioned embodiment, which will not be described in detail here.

[0022] In a fourth aspect, the present application provides a computer program product. When the computer program product runs on an infrared target detection system, the infrared target detection system can implement an infrared target detection method based on deep learning provided in the above embodiment, which will not be repeated here.

[0023] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. Using the original infrared images marked with target objects and their corresponding environmental data to train the target detection model not only allows the model to learn the visual features of the target in the infrared image, but also takes into account the impact of environmental factors such as temperature and time on the infrared features, thereby improving the adaptability and accuracy of the model in complex environments, thereby improving the accuracy of infrared target detection.

[0024] 2. Divide the training data set into multiple sub-training data sets for different time periods, and perform special model training on the data set for each time period, so that the model can more accurately adapt to the environmental influences brought about by different seasons or day and night changes, such as changes in temperature and lighting conditions. This not only improves the generalization ability of the model, but also ensures that the most appropriate model can be selected according to the specific time during actual deployment, thereby improving the accuracy of infrared detection.

[0025] 3. Label the images with identification errors and use these data for incremental training of the model, continuously optimize and adjust the model, and improve the real-time performance of the target detection model and the accuracy of infrared detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a flowchart of an infrared target detection method based on deep learning in an embodiment of the present application; Figure 2 is another flowchart of an infrared target detection method based on deep learning in an embodiment of the present application; Figure 3 It is a schematic diagram of the principle of the infrared target detection system in the embodiment of the present application performing target detection on the infrared image to be identified; Figure 4 It is a schematic diagram of the structure of a physical device of the infrared target detection system in the embodiment of the present application. DETAILED DESCRIPTION

[0027] The terms used in the following embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to be used as limitations to the present application. As used in the specification and appended claims of the present application, the singular expressions "one", "a kind of", "said", "above", "the" and "this" are intended to also include plural expressions, unless there is a clear indication to the contrary in the context. It should also be understood that the term "and / or" used in the present application refers to any or all possible combinations comprising one or more listed items.

[0028] In the following, the terms "first" and "second" are used for descriptive purposes only and are not to be understood as suggesting or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features, and in the description of the embodiments of the present application, unless otherwise specified, "plurality" means two or more.

[0029] For ease of understanding, the following is a description of the process of the method provided by this implementation. Figure 1 , which is a flow chart of an infrared target detection method based on deep learning in an embodiment of the present application.

[0030] S101: Acquire an original infrared image and environmental data corresponding to the original infrared image.

[0031] The infrared target detection system first obtains data for training and testing the target detection model. These data include raw infrared images and environmental data corresponding to each raw infrared image. Among them, the raw infrared image refers to the thermal image directly acquired by the infrared camera without any preprocessing. The infrared camera can capture the thermal radiation emitted by the object, so that it can also clearly image in harsh environments or at night, that is, the raw infrared image.

[0032] The environmental data corresponding to the original infrared image contains information about various external factors when the image was collected. These environmental data include the geographic coordinates of the shooting location (such as longitude and latitude), the temperature and humidity at the time, and the location and category of the landmarks.

[0033] For example, in a specific embodiment, the infrared target detection system is deployed on a forest fire monitoring drone, and the original infrared images of the ground are collected through an onboard infrared camera. At the same time, the drone is also equipped with a GPS positioning system that can record the coordinates of the shooting location of each image. The fuselage is also equipped with a temperature and humidity sensor to collect the temperature and humidity data of the environment in real time. For each original infrared image collected, the system matches it with the synchronously recorded GPS coordinates, temperature and humidity data one by one, and uses them together as training data. In addition, the system can also automatically mark easily recognizable objects in the infrared image, such as roads, houses, etc., as visually significant landmarks through image recognition algorithms. The categories and coordinates of these landmarks will also be added to the environmental information of the image.

[0034] It is understandable that geographic coordinates can help the model learn the differences in background features in different regions; temperature and humidity data can reflect the impact of meteorological conditions on the appearance and infrared characteristics of the target; and landmark information provides spatial reference and scale information for the target. Combining environmental data with original infrared images can provide richer semantic information for the training data set of the target prediction model, thereby improving the robustness of target detection.

[0035] S102: Mark the target object appearing in the original infrared image to obtain target marking information.

[0036] After acquiring the original infrared image and its environmental data, the infrared target detection system also needs to mark the target of interest in the original infrared image. For example, if the detected target is an animal, the type and position of the animal appearing in the original infrared image need to be marked. Specifically, the type and position of the target can be marked directly manually or with the help of a marking tool, which is not limited here.

[0037] S103: training a deep learning model for target detection based on the original infrared image, environmental data, and target annotation information to obtain a target detection model.

[0038] The infrared target detection system generates a training data set (including training set, test set and validation set) based on the acquired original infrared image, environmental parameter data and target annotation information, builds a deep learning model (such as Faster R-CNN, YOLO, etc.), inputs the training data set into the deep learning model for training, and automatically learns the key feature representation of the detection target in the training data set through iterative optimization. After multiple rounds of iterative training, when the performance index of the model on the validation set reaches the preset threshold, the model training is considered to be completed, and the final infrared target detection model is obtained. This infrared target detection model can automatically identify each target instance of interest in the infrared image based on the input infrared image and the corresponding environmental data, and give its location coordinates and category label.

[0039] For example, in the forest fire monitoring scenario, the infrared target detection system can use the YOLO v5 model to build an infrared fire detection algorithm. The original infrared images, environmental parameters, and fire point annotation information of multiple scenes and time scales are divided into training, validation, and test sets in proportion. After preprocessing such as scale normalization and data enhancement, the original infrared image is input into the YOLO v5 network together with the environmental parameter features. Through multiple rounds of iterative optimization on the training set, the model gradually learns the infrared imaging characteristics of the fire point target and establishes the robust discrimination ability of fire conditions in different environments. The training process monitors the model performance in real time through the validation set, and finally obtains a target detection model that can efficiently and accurately detect the location of the fire point in the infrared image.

[0040] In addition, in the actual deployment process, the infrared target detection system can select detection models of different performance levels according to specific mission requirements, and balance between detection accuracy and real-time performance to meet the requirements of different occasions. At the same time, the infrared target detection system can also continuously obtain new measured data, regularly iterate and optimize the model, and continuously improve its detection performance and environmental adaptability to ensure the reliability and effectiveness of the detection results.

[0041] S104, inputting the preprocessed infrared image to be identified and the environmental data corresponding to the infrared image into a target detection model to obtain an infrared target detection result.

[0042] When the infrared target detection system completes the training of the detection model, the model can be used to perform real-time detection and recognition on newly acquired infrared image data.

[0043] Specifically, when the infrared target detection system obtains a new infrared image to be detected, it first preprocesses the infrared image, including image enhancement, color optimization, and texture optimization. At the same time, the system obtains environmental parameter data corresponding to the infrared image, such as the geographic coordinates of the shooting location, the current temperature and humidity, etc., and encodes it into a feature vector.

[0044] Next, the system inputs the preprocessed infrared image and the corresponding environmental parameter features into the trained target detection model. The detection model automatically analyzes and identifies the input infrared image, locates the position of each target of interest in the infrared image using the learned target feature pattern, and gives its corresponding category label and confidence score. The system automatically filters out the detection results with confidence scores higher than the threshold according to the pre-set confidence threshold as the final target recognition result. The output results of the target detection model are usually given in the form of bounding boxes and category labels. Each detection result contains the category information and location coordinate information of the target, which can be directly visualized and displayed on the original image.

[0045] In the above embodiment, the infrared target detection system uses the original infrared image marked with the target object and its corresponding environmental data to train the target detection model, and uses the trained model to perform target detection on the preprocessed infrared image. This method can help the target detection model learn the characteristic patterns of infrared detection targets in different environments, consider the influence of environmental data on infrared detection, improve the accuracy and efficiency of infrared target detection, and is more suitable for efficient infrared target recognition in complex environments.

[0046] The following is a more detailed description of the process of the method provided by this implementation. Figure 2 , is another flow chart of an infrared target detection method based on deep learning in an embodiment of the present application.

[0047] S201: Label the original infrared image sample to determine the location and category of the marker in the original infrared image sample.

[0048] The infrared target detection system annotates a batch of original infrared image samples. The markers refer to some prominent and easily identifiable objects or areas in the image, such as buildings, roads, flowers, etc. The specific markers can be determined according to actual needs. These markers can provide important spatial reference information for subsequent target detection.

[0049] Specifically, the annotation process can be done manually or with the help of some semi-automatic image annotation tools. The annotation content mainly includes two aspects: the location and category of the marker. Among them, the location information can be given in the form of a bounding box, that is, marking the approximate range of the marker in the image.

[0050] S202: Use the labeled original infrared image samples to train a convolutional neural network model to obtain an image recognition model.

[0051] The infrared target detection system uses the original infrared image samples annotated in step S201 to train an image recognition model for identifying the marker information in the original infrared image.

[0052] Specifically, the annotated original infrared images are divided into two parts: training set and validation set. The infrared target detection system uses the training set images to train the CNN model, that is, through the back-propagation algorithm, the internal parameters of the model are adjusted so that it can correctly extract and identify the features of various types of markers from the infrared images. During the training process, the validation set data does not participate in back-propagation, but is only used to evaluate the performance of the model so that the training can be terminated in time to prevent overfitting. The trained CNN model (i.e., image recognition model) can detect and locate different categories of markers from any input infrared image.

[0053] In the above embodiment, the infrared target detection system can build an image recognition model based on the original infrared images acquired historically before acquiring the landmark features of the original infrared images, which is used to perform landmark features on the original images in the training data set when subsequently constructing the target detection model, thereby improving the efficiency of constructing the training data set.

[0054] S203, obtaining the original infrared image, geographic location coordinates, and temperature and humidity sent by the infrared detection device.

[0055] Specifically, the original infrared image can be acquired in real time through an infrared camera. The infrared camera can capture the thermal radiation emitted by the object and convert it into a digital image signal to transmit to the processing unit. At the same time, the detection equipment also needs to support positioning and environmental sensing functions. Among them, the positioning function can be achieved through positioning systems such as GPS to record the shooting location of the image. The environmental sensing function can be achieved through a temperature and humidity sensor to record the ambient temperature and humidity during shooting. Then the infrared detection device will send the collected original image and the corresponding position, temperature and humidity data to the infrared target detection system as input data for the target detection model.

[0056] S204: Identify the original infrared image based on the image recognition model to obtain the features of the markers in the original infrared image.

[0057] Specifically, when the infrared target detection system receives a new frame of original infrared image, it is first input into the image recognition model, and the position and category information of different markers in the original infrared image is output.

[0058] In the above embodiment, the infrared target detection system extracts features from the original infrared image through an image recognition model, thereby improving the efficiency of extracting marker features.

[0059] S205: construct a training data set based on the original infrared image, environmental data and target annotation information.

[0060] The infrared target detection system aggregates the original infrared images, environmental parameters, and manually annotated target information to construct a training data set. The original infrared images are collected in real time by infrared cameras. Environmental parameters include the geographic location, temperature, and humidity when the image was taken, which correspond to the original infrared images. The target annotation information is manually marked in each original infrared image to indicate the location and category of the target of interest.

[0061] In the above embodiment, the infrared target detection system constructs a training data set including the original infrared image and its corresponding environmental data to help the target detection model learn the characteristic patterns of infrared detection targets in different environments, thereby improving the adaptability and accuracy of the target detection model in detecting targets in the original infrared images under different environments.

[0062] S206: Divide the collection time of environmental data into multiple time periods.

[0063] Specifically, the infrared target detection system can select appropriate time granularity for division according to the application scenario, such as dividing into four time periods of spring, summer, autumn and winter according to seasons, or dividing into two time periods of day and night according to day and night, etc., which are not limited here.

[0064] S207: Divide the training data set according to multiple time periods to obtain multiple sub-training data sets.

[0065] After completing the division of time periods according to step S206, the infrared target detection system divides the existing training data set accordingly according to the time period corresponding to the acquisition timestamp of each training sample (including the original infrared image and environmental parameters), thereby obtaining a sub-training data set corresponding to each time period.

[0066] After building multiple sub-training data sets, the infrared target detection system further divides each sub-training set into three parts: training set, validation set, and test set. The training set is used for model training, the validation set is used to evaluate model performance and adjust parameters during training, and the test set is used to evaluate the final performance indicators of the trained model. Generally, the ratio of 7:2:1 or 8:1:1 can be used, and there is no limitation here.

[0067] S208. Use multiple sub-training data sets to train the deep learning model respectively to obtain the target detection model corresponding to each time period.

[0068] The process of using the sub-training data set for the deep learning model is basically the same as step S103 and will not be repeated here.

[0069] In the above embodiment, the infrared target detection system divides the training data set into multiple sub-training data sets in different time periods and performs model training, so that different target detection models can adapt to the characteristic patterns of infrared detection targets in different time environments (such as winter, summer, etc.), making the target detection model more targeted.

[0070] S209: Determine the acquisition time corresponding to the infrared image to be identified.

[0071] Since each infrared image is marked with a timestamp when it is collected, the infrared target detection system can directly read the timestamp and save it together with the environmental parameter data as the metadata of the image. When the infrared image to be identified needs to be detected, the infrared target detection system only needs to query its acquisition timestamp to determine which pre-divided time period it belongs to.

[0072] S210 , using a target detection model that matches the acquisition time to identify the infrared image to be identified, to obtain an infrared target detection result.

[0073] After step S208, the infrared target detection system already has multiple target detection models for different time periods. When actually performing infrared target detection, the infrared target detection system performs infrared target detection based on the target detection model of the corresponding time period according to the timestamp marked when each original infrared image is collected. The specific detection process has been described in step S104 and will not be repeated here.

[0074] In the above embodiment, the infrared target detection system can more accurately handle environmental factors related to specific times, such as different conditions at night and during the day, by selecting a model that matches the image acquisition time for infrared detection, thereby further improving the accuracy of detection.

[0075] In addition, when the infrared target detection system uses the trained target detection model to detect new infrared images, some recognition errors are inevitable. In this regard, the infrared target detection system monitors the output results of the target detection model in real time. When it is found that the detection result does not meet expectations, or the confidence is too low, it is determined to be a recognition error. At this time, the infrared target detection system automatically starts the labeling process and re-labels the incorrectly detected targets in the image through artificial intelligence algorithms or manual review. The re-labeled erroneous samples are automatically added to an independent incremental training set. When the sample size of the incremental training set accumulates to a certain extent, that is, exceeds the preset capacity threshold, the system uses the newly collected samples to fine-tune or locally optimize the model so that it can adapt to the new data distribution, thereby improving the real-time performance of the target detection model and the accuracy of infrared detection.

[0076] The following is a schematic diagram of the principle of the infrared target detection system in the embodiment of the present application performing target detection on the infrared image to be identified. Figure 3 shown.

[0077] In the embodiment of the present application, the infrared target detection system performs target detection on the infrared image to be identified in two processes, one of which is the process of constructing the target detection model, and the other is the process of performing target detection on the infrared image to be identified based on the constructed target detection model. Among them, the target detection model construction process is based on the original infrared image and environmental data collected by the infrared detection equipment, as well as the target annotation information of the original infrared image to construct a training data set, and the training data set is used to train the deep learning model, and the deep learning model whose performance index reaches the preset threshold is screened and evaluated as the target detection model. When performing infrared target detection, the infrared image to be identified and the corresponding collected environmental data are put into the target detection model, and the infrared target detection result, that is, the category information and position coordinate information of the target is output.

[0078] The infrared target detection system of the embodiment of the present invention is applied to electronic equipment. Figure 4 A schematic diagram of the architecture of an electronic device suitable for implementing an embodiment of the present invention is shown.

[0079] It should be noted that Figure 4 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0080] Those skilled in the art will appreciate that all or part of the steps in the various methods of the above embodiments can be completed by instructions (computer programs), or by controlling related hardware through instructions (computer programs), and the instructions can be stored in a computer-readable storage medium and loaded and executed by a processor. The electronic device of this embodiment includes a storage medium and a processor, wherein a plurality of instructions are stored in the storage medium, and the instructions can be loaded by the processor to execute any step of the method provided in the embodiment of the present invention.

[0081] Specifically, the storage medium and the processor are electrically connected directly or indirectly to realize data transmission or interaction. For example, these elements can be electrically connected to each other through one or more signal lines. The storage medium stores computer execution instructions for implementing the data access control method, including at least one software function module that can be stored in the storage medium in the form of software or firmware. The processor executes various functional applications and data processing by running the software program and module stored in the storage medium. The storage medium can be, but is not limited to, random access storage medium (Random Access Memory, referred to as: RAM), read-only storage medium (Read Only Memory, referred to as: ROM), programmable read-only storage medium (Programmable Read-Only Memory, referred to as: PROM), erasable read-only storage medium (Erasable Programmable Read-Only Memory, referred to as: EPROM), electrically erasable read-only storage medium (Electric Erasable Programmable Read-Only Memory, referred to as: EEPROM), etc. Among them, the storage medium is used to store programs, and the processor executes the program after receiving the execution instruction.

[0082] Furthermore, the software programs and modules in the above-mentioned storage medium may also include an operating system, which may include various software components and / or drivers for managing system tasks (such as memory management, storage device control, power management, etc.), and may communicate with various hardware or software components to provide an operating environment for other software components. The processor may be an integrated circuit chip having signal processing capabilities. The above-mentioned processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc., which may implement or execute the various methods, steps, and logic flow diagrams disclosed in this embodiment. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0083] Since the instructions stored in the storage medium can execute the steps in any method provided in the embodiments of the present invention, the beneficial effects of any method provided in the embodiments of the present invention can be achieved. Please refer to the previous embodiments for details and will not be repeated here.

[0084] 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 a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A deep learning-based infrared target detection method, applied to an infrared target detection system, characterized in that: The method comprises: Acquire an original infrared image and environmental data corresponding to the original infrared image, wherein the environmental data includes geographic location coordinates, temperature and humidity, and features of markers in the image, wherein the features of the markers include the location and category of the markers; Annotating the target object appearing in the original infrared image to obtain target annotation information; Training a deep learning model for target detection based on the original infrared image, the environmental data, and the target annotation information to obtain a target detection model; The preprocessed infrared image to be identified and the environmental data corresponding to the infrared image are input into the target detection model to obtain the infrared target detection result, wherein the preprocessing includes image enhancement, color optimization and texture optimization.

2. The method according to claim 1, characterized in that: The step of obtaining the original infrared image and the environmental data corresponding to the original infrared image specifically includes: Acquire the original infrared image, geographic location coordinates, and temperature and humidity sent by the infrared detection device, wherein the infrared detection device is equipped with a positioning system, a temperature sensor, and a humidity sensor; The original infrared image is recognized based on an image recognition model to obtain the features of the markers in the original infrared image.

3. The method according to claim 2, characterized in that Before the step of identifying the original infrared image based on the image recognition model to obtain the features of the markers in the original infrared image, the method further includes: Annotating the original infrared image sample to determine the location and category of the marker in the original infrared image sample; The convolutional neural network model is trained using the annotated original infrared image samples to obtain an image recognition model, wherein the image recognition model is used to recognize the features of the markers in the original infrared image.

4. The method according to claim 1, characterized in that The step of training a deep learning model for target detection based on the original infrared image, the environmental data, and the target annotation information to obtain a target detection model specifically includes: Constructing a training data set based on the original infrared image, the environmental data and the target annotation information, wherein the training data set includes a training set, a test set and a validation set; The deep learning model is trained using the training data set to obtain a target detection model, where the target detection model is used to identify the target object in the original infrared image based on the input original infrared image and environmental data.

5. The method according to claim 4, characterized in that The step of using the training data set to train the deep learning model to obtain the target detection model specifically includes: Dividing the collection time of the environmental data into multiple time periods; Dividing the training data set according to the multiple time periods to obtain multiple sub-training data sets; The deep learning model is trained respectively using the multiple sub-training data sets to obtain a target detection model corresponding to each time period.

6. The method according to claim 5, characterized in that After the step of using the multiple sub-training data sets to train the deep learning model respectively to obtain the target detection model corresponding to each time period, the method further includes: Determine the acquisition time corresponding to the infrared image to be identified, the acquisition time of the infrared image and the environmental data corresponding to the infrared image being the same; The infrared image to be identified is identified using a target detection model that matches the acquisition time to obtain an infrared target detection result.

7. The method according to claim 1, characterized in that After the step of inputting the preprocessed infrared image to be identified and the environmental data corresponding to the infrared image into the target detection model to obtain the infrared target detection result, the method further includes: Annotating infrared images with recognition errors, and combining environmental data corresponding to the infrared images to expand the incremental training set; After detecting that the incremental training set exceeds a preset capacity threshold, the object detection model is incrementally trained using the incremental training set.

8. An infrared target detection system, characterized in that: The infrared target detection system includes: one or more processors and memory; The memory is coupled to the one or more processors, and the memory is used to store computer program codes, wherein the computer program codes include computer instructions, and the one or more processors call the computer instructions to enable the infrared target detection system to execute the method according to any one of claims 1 to 7.

9. A computer-readable storage medium comprising instructions, characterized in that: When the instruction is executed on the infrared target detection system, the infrared target detection system is caused to execute the method as claimed in any one of claims 1 to 7.

10. A computer program product, characterized in that When the computer program product runs on an infrared target detection system, the infrared target detection system is enabled to execute the method according to any one of claims 1 to 7.

Images