Space infrared point target recognition method and system based on similarity measure

Abstract

The application relates to the technical field of infrared detection, in particular to a space infrared point target recognition method and system based on similarity measurement and a storage medium, which comprises the following steps: respectively pre-processing a to-be-recognized target infrared radiation intensity time sequence collected by a space infrared point target and a plurality of template infrared radiation intensity time sequences in a template library; respectively cutting a key interval from the pre-processed to-be-recognized target infrared radiation intensity time sequence and the plurality of template infrared radiation intensity time sequences; respectively calculating shape similarity and length similarity of the to-be-recognized target infrared radiation intensity time sequence and the plurality of template infrared radiation intensity time sequences in the key interval; performing similarity measurement on the space infrared point target based on the shape similarity and the length similarity, obtaining a recognition result, and accurately recognizing the space infrared point target by using radiation intensity time sequence characteristics.

Description

Technical Field

[0001] This application belongs to the field of infrared detection technology, specifically a method, system, and storage medium for spatial infrared point target identification based on similarity measurement. Background Technology

[0002] When an infrared detector is far from a target, the target appears as a light spot or even a point of light, losing detailed information such as its shape and texture. Furthermore, monocular infrared detection systems also lose motion information. When the target is far from the infrared detection system, it often appears as a single-pixel bright spot on the imaging plane. Even considering the point spread effect of energy, the target size is generally no larger than 8×8 pixels, hence it is called a point target. These factors pose significant challenges to the identification of point targets in space infrared imaging. Therefore, in the identification of point targets in space infrared imaging, since image data lacks information about the target's shape and size, infrared sensors can only obtain a time series of the target's infrared radiation intensity. Although the one-dimensional sequence signal contains very limited information, observations have shown that the amplitude and frequency of fluctuations in the target's infrared radiation intensity time series can reflect the motion characteristics and inherent physical properties of space infrared point targets, and can be used to distinguish and identify different target types.

[0003] Identifying spatial infrared point targets is a highly challenging task, requiring effective analysis of radiation intensity time-series characteristics, extraction of key identification features, and rational utilization of these features. The main technical difficulties lie in the following three aspects: 1. Spatial infrared point targets lack physical attributes such as shape, structure, size, and texture. Research can only utilize the target's grayscale information on the imaging plane and its prior motion information. 2. Due to the considerable distance from the detector, the signal-to-noise ratio of the target on the imaging plane is often low, or even completely obscured by noise and background clutter. 3. In practical applications, the recognition algorithm requires high real-time performance, while the computational and storage resources of the hardware processing system limit its computational complexity. Therefore, effectively utilizing the radiation intensity time-series characteristics under limited computational resources, eliminating the influence of distance and environment, and accurately identifying spatial infrared point targets has become an urgent problem to be solved. Summary of the Invention

[0004] The purpose of one or more embodiments of this specification is to provide a method for identifying spatial infrared point targets based on similarity measurement, which utilizes the time series characteristics of radiation intensity to accurately identify spatial infrared point targets.

[0005] To solve the above-mentioned technical problems, one or more embodiments of this specification are implemented as follows: Firstly, a method for identifying spatial infrared point targets based on similarity measurement is provided, comprising the following steps: preprocessing the infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of multiple templates in a template library; extracting key intervals from the preprocessed infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of the multiple templates; calculating shape similarity and length similarity for the infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of the multiple templates within the key intervals; and performing similarity measurement on the spatial infrared point target based on the shape similarity and the length similarity to obtain the identification result.

[0006] Secondly, a spatial infrared point target recognition system based on similarity measurement is proposed, comprising: a preprocessing module for preprocessing the infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of multiple templates in a template library; a key region extraction module for extracting key intervals from the preprocessed infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of the multiple templates; a similarity calculation module for calculating shape similarity and length similarity for the infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of the multiple templates within the key intervals; and a result recognition module for performing similarity measurement on the spatial infrared point target based on the shape similarity and the length similarity to obtain the recognition result.

[0007] Thirdly, a storage medium is proposed for computer-readable storage, wherein the storage medium stores one or more types of programs, which, when executed by one or more types of processors, implement the above-mentioned spatial infrared point target recognition method based on similarity measurement.

[0008] As can be seen from the technical solutions provided by one or more embodiments of this specification, the spatial infrared point target recognition method based on similarity measurement provided by the embodiments of the present invention innovatively proposes a new similarity measurement method. Utilizing the characteristics of radiation intensity time series, it comprehensively considers the shape similarity and length similarity between the infrared radiation intensity time series of the spatial infrared point target to be identified and the infrared radiation intensity time series of various types of templates in the template library. The recognition accuracy can reach 85%, and the time required for a single recognition task is only 0.5 seconds. It achieves high recognition accuracy while eliminating the influence of distance and environment, meets timeliness requirements, has low computational resource requirements, and is easy to deploy. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in one or more embodiments or prior art of this specification, the accompanying drawings used in the description of one or more embodiments or prior art will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a flowchart illustrating a spatial infrared point target recognition method based on similarity measurement according to an embodiment of the present invention; Figure 2(a) is a schematic diagram of the simulation data curve of the infrared radiation intensity time series of a target to be identified under one type in the spatial infrared point target identification method based on similarity measurement provided in the embodiment of the present invention. Figure 2(b) is a schematic diagram of the simulation data after normalization of the time series infrared radiation intensity of another target to be identified under the same type as Figure 2(a); Figure 3(a) is a schematic diagram of the time series simulation data curve of the infrared radiation intensity of a target to be identified under type 1(a) in the spatial infrared point target identification method based on similarity measurement provided in the embodiment of the present invention; Figure 3(b) is a schematic diagram of the time series simulation data curve of the infrared radiation intensity of another target to be identified under type 1(b) of Figure 3(a); Figure 3(c) is a schematic diagram of the time series simulation data curve of the infrared radiation intensity of a target to be identified under type 2 in the spatial infrared point target identification method based on similarity measurement provided in the embodiment of the present invention; Figure 4 This is a schematic diagram of the process of measuring the similarity between a spatial infrared point target and a template in the spatial infrared point target recognition method based on similarity measurement provided in the embodiments of the present invention; Figure 5 This is a schematic diagram illustrating the filtering effect of the filtering process in the spatial infrared point target recognition method based on similarity measurement provided in an embodiment of the present invention; Figure 6(a) is a schematic diagram of the time series curve of the infrared radiation intensity of the target to be identified in the spatial infrared point target identification method based on similarity measurement provided in the embodiment of the present invention; Figure 6(b) is a schematic diagram of the time series curve of template infrared radiation intensity in the spatial infrared point target recognition method based on similarity metric provided in the embodiment of the present invention; Figure 7(a) is a schematic diagram of the test key interval extracted from the time series curve of the infrared radiation intensity of the target to be identified in the spatial infrared point target identification method based on similarity measurement provided in the embodiment of the present invention; Figure 7(b) is a schematic diagram of extracting the key interval of the template from the time series curve of the template infrared radiation intensity in the spatial infrared point target recognition method based on similarity measurement provided in the embodiment of the present invention. Figure 8(a) is a schematic diagram of point matching based on Euclidean distance; Figure 8(b) is a schematic diagram of point matching based on the dynamic time warping method in the spatial infrared point target recognition method based on similarity measurement provided in the embodiment of the present invention. Detailed Implementation

[0011] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described one or more embodiments are merely some embodiments of this specification, and not all embodiments. All other embodiments obtained by those skilled in the art based on one or more embodiments of this specification without creative effort should fall within the protection scope of this document.

[0012] This invention provides a spatial infrared point target recognition method based on similarity measurement, which can achieve a high recognition rate. The method and its steps are described in detail below.

[0013] Example 1 Reference Figure 1 As shown in the illustration, a spatial infrared point target recognition method based on similarity measurement is provided by an embodiment of the present invention. This method is based on the time series characteristics of infrared radiation intensity. Due to human cognitive patterns, two-dimensional images are naturally more intuitively perceived. However, as an information carrier, one-dimensional time series representation is relatively abstract for humans. Therefore, analyzing the characteristics of the infrared radiation intensity time series of spatial infrared point targets helps to deepen the understanding of the information contained in the time series, which is the foundation and key to target feature extraction and recognition research.

[0014] The radiation signal received by the infrared detector is determined by factors such as temperature, infrared emissivity, projected area, and observation angle. Assuming the target surface exhibits gray-body radiation and diffuse reflection characteristics, according to Planck's law, the infrared radiation intensity received by the detector's focal plane within the specified band can be approximated as: (1) in, , For the band range, The entrance pupil area of ​​the detector. Let be the projected area of ​​the target in the direction of the detection line of sight. This represents the distance between the detector and the target. The radiation value is the value when the target surface temperature is T. The observation integration time.

[0015] Use constant coefficients for all fixed parameters This means that the definite integral result is expressed as a function of temperature T. If we express that, then equation (1) can be expressed as: (2) As can be seen from equation (2), the infrared radiation intensity is mainly determined by the target surface temperature T, the detection distance R, and the geometric projection area of ​​the target within the line of sight. These factors determine the fluctuations and trends in the time series of infrared radiation intensity.

[0016] The characteristics of the data collected by the infrared detector are analyzed as follows: 1) The sample size is small and the distribution is uneven. In real-world scenarios, acquiring time-series data on the infrared radiation intensity of spatial infrared point targets is extremely costly and opportunities are limited. The sample data used in this embodiment of the invention comprises 10 types and 81 data entries (61 template data entries from the template library and 20 data entries of spatial infrared point targets to be identified). The sample distribution is uneven; the template library contains a maximum of 10 template data entries for each type, and a minimum of only 2. This application does not perform data augmentation because fluctuations and trends contain important and hidden feature information for one-dimensional time series. Randomly cropping, splicing, or adding random noise to these data can easily lead to information loss or even destruction.

[0017] 2) Discontinuous timestamps lead to data loss. Complex environmental disturbances, such as cloud cover, can cause targets to disappear over a period of time. This results in discontinuous timestamps in the target radiation intensity time series, leading to data loss. This data loss is random and cannot be predicted or corrected in advance.

[0018] 3) Noise interference exists. Noise in the infrared radiation intensity time series mainly comes from two aspects: First, there is the noise of the infrared detector itself, such as stripe noise and spot noise. This noise will randomly affect the grayscale value of spatial infrared point targets; second, spatial infrared point targets may undergo random movements such as rolling and rotation during their movement due to maneuvering control or uneven force, and the radiation values ​​of different parts of the target are different, causing the target radiation value captured by the infrared sensor to jump, affecting the recognition result. This is illustrated by the tooth-like undulations in Figure 2.

[0019] 4) The time series of radiation intensity of similar type of spatial infrared point targets show significant differences. The significant differences in the length of the radiation intensity time series for similar space infrared point targets stem primarily from two factors: First, although they are the same type of space infrared point target, their flight time and motion processes differ in different missions; second, when a detector acquires a space infrared point target, it needs to scan and detect it in real time over a large area before monitoring and tracking it. This may prevent the detector from capturing the complete motion process of the space infrared point target. Figures 2(a) and (b) show the simulation data curves after normalizing the radiation intensity time series of two targets of the same type. As can be seen from Figures 2(a) and (b), the shapes of the two sets of data differ due to the different missions performed. Furthermore, the detector's failure to acquire the target in a timely manner results in a significant difference in the length of the two sets of data. This increases the difficulty of type identification.

[0020] 5) The time series of spatial infrared point target radiation intensity are similar among different types. Space infrared point targets belonging to the same series but of different types have very similar infrared radiation intensity time series curves due to their similar structure and principles. As shown in Figures 3(a), (b), and (c), these are simulated data curves of radiation intensity time series after normalization processing for two different types of targets. Intuitively, type 1(a) is more similar to type 2 than type 1(b), which poses a significant challenge for type recognition tasks.

[0021] The spatial infrared point target recognition method based on similarity measurement provided in this invention includes the following steps: S10: Preprocess the infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of multiple templates in the template library, which are collected from spatial infrared point targets respectively; The purpose of preprocessing is to ensure consistency between the infrared radiation intensity time series of the target to be identified and the various types of template infrared radiation intensity time series in the template library, taking into account the aforementioned influencing factors, so as to achieve a unified data standard for subsequent similarity measurement. Preprocessing may include: interpolation completion, filtering, and normalization.

[0022] S20: Extract key intervals from the preprocessed infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of multiple templates, respectively. Peak points of radiation intensity can be found in the preprocessed infrared radiation intensity time series of the target to be identified. Then, a radiation threshold T is set, and the effective range near the peak point of the infrared radiation intensity time series to be identified is extracted as the key interval. Similarly, the above steps are performed on the infrared radiation intensity time series of n types of templates in the template library to obtain the corresponding key intervals.

[0023] S30: Within the key interval, calculate the shape similarity and length similarity of the infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of various templates, respectively; using the Dynamic Time Warping (DTW) algorithm, sequentially calculate the shape similarity between the preprocessed infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of each type of template in the template library. Where n is the number of template types in the template library. Additionally, drawing on the concept of IoU from two-dimensional target detection, the length L of the segment extracted from the infrared radiation intensity time series of the target to be identified is calculated, along with the segment lengths of the infrared radiation intensity time series extracted from each type of template in the template library. Length similarity , where n is the number of template types in the template library.

[0024] S40: Similarity measurement of spatial infrared point targets is performed based on shape and length similarity to obtain the recognition results. The similarity measurement function S is designed as follows: Where i = 1, 2, ..., n, and λ is the weight of shape similarity in the final similarity score. Experimentally, λ = 0.75 was chosen. The maximum value was obtained. The corresponding type k is the identification result of the spatial infrared point target to be identified.

[0025] Optionally, in the spatial infrared point target identification method provided in this embodiment of the invention, the time series of infrared radiation intensity of the target to be identified and the time series of infrared radiation intensity of multiple types of templates in the template library are preprocessed respectively, specifically including: First, linear interpolation is performed to complete the time series of infrared radiation intensity of the target to be identified. To address the data loss problem mentioned above, this embodiment of the invention uses linear interpolation to process the original data of the time series of infrared radiation intensity of the target to be identified. The original data format is... ,in, The absolute time when the infrared detector captures the current frame. Let be the current radiation intensity of the target to be identified. The detector's sampling frequency is n frames / second, so the original data is padded to n sets / second. Given that two adjacent sets of data in the original dataset are... and If the time interval between the two sets of data is greater than the sampling interval (seconds) means: Then, the missing data within the interval needs to be filled in, and the supplemented radiation intensity needs to be filled in. The calculation formula is: (3) Secondly, the infrared radiation intensity time series of the target to be identified, after linear interpolation, is filtered. To address noise interference, this embodiment of the invention selects to filter the original data after linear interpolation to reduce fluctuations in radiation intensity. Common data filtering methods include moving average methods (simple moving average, weighted moving average), exponential sliding average, etc. However, noise interference is irregular. Therefore, to maximize the preservation of the trend of the original data and achieve a smoothing effect, this embodiment of the invention uses a Savitzky-Golay filter for optimization. The Savitzky-Golay filter is a filtering method based on local polynomial least squares fitting in the time domain, which can filter out noise while ensuring that the shape and width of the signal remain unchanged. To adapt to infrared radiation intensity time series of different lengths, the window length (must be an odd number) is set to approximately 1 / 20 of the sequence length, and the order m of the polynomial fitting is set to 5, at which point the effect is optimal. The filtering effect is shown in Figure 6.

[0026] Finally, the time series of infrared radiation intensity of the target to be identified after filtering is normalized. Since shape similarity is calculated using the DTW algorithm, which is sensitive to the amplitude of the time series, normalization is necessary. The normalization formula used in this invention is: (4) Optionally, in the spatial infrared point target identification method provided in this embodiment of the invention, key intervals are extracted from the preprocessed infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of the multiple templates, respectively. Specifically, this includes: determining the amplitude peak point in the infrared radiation intensity time series of the target to be identified; setting amplitude thresholds on both sides of the amplitude peak point, and extracting the test key interval between the amplitude thresholds. Figures 7(a) and (b) show the infrared radiation intensity time series curves of the target to be identified. and the time series curve of infrared radiation intensity of a certain template in the template library The diagram is a preprocessed representation. The infrared radiation intensity time series in this paper is one-dimensional data obtained by sampling at set time intervals and processing. The size of the data represents the magnitude of the radiation value. Plotting this one-dimensional sequence in chronological order yields the corresponding infrared radiation intensity time series curve. Therefore, length similarity and shape similarity can be used to characterize the similarity of one-dimensional time series, as well as the similarity of infrared radiation intensity time series curves.

[0027] Taking Figures 7(a) and (b) as examples, the steps of calculating the shape similarity between the infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of the templates in the template library using the DTW algorithm in this invention are explained as follows: Through observation and extensive experiments, it was found that only data with radiation intensity exceeding a certain threshold in the sequence affects the recognition results. Other data not only do not help the recognition results but may also lower the similarity score between the two sequences and increase the computational load. Therefore, before calculating shape similarity, it is necessary to extract key intervals that are highly correlated with the type of infrared radiation intensity time series. The method for extracting the infrared radiation intensity time series of the target to be identified is as follows: ① Finding the peak point By iterating through all points in the time series of infrared radiation intensity of the target to be identified, the point with the maximum amplitude is found, which is the peak point, denoted as . As shown by the asterisk in Figure 7.

[0028] ② Set the threshold T Traverse both sides of the peak point to find the first point where the radiation intensity is less than the threshold T; this is the key testing interval, denoted as [reference needed]. , Experiments showed that the recognition effect was best when the threshold T=0.1.

[0029] The method for determining the key intervals of templates from the infrared radiation intensity time series of multiple template types in the template library is the same as described above. The endpoints of the key intervals are denoted as follows: , The results of the cropping are shown in Figures 8(a) and (b).

[0030] Optionally, in the spatial infrared point target recognition method provided in this embodiment of the invention, the shape similarity and length similarity are calculated for the infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of multiple templates within the key interval, respectively, specifically including: Based on the dynamic time warping method, the curve segment of the infrared radiation intensity time series curve of the target to be identified in the test key interval is matched with the curve segment of the infrared radiation intensity time series curve of the multi-type template in the template key interval, and the shortest Euclidean distance of the matching points is calculated. The shortest Euclidean distance is remapped to obtain the shape similarity between the infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of the multi-type template.

[0031] Dynamic Time Warping (DTW) differs from Euclidean distance in that it measures sequence similarity based on time points. DTW matches data points in a time series by bending the time points of the time series. A schematic diagram of the matching between the two methods is shown in Figure 8. The core of the DTW method is a point matching technique based on the sequence shape (matching sequence peaks and troughs as closely as possible), and then performing traditional similarity calculations such as Euclidean distance based on these correspondences. The calculation involves the infrared radiation intensity time series curve of the target to be identified. In the interval The curve segment within and the time series curve of infrared radiation intensity of templates in the template library In the interval The shortest Euclidean distance of a curve segment within a given point matching scheme is denoted as . .

[0032] For the final similarity measurement result, the larger the value, the higher the similarity between the two sets of sequences, and the more likely they are to belong to the same type. The result obtained by the DTW algorithm is the Euclidean distance between the two sets of sequences after point matching; the larger the value, the lower the similarity in shape between the two sets of sequences, and the less likely they are to belong to the same type. Therefore, the reciprocal of the DTW calculation result is taken: 1 / .

[0033] Following the steps above, the DTW distance between the infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of various templates in the template library is calculated sequentially, and the reciprocal is taken. Finally, the result is normalized. Ultimately, the shape similarity between the infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of various templates in the template library is obtained. , This represents the number of types in the template library.

[0034] Optionally, the spatial infrared point target identification method provided in this embodiment of the invention calculates shape similarity and length similarity for the infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of multiple templates within a key interval. Specifically, this includes: calculating the intersection-union ratio (IUR) of the test key interval corresponding to the infrared radiation time series curve of the target to be identified and the template key interval corresponding to the infrared radiation time series curve of the multiple templates, thereby obtaining the length similarity between the infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of the multiple templates. The length similarity refers to the IUR of the lengths of the key intervals of the infrared radiation intensity time series curve of the target to be identified and the key intervals of the infrared radiation intensity time series curve of the multiple templates.

[0035] Because some types are extremely similar in shape, and because sample data of the same type may have higher shape similarity between two groups of sequences belonging to different types due to different tasks performed and various noise interferences, using shape similarity alone as a similarity measure has drawbacks.

[0036] Observation revealed that while the lengths of the infrared radiation intensity time series within the same type varied, the differences were relatively small; however, the lengths of the infrared radiation intensity time series across different types differed considerably, as shown in Type 1(a) and Type 2 in Figure 3. Therefore, length similarity was incorporated as another similarity metric.

[0037] Similarly, to avoid the influence of redundant data, the length similarity is calculated using the truncated key interval, i.e., the test key interval truncated from the infrared radiation intensity time series of the target to be identified is... The key intervals of the infrared radiation intensity time series of various templates in the template library are as follows: .

[0038] In this embodiment of the invention, the method for calculating length similarity is derived from the method for calculating the intersection-over-union (IoU) between the predicted region and the ground truth box in two-dimensional image object detection. The formula for calculating IoU in two-dimensional image object detection is shown in equation (5): (5) Where A is the prediction region and B is the truth box.

[0039] Define the time series curve of infrared radiation intensity of the target to be identified. and the time series curve of infrared radiation intensity of a certain template in the template library The formula for calculating length similarity is: (6) Using the above formula (6), the length similarity between the infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of each type of template in the template library is calculated and denoted as: , L represents the number of types in the template library. test L refers to the length of the critical interval in the time series curve of the infrared radiation intensity of the target to be identified; model It is the length of the critical interval of the infrared radiation intensity time series curve of a certain template in the template library. The length similarity S in the formula... 2n It is equal to the Intersection over Union (IoU), which measures length similarity.

[0040] Optionally, in the spatial infrared point target recognition method provided in this embodiment of the invention, the spatial infrared point target is similar to shape similarity and length similarity to obtain the recognition result. Specifically, this includes: assigning similarity measurement factors to shape similarity and length similarity; calculating the similarity measurement between the infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of multiple templates respectively; and selecting the type of the template infrared radiation intensity time series corresponding to the maximum value in the similarity measurement as the recognition result of the spatial infrared point target.

[0041] The time series curve of infrared radiation intensity of the target to be identified was obtained. and the time series curve of infrared radiation intensity of a certain template in the template library The similarity measures are shape similarity and length similarity, and both similarity factors are in the range [0,1] and have the same order of magnitude. The following similarity measure function S is designed: , (7) in, This represents the weight of shape similarity in the final similarity score. For the target to be identified and the template in the template library Similarity measures of class sequences, through experimental comparison. The recognition effect is best at that time. Ultimately, the maximum value... The corresponding type k is the identification result of the spatial infrared point target to be identified.

[0042] The above analysis shows that the spatial infrared point target recognition method based on similarity measurement provided in this invention innovatively proposes a new similarity measurement method. Utilizing the characteristics of radiation intensity time series, it comprehensively considers the shape and length similarity between the radiation intensity time series of the spatial infrared point target to be identified and the infrared radiation intensity time series of various types of templates in the template library. The recognition accuracy can reach 85%, and the time required for a single recognition task is only 0.5 seconds. It achieves high recognition accuracy while eliminating the influence of distance and environment, meets timeliness requirements, has low computational resource requirements, and is easy to deploy.

[0043] Example 2 This invention provides a spatial infrared point target recognition system based on similarity measurement, the spatial infrared point target recognition system comprising: The preprocessing module is used to preprocess the infrared radiation intensity time series of the target to be identified, acquired from spatial infrared point targets, and the infrared radiation intensity time series of various templates in the template library. The purpose of preprocessing is to ensure consistency between the infrared radiation intensity time series of the target to be identified and the various template infrared radiation intensity time series in the template library used for similarity measurement, taking into account the aforementioned influencing factors, and achieving a unified data standard for subsequent similarity measurement. Preprocessing may include interpolation completion, filtering, and normalization.

[0044] The key region extraction module is used to extract key intervals from the preprocessed infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of the multiple templates, respectively. It can find the peak point of radiation intensity in the preprocessed infrared radiation intensity time series of the target to be identified; then, a radiation threshold T is set, and the effective range near the peak point of the infrared radiation intensity time series to be identified is extracted as the key interval. Similarly, the above steps are performed on the infrared radiation intensity time series of n types of templates in the template library to obtain the corresponding key intervals.

[0045] The similarity calculation module is used to calculate the shape similarity and length similarity of the infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of various templates within key intervals. Using the Dynamic Time Warping (DTW) algorithm, the shape similarity is calculated sequentially between the preprocessed infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of each type of template in the template library. Where n is the number of template types in the template library. Additionally, drawing on the concept of IoU from two-dimensional target detection, the length L of the segment extracted from the infrared radiation intensity time series of the target to be identified is calculated, along with the segment lengths of the infrared radiation intensity time series extracted from each type of template in the template library. Length similarity , where n is the number of template types in the template library.

[0046] The result recognition module is used to measure the similarity of spatial infrared point targets based on shape and length similarity, and obtain the recognition result. The similarity measurement function S is designed as follows: Where i = 1, 2, ..., n, and λ is the weight of shape similarity in the final similarity score. Experimentally, λ = 0.75 was chosen. The maximum value was obtained. The corresponding type k is the identification result of the spatial infrared point target to be identified.

[0047] Optionally, in the spatial infrared point target recognition system provided in this embodiment of the invention, the similarity calculation module is specifically used for: performing point matching based on the dynamic time warping method on the curve segment of the infrared radiation time series curve of the target to be identified in the test key interval and the curve segment of the infrared radiation time series curve of the multi-type template infrared radiation time series in the template key interval, and calculating the shortest Euclidean distance of the matching points; remapping the shortest Euclidean distance to obtain the shape similarity between the infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of the multi-type template.

[0048] Dynamic Time Warping (DTW) differs from Euclidean distance in that it measures sequence similarity based on time points. DTW matches data points in a time series by bending the time points of the time series. A schematic diagram of the matching between the two methods is shown in Figure 8. The core of the DTW method is a point matching technique based on the sequence shape (matching sequence peaks and troughs as closely as possible), and then performing traditional similarity calculations such as Euclidean distance based on these correspondences. The calculation involves the infrared radiation intensity time series curve of the target to be identified. In the interval The curve segment within and the time series curve of infrared radiation intensity of templates in the template library In the interval The shortest Euclidean distance of a curve segment within a given point matching scheme is denoted as . .

[0049] For the final similarity measurement result, the larger the value, the higher the similarity between the two sets of sequences, and the more likely they are to belong to the same type. The result obtained by the DTW algorithm is the Euclidean distance between the two sets of sequences after point matching; the larger the value, the lower the similarity in shape between the two sets of sequences, and the less likely they are to belong to the same type. Therefore, the reciprocal of the DTW calculation result is taken: 1 / .

[0050] Following the steps above, the DTW distance between the infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of various templates in the template library is calculated sequentially, and the reciprocal is taken. Finally, the result is normalized. Ultimately, the shape similarity between the infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of various templates in the template library is obtained. , This represents the number of types in the template library.

[0051] Optionally, in the spatial infrared point target recognition system provided in this embodiment of the invention, the similarity calculation module is further specifically used to: calculate the intersection-union ratio of the test key interval corresponding to the infrared radiation time series curve of the target to be identified and the template key interval corresponding to the infrared radiation time series curve of multiple templates, so as to obtain the length similarity between the infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of multiple templates.

[0052] Because some types are extremely similar in shape, and because sample data of the same type may have higher shape similarity between two groups of sequences belonging to different types due to different tasks performed and various noise interferences, using shape similarity alone as a similarity measure has drawbacks.

[0053] Observation revealed that while the lengths of the infrared radiation intensity time series within the same type varied, the differences were relatively small. However, the lengths of the infrared radiation intensity time series within different types differed relatively significantly, as shown in Type 1(a) and Type 2 in Figure 3. Therefore, length similarity was incorporated as another similarity metric.

[0054] Similarly, to avoid the influence of redundant data, the length similarity is calculated using the truncated key interval, i.e., the test key interval truncated from the infrared radiation intensity time series of the target to be identified is... The key intervals of the infrared radiation intensity time series of various templates in the template library are as follows: .

[0055] In this embodiment of the invention, the method for calculating length similarity is derived from the method for calculating the intersection-over-union (IoU) between the predicted region and the ground truth box in two-dimensional image object detection. The formula for calculating IoU in two-dimensional image object detection is shown in equation (5): (5) Where A is the prediction region and B is the truth box.

[0056] Define the time series curve of infrared radiation intensity of the target to be identified. and the time series curve of infrared radiation intensity of a certain template in the template library The formula for calculating length similarity is: (6) Using the above formula (6), the length similarity between the infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of each type of template in the template library is calculated and denoted as: , This represents the number of types in the template library.

[0057] The above analysis shows that the spatial infrared point target recognition method based on similarity measurement provided in this invention innovatively proposes a new similarity measurement method. Utilizing the characteristics of radiation intensity time series, it comprehensively considers the shape and length similarity between the radiation intensity time series of the spatial infrared point target to be identified and the infrared radiation intensity time series of various types of templates in the template library. The recognition accuracy can reach 85%, and the time required for a single recognition task is only 0.5 seconds. It achieves high recognition accuracy while eliminating the influence of distance and environment, meets timeliness requirements, has low computational resource requirements, and is easy to deploy.

[0058] Example 3 This invention provides a storage medium for computer-readable storage, wherein the storage medium stores one or more types of programs, which, when executed by one or more types of processors, implement the spatial infrared point target recognition method based on similarity metric as described above.

[0059] The above analysis shows that the spatial infrared point target recognition method based on similarity measurement provided in this invention innovatively proposes a new similarity measurement method. Utilizing the characteristics of radiation intensity time series, it comprehensively considers the shape and length similarity between the radiation intensity time series of the spatial infrared point target to be identified and the infrared radiation intensity time series of various types of templates in the template library. The recognition accuracy can reach 85%, and the time required for a single recognition task is only 0.5 seconds. It achieves high recognition accuracy while eliminating the influence of distance and environment, meets timeliness requirements, has low computational resource requirements, and is easy to deploy.

[0060] In summary, the above description is merely a preferred embodiment of this specification and is not intended to limit the scope of protection of this specification. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this specification should be included within the scope of protection of this specification.

[0061] The systems, apparatuses, modules, or units described in one or more of the above embodiments may be implemented by a computer chip or entity, or by a product having a certain function. A typical implementation device is a computer. Specifically, a computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or any combination of these devices.

[0062] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0063] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0064] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

Claims

1. A method for identifying spatial infrared point targets based on similarity measurement, characterized in that, Includes the following steps: The infrared radiation intensity time series of the target to be identified, acquired from spatial infrared point targets, and the infrared radiation intensity time series of multiple templates in the template library are preprocessed respectively. The preprocessing specifically includes: performing linear interpolation to complete the infrared radiation intensity time series of the target to be identified; filtering the infrared radiation intensity time series of the target to be identified after linear interpolation; normalizing the infrared radiation intensity time series of the target to be identified after filtering; and extracting key intervals from the infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of the multiple templates respectively. Within the key interval, the shape similarity and length similarity are calculated for the infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of the multiple templates, respectively. Determine the peak amplitude point in the infrared radiation intensity time series of the target to be identified; set amplitude thresholds on both sides of the peak amplitude point, and extract the test key interval between the amplitude thresholds; determine the template key interval corresponding to the infrared radiation intensity time series of the multiple templates; The shape similarity is passed. Based on the dynamic time warping method, the curve segment of the infrared radiation time series curve of the target to be identified corresponding to the infrared radiation intensity time series of the target to be identified in the test key interval is matched with the curve segment of the infrared radiation time series curve of the multi-type template corresponding to the infrared radiation intensity time series of the multi-type template in the template key interval. The shortest Euclidean distance of the matching points is calculated and the shortest Euclidean distance is remapped to obtain the result. The length similarity is the intersection-union ratio of the test key interval and the template key interval; The similarity of the spatial infrared point target is measured based on the shape similarity and the length similarity to obtain the recognition result; Infrared radiation intensity time series is one-dimensional data obtained by sampling and processing at set time intervals. The size of the data represents the size of the radiation value. Plotting the infrared radiation intensity time series in chronological order yields the corresponding infrared radiation intensity time series curve.

2. The spatial infrared point target identification method according to claim 1, characterized in that, The similarity of the spatial infrared point target is measured based on the shape similarity and the length similarity to obtain the recognition result, specifically including: Assign similarity measurement factors to the shape similarity and the length similarity; Calculate the similarity measure between the infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of the multiple templates; The type of the template infrared radiation intensity time series corresponding to the maximum value in the similarity metric is selected as the identification result of the spatial infrared point target.

3. A spatial infrared point target recognition system based on similarity measurement, characterized in that, include: The preprocessing module is used to preprocess the infrared radiation intensity time series of the target to be identified collected from spatial infrared point targets and the infrared radiation intensity time series of multiple templates in the template library, respectively. The key region extraction module is used to extract key intervals from the preprocessed infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of the multiple templates, respectively. Determine the peak amplitude point in the infrared radiation intensity time series of the target to be identified; set amplitude thresholds on both sides of the peak amplitude point, and extract the test key interval between the amplitude thresholds; determine the template key interval corresponding to the infrared radiation intensity time series of the multiple templates; The similarity calculation module is used to calculate the shape similarity and length similarity of the infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of the multi-type templates within the key intervals extracted from the preprocessed infrared radiation intensity time series of the target to be identified and the infrared radiation intensity time series of the multi-type templates, respectively. The shape similarity is passed. Based on the dynamic time warping method, the curve segment of the infrared radiation time series curve of the target to be identified corresponding to the infrared radiation intensity time series of the target to be identified in the test key interval is matched with the curve segment of the infrared radiation time series curve of the multi-type template corresponding to the infrared radiation intensity time series of the multi-type template in the template key interval. The shortest Euclidean distance of the matching points is calculated and the shortest Euclidean distance is remapped to obtain the result. The length similarity is the intersection-union ratio of the test key interval and the template key interval; The result recognition module is used to measure the similarity of the spatial infrared point target based on the shape similarity and the length similarity to obtain the recognition result.

4. A storage medium, characterized in that, For computer-readable storage, the storage medium stores one or more programs, which, when executed by one or more processors, implement the spatial infrared point target identification method as described in any one of claims 1 to 2.

Images