A method for individual identification of manis pentadactyla

Abstract

The application discloses a method for individual distinguishing and identifying of Manis pentadactyla. The method is characterized in that the number, distribution position, shape and size of the first to third rows of scales in the dorsal starting scale area of the head of Manis pentadactyla are measured as the first parameter index, and the length of the nose tip of Manis pentadactyla is measured as the second parameter index, so as to realize individual distinguishing and identifying of Manis pentadactyla. The method is simple and easy to operate, and is a non-contact measurement method, which can effectively and accurately distinguish and identify different Manis pentadactyla individuals and provide a reference for protection personnel.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of animal individual identification, and particularly relates to a method for individual identification of Manis pentadactyla. BACKGROUND

[0002] As a national first-class protected animal, Manis pentadactyla is listed in the Red List of the International Union for Conservation of Nature (IUCN) as critically endangered (CR) together with other seven species of pangolin. It is of great importance to understand the population size and living conditions of Manis pentadactyla in natural habitats for the protection of the species. However, due to its nocturnal behavior, small size, and biological characteristics of being covered with scales, it is difficult for protection personnel to monitor its activities in the wild. Even if the image data of the activities in the wild are captured, it is difficult to distinguish and identify different individuals. Individual identification technology has been widely used in wild animals, which can be identified according to unique biological characteristics such as hair morphology and facial features. However, as a kind of animal covered with scales, Manis pentadactyla lacks hair characteristics. At present, only the unique morphological characteristics of each species of pangolin can be used to distinguish between species, and there is no specific method for individual identification of a single species of pangolin. Therefore, there is an urgent need for a method for individual identification of Manis pentadactyla to accurately identify different individuals of Manis pentadactyla. SUMMARY

[0003] In order to accurately identify individuals of Manis pentadactyla, a method for individual identification of Manis pentadactyla is disclosed to solve the above problems.

[0004] The first object of the present application is to provide a method for individual identification of Manis pentadactyla, comprising the following steps:

[0005] S1. Obtaining a dorsal head image of Manis pentadactyla and performing pretreatment;

[0006] S2. Collecting the number of each row of scales in the starting scale area of the dorsal head of Manis pentadactyla, the symmetry of the scales on both sides of the median sagittal axis, the number of scales on the median sagittal axis, the number of scales on both sides of the median sagittal axis, and the long axis length and short axis length of the first row of scales as first parameter index data using the dorsal head image of step S1.

[0007] If the first row of scales is on the median sagittal axis, the long axis length and short axis length of the first row of scales are the long axis length and short axis length of the scales on the median sagittal axis of the first row of scales. If the first row of scales is not symmetrically distributed or there are no scales on the median sagittal axis, the long axis length and short axis length of the first row of scales are the long axis length and short axis length of the scale with the largest surface area in the first row.

[0008] S2. Collecting the data of the length of the nose tip of Manis pentadactyla as the second parameter index data; the length of the nose tip is the distance from the starting point of the nose to the nearest point of the first scale on the dorsal side of the median sagittal axis of the head;

[0009] S3. Combining the first parameter index data and the second parameter index data to construct an individual identification decision tree, so as to distinguish and identify different individuals of Manis pentadactyla.

[0010] Preferably, the individual identification decision tree comprises the following steps:

[0011] S3.1 If the number and arrangement of the 1st-3rd rows of scales of Manis pentadactyla in different dorsal images of the head are different, it is determined that the Manis pentadactyla in different dorsal images of the head are different individuals of Manis pentadactyla; if the number and arrangement of the 1st-3rd rows of scales of Manis pentadactyla in different dorsal images of the head are the same, it is further determined whether the symmetry of the scales on both sides of the median sagittal axis of the Manis pentadactyla in different dorsal images of the head is the same;

[0012] S3.2 If the symmetry of the scales on both sides of the median sagittal axis of Manis pentadactyla in different dorsal images of the head is different, it is determined that the Manis pentadactyla in different dorsal images of the head are different individuals of Manis pentadactyla; if the symmetry of the scales on both sides of the median sagittal axis of Manis pentadactyla in different dorsal images of the head is the same, it is further determined whether the number of scales on the median sagittal axis of the Manis pentadactyla in different dorsal images of the head is the same;

[0013] S3.3 If the number of scales on the median sagittal axis of Manis pentadactyla in different dorsal images of the head is different, it is determined that the Manis pentadactyla in different dorsal images of the head are different individuals of Manis pentadactyla; if the number of scales on the median sagittal axis of Manis pentadactyla in different dorsal images of the head is the same, it is further determined whether the number of scales on both sides of the median sagittal axis of the Manis pentadactyla in different dorsal images of the head is the same;

[0014] S3.4 If the number of scales on both sides of the median sagittal axis of Manis pentadactyla in different dorsal images of the head is different, it is determined that the Manis pentadactyla in different dorsal images of the head are different individuals of Manis pentadactyla; if the number of scales on both sides of the median sagittal axis of Manis pentadactyla in different dorsal images of the head is the same, it is further determined the length of the major axis and the length of the minor axis of the 1st row of scales of the Manis pentadactyla in different dorsal images of the head;

[0015] S3.5 If the long axis length and the short axis length of the first row of scales of the Chinese pangolin in different head dorsal images differ by >0.1 mm, it is determined that the Chinese pangolins in different head dorsal images are different Chinese pangolin individuals; if the long axis length and the short axis length of the first row of scales of the Chinese pangolin in different head dorsal images differ by ≤0.1 mm, it is further determined that the nose tip length of the Chinese pangolin in different head dorsal images;

[0016] S3.6 If the nose tip length of the Chinese pangolin in different head dorsal images differs by >0.5 mm, it is determined that the Chinese pangolins in different head dorsal images are different Chinese pangolin individuals.

[0017] Preferably, the preprocessing is adjusting the offset, unifying the angle and / or geometric normalization processing through affine transformation.

[0018] A second object of the present application is to provide an electronic device capable of implementing the method to realize the differentiation and identification of Chinese pangolins.

[0019] Preferably, the head dorsal image of the Chinese pangolin is extracted using a deep learning image classification model; the first to third rows of scales, the median sagittal axis and the scale regions on both sides of the median sagittal axis are divided using a deep learning segmentation model; and the first parameter index data and the second parameter index data are captured and identified by a deep learning image recognition model.

[0020] Preferably, the image classification model in deep learning includes a feature extraction network and a classification system.

[0021] Preferably, the electronic device can construct a Chinese pangolin individual feature database from the obtained head dorsal images of Chinese pangolins and / or the first parameter index and the second parameter index collected using the obtained head dorsal images of Chinese pangolins.

[0022] Preferably, when a new head dorsal image of a Chinese pangolin is obtained, the electronic device collects the first parameter index and the second parameter index using the new head dorsal image of the Chinese pangolin and enters the Chinese pangolin individual feature database, compares with the original data in the Chinese pangolin individual feature database, and identifies each Chinese pangolin individual based on the comparison result.

[0023] The present application has the following beneficial effects:

[0024] The application utilizes the more easily obtained image of the pangolin head, effectively and accurately distinguishes and identifies different Chinese pangolin individuals by finding the distribution and size of the characteristic scales on the dorsal side of the Chinese pangolin head and the length of the nose tip as a parameter index, and utilizes the dorsal side image of the Chinese pangolin head to realize the distinction and identification of Chinese pangolin individuals without contact, the method provided by the application is not affected by color errors caused by different lighting environments, and the measurement method is simple and easy to operate. BRIEF DESCRIPTION OF DRAWINGS

[0025] Figure 1 is the measurement of the nose tip length of the Chinese pangolin head and the division of the starting scale area, (A) is a schematic diagram of the long axis and short axis length of the first row of scales and the length of the nose tip, (B) is a schematic diagram of the starting scale area, and (C) is a schematic diagram of the division of the first to third rows of scales and the median sagittal axis.

[0026] Figure 2 is the image of the dorsal side starting scale area of the Chinese pangolin head numbered MP-1 to MP-18 in Example 1.

[0027] Figure 3 is the decision tree for individual identification of Chinese pangolins.

[0028] Figure 4 is the whole image of the Chinese pangolin head collected in Example 2, which is used for distinguishing and identifying different Chinese pangolin individuals.

[0029] Figure 5 is the image of the 10 Chinese pangolins numbered C1 to C10 after being distinguished and identified in Example 2. DETAILED DESCRIPTION

[0030] The following examples are further illustrations of the application, but are not limitations of the application.

[0031] The following examples are further illustrations of the application, but are not limitations of the application. Figure 1 The following examples are further illustrations of the application, but are not limitations of the application.

[0032] Example 1

[0033] Due to the small size and distinct spindle shape of the Chinese pangolin, images captured by cameras at night tend to appear grayish-white, making it difficult to distinguish the color and patterns of individual scales. To facilitate individual differentiation and identification under varying lighting conditions, it's crucial to identify the distribution and size of regionally characteristic scales, which is more easily discernible. Head images are relatively easier to obtain and possess unique biological characteristics of the species, making them suitable for individual identification under different circumstances. Furthermore, long-term observation of captive Chinese pangolins revealed that the distribution characteristics of their scales are already clear before birth. Therefore, data were collected from head images of 18 captive Chinese pangolins (numbered MP-1 to MP-18) to measure indicators (including the number of scales in each of the first to third rows of the dorsal initial scale area, the symmetry of scales on both sides of the midsagittal axis, the number of scales on and on both sides of the midsagittal axis, the long and short axis lengths of the first row of scales, and the nasal tip length). An identification method and characteristic database were established. This included 4 juveniles (age ≤ 0.5 years), 4 subadults (age 0.5–1.5 years), and 10 adults (age > 1.5 years). Detailed animal information is shown in Table 1. Because the growth changes in juveniles are relatively more pronounced than in subadults and adults, all data collection was completed within 7 days to ensure the accuracy and scientific rigor of the experiment.

[0034] 3D reconstruction was performed on CT scan images of 18 Chinese pangolins (numbered MP-1 to MP-18) using Radiant Viewer software. Dorsal and ventral images of their heads were acquired. Labelme tools were used to process the dorsal and ventral images of the pangolin heads, cropping and collecting images of the initial scale area on the dorsal side of the pangolin's head. Specifically, images showing the morphology and distribution of the first to third rows of scales were collected. Figure 1 Images of the dorsal lateral scale region of the head of Chinese pangolins MP-1 to MP-18 are shown below. Figure 2 As shown.

[0035] The images of the Chinese pangolin were preprocessed to correct the images: the offset was adjusted by affine transformation to unify the angle of all images, thereby reducing the influence of the scanning or shooting angle on the distance measurement image; geometric normalization was performed to effectively extract the morphological and distribution characteristics of the first to third rows of scales in the dorsal scale region of the head, reducing the influence of the shooting distance.

[0036] The collected head dorsal photos of Manis pentadactyla are processed by using a deep learning image classification model. The deep learning image classification model mainly consists of a feature extraction network and a classification system. The classification system identifies different individuals according to important features obtained by the feature extraction network, including the number and distribution of scales in the first to third rows of the starting scale area on the dorsal side of the head of Manis pentadactyla. Further, the first to third rows of scales in the starting scale area on the dorsal side of the head of Manis pentadactyla are segmented by using a deep learning segmentation model along the median sagittal axis to obtain corresponding scale part images. The number of scales on the median sagittal axis and the number of scales on the left and right sides of each row of the median sagittal axis are captured and identified by using a deep learning image recognition model.

[0037] Specifically, the deep learning image classification model is used to locate and divide the key scale area (i.e., the starting scale area) from the corrected image of the dorsal side of the head of Manis pentadactyla. A deep learning segmentation model is used to divide the first to third rows of scales and their left and right scale areas along the median sagittal axis. A target detection model is trained to identify and classify each scale in the image. The segmentation and detection results are integrated, and the scale counting is performed through post-processing logic. The scale area segmentation data set is divided, and the training set and the validation set are created. The segmentation model is trained, and the training set is used for model learning. The accuracy of the area division and scale counting is evaluated on the validation set. The model parameters and post-processing rules are adjusted to optimize the counting accuracy and robustness. The trained model pipeline is applied to automatically count the scales of Manis pentadactyla in new images. The counting results, including the number of scales on each side of each row and the total overview, are recorded. The automatic counting results are manually reviewed and corrected.

[0038] The Excel is used for statistics to count the number of scales in the first to third rows of the starting scale area on the dorsal side of the head, the number and distribution of scales on the median sagittal axis and its two sides, and the distribution rule. The number of scales in the first to third rows on the median sagittal axis and the number of scales on the left and right sides of each row of the first to third rows are compared to evaluate the symmetry and distribution rule of scale distribution.

[0039] According to the distance calculation formula, the Euclidean distance between the longest and shortest axes of the scales on the median sagittal axis of the first row of pangolin scales is calculated first (i.e., the length of the major and minor axes of the first row of scales). Each scale is measured five times consecutively, and the average value is taken to determine the shape characteristics of each scale in the first row, thus identifying shape differences in the first row of scales among different individuals. If the scales in the first row are not symmetrically distributed or there are no scales distributed on the median sagittal axis, then the Euclidean distance between the longest and shortest axes of the scale with the largest surface area in the first row is measured. These are collectively referred to as the length of the major and minor axes of the first row of scales. In summary, using the number of scales in each of the first to third rows of scales in the dorsal scale region of the head, the symmetry of scales on both sides of the median sagittal axis (determined by whether the number of scales on both sides of the median sagittal axis is the same), the number of scales on both sides of the median sagittal axis and on the axis, and the length of the major and minor axes of the first row of scales as image features, a feature database of scales in the initial scale region of the Chinese pangolin is established, which serves as the first parameter index for distinguishing and identifying different Chinese pangolins.

[0040] CT scan images of 18 Chinese pangolins were processed using Radiant Viewer software. The nasal tip length was calculated as the distance from the nasal origin to the nearest point of the first scale on the dorsal side of the midsagittal axis of the head. The nasal tip length of each pangolin was measured five times and the average value was taken. Data on the nasal tip length of each pangolin's head were statistically analyzed using Excel and SPSS 25 software. Significant differences were found between nasal tip length and age, sex, and weight (p<0.01). Detailed data are shown in Table 1. Nasal tip length was used as a secondary parameter for distinguishing and identifying different pangolins.

[0041] After summing the first and second parameter indicators, a method for distinguishing and identifying different pangolin individuals is finally obtained, such as... Figure 3 The individual identification decision tree shown illustrates the following steps for identifying and distinguishing Chinese pangolins based on images:

[0042] The 18 Chinese pangolins can be divided into five groups by comparing the number of scales in the first to third rows of the dorsal starting scale area of the head of the Chinese pangolin, i.e., the first group has a total of 8 scales (1, 3, and 4 in the first to third rows, respectively), the second group has a total of 9 scales (1, 3, and 5 in the first to third rows, respectively), the third group has a total of 6 scales (1, 2, and 3 in the first to third rows, respectively), the fourth group has a total of 7 scales (1, 3, and 3 in the first to third rows, respectively), and the fifth group has a total of 9 scales (3, 3, and 3 in the first to third rows, respectively). The first group includes five Chinese pangolin individuals, i.e., MP-1, MP-2, MP-3, MP-8, and MP-18; the second group includes three Chinese pangolin individuals, i.e., MP-4, MP-6, and MP-9; the third group includes three Chinese pangolin individuals, i.e., MP-5, MP-7, and MP-10; the fourth group includes six Chinese pangolin individuals, i.e., MP-11, MP-12, MP-13, MP-14, MP-15, and MP-17; and the fifth group includes one Chinese pangolin individual, i.e., MP-16. This step can distinguish the Chinese pangolin MP-16.

[0043] According to whether the scales in the first to third rows of the dorsal starting scale area of the head of the Chinese pangolin are symmetrically distributed about the median sagittal axis, the individuals in the first group can be divided into two groups, i.e., MP-1, MP-2, MP-3, and MP-8, in which the scales in the first to third rows are symmetrically distributed about the median sagittal axis (i.e., the number of scales in the first to third rows on both sides of the median sagittal axis is the same), and MP-18, in which the scales in the first to third rows are not symmetrically distributed about the median sagittal axis (i.e., the number of scales in the first to third rows on both sides of the median sagittal axis is different). The individuals in the second group can be divided into one group, i.e., MP-4, MP-6, and MP-9, in which the scales in the first to third rows are symmetrically distributed about the median sagittal axis. The individuals in the third group can be divided into one group, i.e., MP-5, MP-7, and MP-10, in which the scales in the first to third rows are not symmetrically distributed about the median sagittal axis. The individuals in the fourth group can be divided into one group, i.e., MP-11, MP-12, MP-13, MP-14, MP-15, and MP-17, in which the scales in the first to third rows are symmetrically distributed about the median sagittal axis. This step can distinguish the Chinese pangolin MP-18.

[0044] According to the number of scales on the median sagittal axis of the first to third rows of scales in the starting scale area on the dorsal side of the head of the Chinese pangolin, the remaining individuals (MP-1, MP-2, MP-3, MP-8) in the first group were divided into two groups, MP-1 (2 scales on the median sagittal axis), MP-2, MP-3, MP-8 (3 scales on the median sagittal axis). The individuals in the second group were divided into one group, MP-4, MP-6, MP-9 (3 scales on the median sagittal axis). The individuals in the third group were divided into two groups, MP-5, MP-7 (3 scales on the median sagittal axis), MP-7 (1 scale on the median sagittal axis). The individuals in the fourth group were divided into two groups, MP-11, MP-12, MP-13, MP-14, MP-15 (3 scales on the median sagittal axis), MP-17 (2 scales on the median sagittal axis). In this way, MP-1, MP-7, and MP-17 can be distinguished.

[0045] According to the number of scales on the median sagittal axis of the first to third rows of scales in the starting scale area on the dorsal side of the head of the Chinese pangolin, the remaining individuals (MP-2, MP-3, MP-8) in the first group were divided into two groups, MP-2, MP-8 (2 scales on the left side, 3 scales on the right side), MP-3 (3 scales on the left side, 2 scales on the right side). The individuals in the second group were divided into one group, MP-4, MP-6, MP-9 (3 scales on the left side, 3 scales on the right side). The remaining individuals (MP-5, MP-7) in the third group were divided into two groups, MP-5 (1 scale on the left side, 2 scales on the right side), MP-7 (2 scales on the left side, 3 scales on the right side). The remaining individuals in the fourth group were divided into one group, MP-11, MP-12, MP-13, MP-14, MP-15 (2 scales on the left side, 2 scales on the right side). In this way, MP-3, MP-5, and MP-7 can be distinguished.

[0046] According to the shape and size of the first row of scales in the dorsal head of the Chinese pangolin, the long axis length and the short axis length of the first row of scales are used as indicators for differentiation. When the difference between the long axis length and the short axis length is within the error range of ≤0.1 mm, it is considered that different individuals cannot be distinguished. The long axis length and the short axis length of the first row of scales distinguish the remaining individuals (MP-2, MP-8) in the first group into two groups, MP-2 (first scale long axis length 6.976±0.015 mm, first scale short axis length 5.406±0.011 mm), MP-8 (first scale long axis length 4.310±0.016 mm, first scale short axis length 4.154±0.011); the individuals (MP-4, MP-6, MP-9) in the second group are divided into three groups, MP-4 (first scale long axis length 6.162±0.013 mm, first scale short axis length 5.516±0.011 mm), MP-6 (first scale long axis length 7.890±0.029 mm, first scale short axis length 5.198±0.019 mm), MP-9 (first scale long axis length 10.120±0.019 mm, first scale short axis length 7.264±0.009 mm); the individuals (MP-11, MP-12, MP-13, MP-14, MP-15) in the fourth group are divided into four groups, MP-11 (first scale long axis length 12.694±0.009 mm, first scale short axis length 7.856±0.005 mm), MP-12 and MP-14 (MP-12: first scale long axis length 4.776±0.011 mm, first scale short axis length 3.974±0.009 mm; MP-14 first scale long axis length 4.462±0.013 mm, first scale short axis length 3.466±0.021 mm), MP-13 (first scale long axis length 8.438±0.016 mm, first scale short axis length 5.766±0.024 mm), MP-15 (first scale long axis length 4.450±0.022 mm, first scale short axis length 2.166±0.011 mm).

[0047] According to the measurement of the length of the Chinese pangolin nose tip, the remaining individuals in the fourth group can be successfully divided into two groups, MP-12 (nose tip length: 7.49±0.03 mm), MP-14 (nose tip length: 9.07±0.05 mm). When the difference in nose tip length is within the error range of ≤0.5 mm, it is considered to be the same individual.

[0048] In summary, according to the decision tree described above, MP-1 to MP-18 can be accurately distinguished and identified. Figure 3

[0049] Table 1 CT measurement of body data of Chinese pangolin MP-1 to MP-18 ​

[0050]

[0051] Example 2

[0052] The method constructed in Example 1 was verified by using high-definition images of the head of the pangolin taken by a camera or a mobile phone. Different from the individuals described in Example 1, the dorsal and ventral images of the head of 10 Chinese pangolins (numbered C1 to C10) were collected, a total of 96 (as shown in FIG. 2). A ruler was given to each image when taking the pictures, so as to measure and calculate the length of the tip of the nose of the pangolin later. All the collected images were collected within 7 days. All the measurement data are shown in Table 2. Figure 4

[0053] The dorsal and ventral images of the head of the Chinese pangolin were processed by using the Labelme tool, and the required dorsal and ventral images of the head were cut out. Then, the offset was adjusted by using affine transformation, the angles of all the images were unified, and finally the geometric normalization processing was performed to effectively extract the shape, distribution position and number characteristics of the first to third rows of scales in the dorsal and ventral starting scale area of the head and the length of the tip of the nose. The image classification model in deep learning was used to process the collected images to obtain the number of each row of scales in the first to third rows of scales in the dorsal and ventral starting scale area of the head, the symmetry of the scales on both sides of the median sagittal axis, the number of scales on the median sagittal axis and on both sides thereof, the long axis length and short axis length of the first row of scales, and the length of the tip of the nose, so as to identify different Chinese pangolin individuals and mark and identify different Chinese pangolin images according to the decision tree as shown in FIG. 3. The body data of the 10 Chinese pangolins (C1 to C10) are shown in Table 2. Figure 3

[0054] The specific implementation steps are as follows:

[0055] ​​First, the first to third rows of scales on the dorsal side of the head of 10 Chinese pangolins were counted. According to the total number of scales in the first to third rows and the arrangement of each row, all images can be divided into 8 groups, namely: the first group has a total of 9 scales (2, 3, and 4 in the first to third rows, respectively), the second group has a total of 7 scales (1, 3, and 3 in the first to third rows, respectively), the third group has a total of 9 scales (1, 3, and 5 in the first to third rows, respectively), the fourth group has a total of 10 scales (3, 3, and 4 in the first to third rows, respectively), the fifth group has a total of 8 scales (1, 3, and 4 in the first to third rows, respectively), the sixth group has a total of 11 scales (2, 3, and 6 in the first to third rows, respectively), the seventh group has a total of 6 scales (1, 2, and 3 in the first to third rows, respectively), and the eighth group has a total of 5 scales (1, 1, and 3 in the first to third rows, respectively). According to the total number of scales in the first to third rows and the arrangement of each row in each group, it can be determined that the first group includes C1 and C3, the second group includes C2 and C9, C4 is in the third group, C5 is in the fourth group, C6 is in the fifth group, C7 is in the sixth group, C8 is in the seventh group, and C10 is in the eighth group.

[0056] Further, by measuring whether the scales on the left and right sides of the first to third rows in the first and second groups are symmetrical about the median sagittal axis and the number of scales distributed on the median sagittal axis, C1 and C3 in the first group can be distinguished, and C2 and C9 in the second group cannot be distinguished.

[0057] The length of the nose tip, the long axis length and the short axis length of the first scale, and the length of the nose tip can be considered as the same individual within the measurement error (the difference between the long axis length and the short axis length of the first scale is ≤0.1 mm, and the difference between the lengths of the nose tips is ≤0.5 mm). The data of the measured individuals are shown in Table 2. According to the length of the nose tip, the long axis length and the short axis length of the first scale, C2 and C9 in the second group can be accurately distinguished.

[0058] Table 2 Measurement of body data of C1 to C10 Chinese pangolins in the photographed images

[0059]

[0060] In summary, according to Figure 3 The decision tree shown in Fig. 6 can accurately distinguish and identify C1 to C10 Chinese pangolins. The images of the distinguished individuals are shown in Figure 5 Fig. 7.

Claims

1. A method for distinguishing and identifying Manis pentadactyla, characterized in that, The method comprises the following steps: S1. Obtain the dorsal head image of the Chinese pangolin and perform pretreatment; S2. Collect the number of scales in each row of the first to third rows of the starting scale area on the dorsal head of the Chinese pangolin, the symmetry of the scales on both sides of the median sagittal axis, the number of scales on the median sagittal axis, the number of scales on both sides of the median sagittal axis, and the long axis length and short axis length of the scales in the first row of scales as the first parameter index data by using the dorsal head image in step S1; If the scales in the first row are on the median sagittal axis, the long axis length and short axis length of the scales in the first row are the long axis length and short axis length of the scales on the median sagittal axis in the first row; if the scales in the first row are not symmetrically distributed or there are no scales on the median sagittal axis, the long axis length and short axis length of the scales in the first row are the long axis length and short axis length of the scales with the largest surface area in the first row; S2. Collect the data of the length of the nose tip of the Chinese pangolin as the second parameter index data; the length of the nose tip is the distance from the starting point of the nose to the closest point of the first scale on the dorsal head median sagittal axis; S3. Combine the first parameter index data and the second parameter index data to construct an individual identification decision tree, so as to distinguish and identify different individuals of the Chinese pangolin; The individual identification decision tree comprises the following steps: S3.1 If the number and arrangement of the scales in each row of the first to third rows of the Chinese pangolin in different dorsal head images are different, it is determined that the Chinese pangolins in different dorsal head images are different Chinese pangolin individuals; if the number and arrangement of the scales in each row of the first to third rows of the Chinese pangolin in different dorsal head images are completely the same, it is further determined whether the symmetry of the scales on both sides of the median sagittal axis of the Chinese pangolins in different dorsal head images is the same; S3.2 If the symmetry of the scales on both sides of the median sagittal axis of the Chinese pangolins in different dorsal head images is different, it is determined that the Chinese pangolins in different dorsal head images are different Chinese pangolin individuals; if the symmetry of the scales on both sides of the median sagittal axis of the Chinese pangolins in different dorsal head images is the same, it is further determined whether the number of scales on the median sagittal axis of the Chinese pangolins in different dorsal head images is the same; S3.3 If the number of scales on the median sagittal axis of the Chinese pangolins in different dorsal head images is different, it is determined that the Chinese pangolins in different dorsal head images are different Chinese pangolin individuals; if the number of scales on the median sagittal axis of the Chinese pangolins in different dorsal head images is the same, it is further determined whether the number of scales on both sides of the median sagittal axis of the Chinese pangolins in different dorsal head images is the same; S3.4 If the number of scales on both sides of the median sagittal axis of the Chinese pangolins in different dorsal head images is different, it is determined that the Chinese pangolins in different dorsal head images are different Chinese pangolin individuals; if the number of scales on both sides of the median sagittal axis of the Chinese pangolins in different dorsal head images is the same, it is further determined the long axis length and short axis length of the scales in the first row of the Chinese pangolins in different dorsal head images. S3.5 If the length of the long axis and the length of the short axis of the first row of scales of the Chinese pangolin in different head dorsal images differ by >0.1 mm, it is determined that the Chinese pangolins in different head dorsal images are different Chinese pangolin individuals; if the length of the long axis and the length of the short axis of the first row of scales of the Chinese pangolin in different head dorsal images differ by ≤0.1 mm, it is further determined that the length of the nose tip of the Chinese pangolin in different head dorsal images; S3.6 If the length of the nose tip of the Chinese pangolin in different head dorsal images differs by >0.5 mm, it is determined that the Chinese pangolins in different head dorsal images are different Chinese pangolin individuals.

2. The method of claim 1, wherein, The preprocessing is: adjusting the offset, unifying the angle and / or geometric normalization processing through affine transformation.

3. An electronic device, comprising: The method of any one of claims 1-2 can be implemented to achieve the differentiation and identification of Chinese pangolins.

4. The electronic device of claim 3, wherein, The head dorsal image of the Chinese pangolin is extracted using a deep learning image classification model; the first to third rows of scales, the median sagittal axis and the scale regions on both sides of the median sagittal axis are divided using a deep learning segmentation model; and the first parameter index data and the second parameter index data are captured and identified by a deep learning image recognition model.

5. The electronic device of claim 4, wherein, The image classification model in deep learning includes a feature extraction network and a classification system.

6. The electronic device of claim 3, wherein, The head dorsal image of the Chinese pangolin and / or the first parameter index and the second parameter index collected using the head dorsal image of the Chinese pangolin can be constructed into a Chinese pangolin individual feature database.

7. The electronic device of claim 3, wherein, When a new head dorsal image of a Chinese pangolin is obtained, the electronic device collects the first parameter index and the second parameter index using the new head dorsal image of the Chinese pangolin and enters them into the Chinese pangolin individual feature database, compares them with the original data in the Chinese pangolin individual feature database, and identifies each Chinese pangolin individual based on the comparison result.

Images