Wild animal body size and age group estimation method and system
By simultaneously acquiring visible light and thermal infrared images and combining them with a lightweight regression network, the problem of equipment complexity and high power consumption for all-weather monitoring of rare wild animals has been solved, enabling accurate estimation of size and age range, and supporting unattended deployment.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-26
- Publication Date
- 2026-04-10
AI Technical Summary
Existing technologies cannot achieve long-term, safe, and all-weather monitoring of rare wild animals. In particular, it is difficult to obtain accurate body size and age information at night and in forest cover. Moreover, the equipment is complex, consumes a lot of power, and is difficult to deploy in the wild for a long time.
By simultaneously acquiring visible light images and thermal infrared temperature maps, the optimal scale factor is determined through instance segmentation and attitude verification. Combined with a lightweight regression network, volume, weight, and age group are estimated. The system is designed to be lightweight and supports calibration-free deployment.
It achieves all-weather monitoring in uncalibrated scenarios, with a volume estimation error of less than 6.2% and an age group discrimination accuracy of up to 89.4%. The system is lightweight, low-power, and supports long-term operation powered by solar energy.
Smart Images
Figure CN121838210A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The application belongs to the technical field of unattended monitoring of wild animals, and particularly relates to a wild animal body size and age estimation method and system. BACKGROUND
[0002] In the practice of wild animal protection, ecological research and species management, it is of great significance to obtain wild animal body size parameters (such as volume and weight) and age information for a long time, automatically and accurately. Traditional methods and existing technologies mainly rely on manual contact measurement or single modal visual sensing means, which have many limitations and cannot meet the long-term, stable and safe monitoring needs in complex outdoor environments.
[0003] Direct contact measurement method can obtain accurate body size and age information by measuring wild animals manually after anesthesia or direct capture, but it is invasive and high-risk to animals, which can easily cause stress reaction, injury or even death, especially for rare and sensitive species. It cannot realize continuous and long-term monitoring, is difficult to form time-series physiological data, and is not suitable for large-scale and distributed monitoring scenarios due to its complex operation and high cost.
[0004] The visual measurement method based on visible light image relies on natural light during the day, and obtains body size related information based on image analysis by shooting visible light images of wild animals. However, this method is invalid at night or in shaded scenes, cannot realize all-weather monitoring, and is greatly affected by environmental light, with poor measurement stability.
[0005] The three-dimensional reconstruction method based on binocular vision or depth camera adopts binocular camera or depth camera to establish a stereo vision system, and restores the body size parameters of wild animals by calculating the depth information of the image. However, this method also has the problems of complex device structure, high power consumption, and the need for a large amount of power supply for long-term deployment in the field, which limits its practicality. SUMMARY
[0006] The purpose of the present application is to solve the problems of the existing wild animal monitoring technology in scene adaptability, functional integrity, device deployment difficulty and other aspects, which cannot meet the needs of long-term, safe and all-weather monitoring of rare wild animals. A wild animal body size and age estimation method and system are proposed.
[0007] The technical solution of the present application is as follows: in the first aspect, a wild animal body size and age estimation method comprises the following steps: synchronously collecting visible light images and thermal infrared temperature images of wild animals; performing instance segmentation on the visible light images to obtain binary masks of wild animals, and extracting pixel-level body size parameters; extracting temperature distribution vectors from the thermal infrared temperature images; verify the posture of the wild animal based on the binary mask, the pixel-level body size parameter, and the temperature distribution vector, and determine an optimal scale factor from pixels to physical space; correct the pixel-level body size parameter using the optimal scale factor, and fuse the corrected pixel-level body size parameter and the temperature distribution vector to obtain a multi-dimensional feature vector; input the multi-dimensional feature vector into a double-branch lightweight regression network, and synchronously output a volume proxy value, a weight estimation value, and an age classification result of the wild animal.
[0008] Preferably, the pixel-level body size parameter includes a pixel area, a pixel body length projection, and a pixel shoulder height; and the temperature distribution vector includes a maximum temperature, a pixel area with a temperature greater than or equal to a preset temperature, and a temperature gradient mean value.
[0009] Preferably, the method for verifying the posture of the wild animal based on the binary mask, the pixel-level body size parameter, and the temperature distribution vector, and determining the optimal scale factor from pixels to physical space specifically comprises: fitting a ground plane straight line according to a lowest point of the binary mask; querying a species shoulder height prior table to obtain a candidate shoulder height set; calculating a ratio of each shoulder height prior value in the candidate shoulder height set to the pixel shoulder height to obtain a candidate scale factor corresponding to each shoulder height prior value; cutting the thermal infrared temperature map according to the binary mask, calculating a ratio of a height direction coordinate of a center of mass of a wild animal thermal infrared region based on the binary mask cutting in an image coordinate system to the pixel shoulder height to obtain a center of mass height ratio; matching the center of mass height ratio with a standard center of mass height ratio under different postures of the wild animal, and selecting a candidate scale factor with the highest matching degree as the optimal scale factor.
[0010] Preferably, the multi-dimensional feature vector includes a pixel body length projection corrected based on the optimal scale factor, a pixel shoulder height corrected based on the optimal scale factor, a pixel area corrected based on the optimal scale factor, a maximum temperature, a pixel area ratio with a temperature greater than or equal to a preset temperature, and a temperature gradient mean value. Preferably, the double-branch lightweight regression network includes a first branch, a second branch, and a third branch; the first branch includes a first fully connected layer, a second fully connected layer, and a third fully connected layer connected in sequence, and is configured to calculate a volume proxy value according to the pixel body length projection corrected based on the optimal scale factor and the pixel shoulder height corrected based on the optimal scale factor; the second branch includes a fourth fully connected layer and a fifth fully connected layer connected in sequence, and is configured to calculate a weight estimation value according to the volume proxy value and the maximum temperature; The third branch comprises a feature extraction layer and a Softmax layer connected in sequence, and is used for age stage classification according to the maximum temperature, the pixel area proportion with a temperature greater than or equal to a preset temperature, and the mean value of the temperature gradient.
[0011] Preferably, the calculation formula of the volume proxy value is: V = π (k H_p / 2)² (k L_p) wherein V represents the volume proxy value, k H_p represents the pixel shoulder height corrected based on the optimal scale factor k, pi represents the circular constant, k L_p represents the pixel body length projection corrected based on the optimal scale factor k. The calculation formula of the weight estimation value is: W = rho V sigma(T_max) wherein W represents the weight estimation value, rho represents the average density of the species, sigma represents the temperature correction coefficient, and T_max represents the maximum temperature.
[0012] Preferably, the basis for age stage classification is specifically: if the maximum temperature is higher than T_adult_threshold and the pixel area proportion with a temperature greater than or equal to a preset temperature is less than 0.3, then the adult is determined; if the maximum temperature is lower than T_juv_threshold and the pixel area proportion with a temperature greater than or equal to a preset temperature is greater than 0.5, then the juvenile is determined; if the maximum temperature is less than or equal to T_adult_threshold and greater than or equal to T_juv_threshold, and the pixel area proportion with a temperature greater than or equal to a preset temperature is greater than or equal to 0.3 and less than or equal to 0.5, then the sub-adult is determined; wherein T_adult_threshold represents the adult core body temperature threshold of the target species, G_threshold represents the temperature gradient threshold corresponding to the hair thickness of the adult of the target species, and T_juv_threshold represents the juvenile core body temperature threshold of the target species.
[0013] The beneficial effects of the present application are: 1. Calibration adaptability: without calibration board, scale restoration is realized through ground plane assumption + species shoulder height priori + thermal infrared posture verification, and the method is suitable for field non-calibration scenes.
[0014] 2. All-weather monitoring: The thermal infrared unit dominates data acquisition, and can still work effectively at night and in obstructed scenes (obstruction rate ≤ 35%), solving the scene limitations of traditional visible light technology.
[0015] 3. Functional integrity: Synchronous estimation of volume, weight and age stage is realized for the first time, filling the functional gap of existing technology, with an age stage discrimination accuracy of ≥ 89.4% and an average weight error of ≤ 6.2%.
[0016] 4. Lightweight deployment: The total weight of the system is < 600g, and the power consumption is < 5W, supporting solar power supply, solving the deployment problems of existing devices with complex and high power consumption, and enabling long-term unattended operation.
[0017] 5. Strong robustness: Thermal infrared can penetrate shrub gaps, and can still maintain high measurement accuracy in obstructed scenes, significantly better than pure visual methods.
[0018] In a second aspect, a wild animal body size and age stage estimation system comprises: a monocular visible light unit for collecting visible light images of wild animals; a thermal infrared unit for collecting thermal infrared temperature maps of wild animals; an edge computing node connected to the monocular visible light unit and the thermal infrared unit, for instance segmentation of the visible light images to obtain binary masks of wild animals, and extraction of pixel-level body size parameters; extraction of temperature distribution vectors from the thermal infrared temperature maps, and running of a double-branch lightweight regression network; a non-calibrated scale restoration module integrated in the edge computing node, for verifying the posture of the wild animal based on the binary mask, the pixel-level body size parameters and the temperature distribution vector, and determining the scale factor from pixels to physical space; a fusion regression module integrated in the edge computing node, for correcting the pixel-level body size parameters using the optimal scale factor, fusing the corrected pixel-level body size parameters and the temperature distribution vector to obtain a multi-dimensional feature vector, and inputting the multi-dimensional feature vector into the double-branch lightweight regression network to simultaneously output the volume proxy value, the weight estimation value and the age stage classification result of the wild animal; a result uploading module connected to the output end of the fusion regression module, for uploading the volume proxy value, the weight estimation value and the age stage classification result of the wild animal to a remote server through wireless communication technology.
[0019] As a preferred embodiment, the monocular visible light unit and the thermal infrared unit have overlapping fields of view and synchronous data acquisition, and the main optical axes of the monocular visible light unit and the thermal infrared unit are parallel and fixedly connected. BRIEF DESCRIPTION OF DRAWINGS
[0020] Figure 1 A flowchart of a wild animal body size and age stage estimation method is shown. Detailed Implementation
[0021] Exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be understood that the embodiments shown and described in the drawings are merely exemplary and are intended to illustrate the principles and spirit of the invention, and are not intended to limit the scope of the invention.
[0022] Example 1: like Figure 1 As shown, a method for estimating the body size and age group of wild animals includes the following steps: S1. Simultaneously acquire visible light images and thermal infrared temperature maps of wild animals; S2. Perform instance segmentation on the visible light image to obtain a binary mask of the wild animal and extract pixel-level volume scale parameters; S3. Extract the temperature distribution vector from the thermal infrared temperature map; S4. Verify the wild animal posture based on binary mask, pixel-level volumetric parameters and temperature distribution vector, and determine the optimal scale factor from pixel to physical space; S5. The pixel-level volumetric parameters are corrected using the optimal scaling factor, and the corrected pixel-level volumetric parameters and the temperature distribution vector are fused to obtain a multidimensional feature vector. S6. Input the multidimensional feature vector into the dual-branch lightweight regression network and simultaneously output the volume proxy value, weight estimate, and age group classification results of the wild animals.
[0023] In this embodiment, the pixel-level body size parameters include pixel area A_p, pixel body length projection L_p, and pixel shoulder height H_p; the temperature distribution vector includes the highest temperature T_max (reflecting core body temperature), the pixel area A_hot with temperature ≥ T_max-2℃, and the mean temperature gradient G (reflecting hair thickness).
[0024] In this embodiment, the method for determining the pixel-to-physical-space scale factor based on binary masks, pixel-level volumetric parameters, and temperature distribution vectors to verify the posture of wild animals is as follows: Fit a straight line to the ground plane based on the lowest point of the binary mask; Query the species shoulder height prior table (e.g., 1.2m for adult takin and 0.7m for juveniles) to obtain the candidate shoulder height set {h_i}; Calculate the candidate scale factor k_i=h_i / H_p corresponding to each prior value of shoulder height in the candidate shoulder height set, where H_p is the pixel shoulder height; Based on the binary mask-cropped thermal infrared temperature map, the centroid height ratio r = y_cog / H_p is calculated; where y_cog represents the height coordinate of the centroid of the wild animal thermal infrared region after binary mask cropping in the image coordinate system. The centroid height ratio r is matched with the species standard pose template r_template, and the candidate scale factor with the highest matching degree is selected as the optimal scale factor.
[0025] The binary mask is the outline of the wildlife region obtained by segmenting visible light image instances (only pixels containing wildlife are foreground, and background is 0). After cropping the thermal infrared temperature map with this mask, the effective thermal infrared region corresponding to the wildlife can be accurately extracted. The centroid (cog=centroid) refers to the geometric center of the cropped thermal infrared region, reflecting the core position of the region in the image. The y-axis in the image coordinate system corresponds to the height direction (whether the y-axis is upward or downward, it represents the "height dimension perpendicular to the ground plane"), and y_cog is the specific coordinate value of the geometric center in the height direction.
[0026] The relationship between the centroid height ratio (r=y_cog / H_p) and the scale factor: The two are directly related through "pose verification". The core logic is that the centroid height ratio is used to select the optimal scale factor that fits the current animal pose. H_p is the "pixel shoulder height" extracted from the visible light image (one of the pixel-level body size parameters, that is, the number of pixels of the wild animal's shoulder height in the image); r is the ratio of centroid height to pixel shoulder height, reflecting the pose characteristics of the wild animal in the image (such as standing, bending over and other poses will cause changes in the ratio of centroid height to shoulder height). First, a candidate shoulder height set is obtained through the "species shoulder height prior table" (e.g., 1.2m for adult takin and 0.7m for juveniles). Then, the "candidate shoulder height prior value / pixel shoulder height H_p" is calculated to generate multiple candidate scale factors (k_i = h_i / H_p, where h_i is the candidate shoulder height). The centroid height ratio r = y_cog / H_p is calculated, and r is matched with the "species standard pose template" (a preset standard centroid height ratio r_template for different poses of the species). The candidate scale factor corresponding to the candidate pose with the highest matching degree is the "optimal scale factor".
[0027] Key findings: Candidate scale factors are preliminary results calculated based on "shoulder height prior + pixel shoulder height", while the centroid height ratio r is a key indicator for verifying the actual posture of the animal. The matching result of r directly filters out the optimal scale factor that "fits the current posture". The two form a correlation through "posture verification" of "calculating candidate factors → posture matching and screening → determining the optimal factor". The centroid height ratio is the "posture verification basis" for the optimal scale factor.
[0028] In this embodiment, the multidimensional feature vector includes the pixel volume projection corrected based on the optimal scale factor, the pixel shoulder height corrected based on the optimal scale factor, the pixel area corrected based on the optimal scale factor, the highest temperature, the quotient of the pixel area A_hot at temperature ≥ T_max-2℃ and the pixel area A_p, and the mean temperature gradient.
[0029] In this embodiment, the dual-branch lightweight regression network includes a first branch (branch α), a second branch (branch β), and a third branch (branch γ). The first branch includes a first fully connected layer, a second fully connected layer, and a third fully connected layer connected in sequence, used to calculate the volume proxy value based on the corrected pixel volume length projection and pixel shoulder height; The second branch includes a fourth fully connected layer and a fifth fully connected layer connected in sequence, which are used to calculate the weight estimate based on the volume proxy value and the highest temperature. The third branch consists of a feature extraction layer and a softmax layer connected in sequence, used to classify age groups based on temperature distribution vectors and pixel areas.
[0030] In this embodiment, the formula for calculating the volume surrogate value is: V=π (k H_p / 2)² (k L_p) Where V represents the volume surrogate value, k represents the optimal scale factor, π represents pi, H_p represents the pixel shoulder height, and L_p represents the pixel volume length projection. The formula for calculating the estimated weight is: W=ρ V σ(T_max) Where W represents the estimated body weight, ρ represents the average species density, σ represents the temperature correction factor (σ>1 for juveniles, σ≈1 for adults), and T_max represents the maximum temperature; In this embodiment, the criteria for age group classification are as follows: If the highest temperature is less than or equal to T_adult_threshold and greater than or equal to T_juv_threshold, and the percentage of pixel area with temperature greater than or equal to the preset temperature is greater than or equal to 0.3 and less than or equal to 0.5, then it is determined to be a sub-adult; where T_adult_threshold represents the core body temperature threshold of the adult of the target species, G_threshold represents the temperature gradient threshold corresponding to the hair thickness of the adult of the target species, and T_juv_threshold represents the core body temperature threshold of the juvenile of the target species.
[0031] The volume proxy value, weight estimate, age group classification results, GPS, and timestamp output by the dual-branch lightweight regression network are packaged into JSON format and uploaded to the cloud via LoRa technology to complete the data output.
[0032] In this embodiment of the invention, the monocular visible light unit can be replaced with a low-power high-definition camera without changing the requirements for field-of-view overlap and synchronous triggering, thus ensuring the accuracy of geometric feature extraction. LoRa can be replaced with NB-IoT, which is suitable for outdoor areas with good base station coverage, and must ensure low power consumption and data transmission stability. The dual-branch lightweight network can be replaced with a CNN+Transformer hybrid architecture, while maintaining its lightweight characteristics to ensure that edge computing nodes can handle it.
[0033] This invention solves the problem that existing technologies rely on calibration boards and cannot achieve physical scale restoration in uncalibrated scenarios; it solves the problem of monitoring failure in complex environments such as nighttime and forest cover, enabling all-weather monitoring; it addresses the functional gap in existing technologies that cannot simultaneously acquire information on the size, weight, and age of wild animals; and it solves the problem that existing equipment is complex, consumes a lot of power, and is difficult to deploy unattended in the field for long periods of time.
[0034] Example 2: Based on Example 1, this embodiment of the invention provides a system for estimating the body size and age group of wild animals, which can be used to implement the method for estimating the body size and age group of wild animals as described in the foregoing embodiments. The system includes: A monocular visible light unit is used to acquire visible light images of wildlife; Thermal infrared unit, used to collect thermal infrared temperature maps of wild animals; Edge computing nodes, connected to monocular visible light units and thermal infrared units, are used to perform instance segmentation on visible light images, obtain binary masks of wild animals, and extract pixel-level body size parameters; temperature distribution vectors are extracted from thermal infrared temperature maps, and a dual-branch lightweight regression network is run. The uncalibrated scale restoration module, integrated into the edge computing node, is used to verify the posture of wild animals based on binary masks, pixel-level body size parameters, and temperature distribution vectors, and to determine the scale factor from pixel to physical space. The uncalibrated scale restoration module does not require a calibration board and realizes the conversion between pixel and physical scale based on the ground plane assumption, species shoulder height prior, and thermal infrared posture verification.
[0035] The fusion regression module, integrated into the edge computing node, is used to correct pixel-level body size parameters using the optimal scale factor. The corrected pixel-level body size parameters and temperature distribution vector are fused to obtain a multidimensional feature vector, which is then input into a dual-branch lightweight regression network. Simultaneously, the volume proxy value, weight estimate, and age group classification results of wild animals are output. The results upload module, connected to the output of the fusion regression module, is used to transmit the volume proxy value, weight estimate, and age group classification results of wild animals to a remote server via wireless communication technology using low-power LoRa technology, thereby realizing remote data transmission.
[0036] The monocular visible light unit and thermal infrared unit coupling design in this invention, with dual-unit synchronous triggering and overlapping field of view layout, achieves precise matching of geometric features and temperature features.
[0037] The wildlife body size and age estimation system weighs less than 600g, consumes less than 5W, supports solar power, and can be directly deployed at unattended monitoring nodes in the wild.
[0038] According to embodiments of the present invention, the present invention also provides an electronic device, a readable storage medium, and a computer program product.
[0039] In an exemplary embodiment, the electronic device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the wildlife body size and age group estimation method as described in Embodiment 1 above.
[0040] In an exemplary embodiment, the readable storage medium may be a non-transitory computer-readable storage medium storing computer instructions for causing a computer to execute the method for estimating the body size and age group of wild animals according to Embodiment 1 above.
[0041] In an exemplary embodiment, the computer program product includes a computer program that, when executed by a processor, implements the method for estimating the body size and age group of wild animals according to Embodiment 1 above.
[0042] The program code used to implement the methods of the present invention can be written in any combination of one or more programming languages. This program code can be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code can be executed entirely on the machine, partially on the machine, as a standalone software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.
[0043] In the context of this invention, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.
[0044] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).
[0045] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with embodiments of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.
[0046] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.
[0047] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of this invention.
Claims
1. A method for estimating the body size and age group of wild animals, characterized in that, Includes the following steps: Simultaneously acquire visible light images and thermal infrared temperature maps of wild animals; Instance segmentation is performed on visible light images to obtain binary masks of wild animals, and pixel-level volume scale parameters are extracted. Extract the temperature distribution vector from the thermal infrared temperature map; Based on binary masks, pixel-level volumetric parameters, and temperature distribution vectors, the posture of wild animals is verified, and the optimal scale factor from pixel to physical space is determined. The pixel-level volumetric parameters are corrected using the optimal scaling factor, and the corrected pixel-level volumetric parameters and the temperature distribution vector are fused to obtain a multidimensional feature vector. The multidimensional feature vector is input into a two-branch lightweight regression network, which simultaneously outputs the volume proxy value, weight estimate, and age group classification results of wild animals.
2. The method for estimating the body size and age group of wild animals according to claim 1, characterized in that, The pixel-level body size parameters include pixel area, pixel body length projection, and pixel shoulder height; the temperature distribution vector includes the highest temperature, the pixel area with a temperature greater than or equal to the preset temperature, and the average temperature gradient.
3. The method for estimating the body size and age group of wild animals according to claim 1, characterized in that, The method for verifying wildlife poses and determining the optimal scale factor from pixel to physical space based on binary masks, pixel-level volumetric parameters, and temperature distribution vectors is as follows: Fit a straight line to the ground plane based on the lowest point of the binary mask; Query the species shoulder height prior table to obtain the candidate shoulder height set; Calculate the ratio of each prior shoulder height value in the candidate shoulder height set to the pixel shoulder height to obtain the candidate scale factor corresponding to each prior shoulder height value; Based on the binary mask cropped thermal infrared temperature map, the ratio of the height direction coordinate of the centroid of the wild animal thermal infrared region after binary mask cropping to the pixel shoulder height is calculated to obtain the centroid height ratio. The centroid height ratio is matched with the preset standard centroid height ratios for different postures of wild animals, and the candidate scale factor with the highest matching degree is selected as the optimal scale factor.
4. The method for estimating the body size and age group of wild animals according to claim 1, characterized in that, The multidimensional feature vector includes the pixel volume projection corrected by the optimal scale factor, the pixel shoulder height corrected by the optimal scale factor, the pixel area corrected by the optimal scale factor, the highest temperature, the percentage of pixel areas with a temperature greater than or equal to a preset temperature, and the mean temperature gradient.
5. The method for estimating the body size and age group of wild animals according to claim 4, characterized in that, The two-branch lightweight regression network consists of a first branch, a second branch, and a third branch; The first branch includes a first fully connected layer, a second fully connected layer, and a third fully connected layer connected in sequence, used to calculate the volume proxy value based on the pixel volume length projection corrected based on the optimal scale factor and the pixel shoulder height corrected based on the optimal scale factor. The second branch includes a fourth fully connected layer and a fifth fully connected layer connected in sequence, which are used to calculate the weight estimate based on the volume proxy value and the highest temperature. The third branch includes a feature extraction layer and a softmax layer connected in sequence, which are used to classify age groups based on the highest temperature, the percentage of pixel areas with a temperature greater than or equal to a preset temperature, and the average temperature gradient.
6. The method for estimating the body size and age group of wild animals according to claim 5, characterized in that, In this embodiment, the formula for calculating the volume surrogate value is: V=π (k H_p / 2)² (k L_p) Where V represents the volume surrogate value, k H_p represents the pixel shoulder height after correction based on the optimal scale factor k, π represents pi, and k L_p represents the pixel volume length projection after correction based on the optimal scale factor k; The formula for calculating the estimated weight is: W=ρ V σ(T_max) Where W represents the estimated body weight, ρ represents the average species density, σ represents the temperature correction factor, and T_max represents the maximum temperature.
7. The method for estimating the body size and age group of wild animals according to claim 1, characterized in that, The specific criteria for classifying age groups are as follows: If the highest temperature is higher than T_adult_threshold and the percentage of pixel area with a temperature greater than or equal to the preset temperature is less than 0.3, then it is determined to be an adult object; If the pixel area with the highest temperature is lower than T_juv_threshold and the temperature is greater than or equal to the preset temperature, the pixel is identified as a juvenile. If the highest temperature is less than or equal to T_adult_threshold and greater than or equal to T_juv_threshold, and the percentage of pixel area with temperature greater than or equal to the preset temperature is greater than or equal to 0.3 and less than or equal to 0.5, then it is determined to be a sub-adult; where T_adult_threshold represents the core body temperature threshold of the adult of the target species, G_threshold represents the temperature gradient threshold corresponding to the hair thickness of the adult of the target species, and T_juv_threshold represents the core body temperature threshold of the juvenile of the target species.
8. A system for estimating the body size and age of wild animals for implementing the method for estimating the body size and age group of wild animals as described in any one of claims 1-7, characterized in that, include: A monocular visible light unit is used to acquire visible light images of wildlife; Thermal infrared unit, used to collect thermal infrared temperature maps of wild animals; Edge computing nodes, connected to monocular visible light units and thermal infrared units, are used to perform instance segmentation on visible light images, obtain binary masks of wild animals, and extract pixel-level body size parameters; temperature distribution vectors are extracted from thermal infrared temperature maps, and a dual-branch lightweight regression network is run. The uncalibrated scale restoration module, integrated into the edge computing node, is used to verify the posture of wild animals based on binary masks, pixel-level volumetric parameters, and temperature distribution vectors, and to determine the scale factor from pixel to physical space. The fusion regression module, integrated into the edge computing node, is used to correct pixel-level body size parameters using the optimal scale factor. The corrected pixel-level body size parameters and temperature distribution vector are fused to obtain a multidimensional feature vector, which is then input into a dual-branch lightweight regression network. Simultaneously, the volume proxy value, weight estimate, and age group classification results of wild animals are output. The results upload module, connected to the output of the fusion regression module, is used to transmit the volume proxy value, weight estimate, and age group classification results of wild animals to a remote server via wireless communication technology.
9. The method for estimating the body size and age group of wild animals according to claim 8, characterized in that, The monocular visible light unit and the thermal infrared unit maintain overlapping fields of view and synchronized data acquisition, and the main optical axes of the monocular visible light unit and the thermal infrared unit are parallel and fixed.