Method and system for habitat suitability evaluation of qian gold monkey ecological corridor based on credibility fusion

Abstract

This invention discloses a method and system for evaluating the habitat suitability of Guizhou golden snub-nosed monkey ecological corridors based on credibility fusion, relating to the field of animal population monitoring and habitat evaluation technology. The method includes: acquiring multi-source occurrence data of Guizhou golden snub-nosed monkeys within the ecological corridor; determining the credibility weight and time-decrease factor of each occurrence point and determining equivalent weights based on the consistency of attributes between observation records from different sources within the same spatiotemporal neighborhood; classifying occurrence points into confirmed occurrence points and uncertain observation points; constructing similarity constraints between uncertain and confirmed occurrence points based on environmental variable similarity; using equivalent weights to weight the objective function of the maximum entropy model; introducing the constraints into the objective function in a regularized form; iteratively solving the habitat suitability index distribution; and evaluating the habitat utilization of each corridor area by combining population size, behavior type, and activity intensity data obtained from monitoring, thereby providing technical support for habitat restoration and food source allocation optimization in the ecological corridor.

Description

Technical Field

[0001] The embodiments of this invention relate to the field of animal population monitoring and habitat assessment technology, and in particular to a method and system for assessing the habitat suitability of the Guizhou golden monkey ecological corridor based on credibility fusion. Background Technology

[0002] The maximum entropy model is one of the mainstream methods for assessing species habitat suitability. Its basic principle is to calculate the probability distribution that maximizes information entropy based on species occurrence point data and environmental variables (vegetation, topography, disturbance factors, etc.) of the study area, while satisfying observational constraints, thereby predicting the potential suitable habitat distribution of the species within the study area. However, when this type of method is directly applied to the monitoring and assessment of the Guizhou golden monkey ecological corridor construction area, the following problems exist:

[0003] The reliability differences of multi-source survey data were not considered. Guizhou golden monkey surveys typically use interview surveys, transect surveys, and infrared camera monitoring to obtain sighting points. These three methods differ significantly in terms of positioning accuracy and sighting authenticity. For example, interview surveys rely on eyewitness recollections, which are highly subjective and lack direct evidence. Transect surveys are affected by observer experience and environmental obstruction, and sighting records may contain species misidentifications. Although infrared camera monitoring can provide image evidence, it is limited by the density of equipment deployment and triggering conditions, resulting in missed images or invalid triggers. Traditional methods indiscriminately input sighting points from different sources into the model, which easily introduces noise and reduces the accuracy of habitat suitability prediction.

[0004] A large number of uncertain observation points have not been effectively utilized. The Guizhou golden monkey is a rare and endangered species, with elusive activities and a low encounter rate. Samples of locations where individual presence can be confirmed are scarce, while the survey yielded numerous "suspected but unconfirmed" observation records, such as feeding traces, activity traces, and single-visit information. Traditional maximum entropy models struggle to incorporate such uncertain data, resulting in insufficient model generalization ability and unstable prediction results under small sample conditions.

[0005] Therefore, there is an urgent need for a habitat suitability assessment method that can adaptively fuse multi-source survey data with credibility and effectively utilize information from uncertain observation points. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method for evaluating the habitat suitability of the Guizhou golden monkey ecological corridor based on credibility fusion. By performing credibility-adaptive weighted fusion on multi-source occurrence point data and introducing uncertain observation points into the maximum entropy habitat suitability model in a semi-supervised regularization manner, it helps to improve the stability of habitat suitability evaluation results under small sample conditions and provides technical support for optimizing vegetation and food source configuration in ecological corridor habitats.

[0007] In a first aspect, embodiments of the present invention provide a method for evaluating the habitat suitability of the Guizhou golden snub-nosed monkey ecological corridor based on credibility fusion, including:

[0008] Multiple survey methods were used to obtain data on the occurrence points of the Guizhou golden monkey and the corresponding environmental variable data within the ecological corridor construction area, and environmental attribute vectors and environmental variable feature vectors were constructed based on the environmental variable data.

[0009] Based on the environmental attribute vector, the credibility of each occurrence point data is evaluated, and the credibility weight and time decay factor of each occurrence point data are determined.

[0010] The corresponding equivalent weight is determined based on the confidence weight and time decay factor of each occurrence point data, and each occurrence point data is divided into confirmed occurrence points and uncertain observation points according to the confidence weight.

[0011] The uncertain observation points are introduced into a pre-constructed maximum entropy habitat suitability model, and the data of each occurrence point in the objective function of the model are weighted according to the equivalent weights. The weighted objective function is solved iteratively to obtain the habitat suitability index distribution.

[0012] Based on the distribution results of the habitat suitability index, the habitat utilization of each ecological corridor construction area is evaluated, and suggestions for optimizing the configuration of corridor vegetation and food sources are proposed.

[0013] In a preferred embodiment, the environmental variable data includes habitat type information, behavioral characteristic information, and vegetation, topography, and disturbance factor information. Based on the environmental variable data, an environmental attribute vector and an environmental variable feature vector are constructed, including:

[0014] The habitat type information is sorted according to the coverage of the main food source plants of the Guizhou golden monkey by each habitat type and converted into normalized habitat type code values.

[0015] The behavioral feature information is converted into normalized behavioral feature encoding values ​​according to the ecological differences corresponding to the behavioral type;

[0016] The normalized habitat type code value, the normalized behavioral feature code value, and the location information of the observation record are used together to form the attribute vector;

[0017] The vegetation, topography, and disturbance factor information are standardized to form the environmental variable feature vector.

[0018] In a preferred embodiment, the credibility of each occurrence point data is evaluated based on the environmental attribute vector to determine the credibility weight and time decay factor of each occurrence point data, including:

[0019] Calculate the Mahalanobis distance between the environmental attribute vectors of point data obtained by different survey methods within the same spatiotemporal neighborhood, and measure the degree of attribute consistency between point data based on the Mahalanobis distance.

[0020] Based on the basic credibility prior of the investigation method to which each point of occurrence data belongs, the basic credibility prior is modified according to the degree of consistency of attributes among the point of occurrence data to determine the credibility weight of the point of occurrence data.

[0021] By weighted and fused multiple activity observation indicators of Guizhou golden monkeys in fixed monitoring plots during each monitoring period, the measured value of habitat use intensity for the current period was obtained.

[0022] Based on the degree of matching between the measured value of habitat use intensity in the current period and the prediction results of historical models, the time decay factor corresponding to the historical data in the current period is adaptively solved.

[0023] In a preferred embodiment, the equivalent weight of each occurrence point data is determined based on its confidence weight and time decay factor, and the occurrence point data is further divided into confirmed occurrence points and uncertain observation points based on the confidence weight, including:

[0024] Multiply the credibility weight of each occurrence point data by the time decay factor, and use the product as the equivalent weight of that occurrence point data;

[0025] Data points with a confidence weight greater than or equal to a preset confidence threshold are classified as confirmed occurrence points, while data points with a confidence weight less than the preset confidence threshold are classified as uncertain observation points.

[0026] As a preferred implementation, the maximum entropy habitat suitability model is constructed through the following steps:

[0027] Using the environmental variable weight vector as the model parameters to be solved, the inner product of the environmental variable feature vector and the weight vector of each environmental unit is used as the linear score of that environmental unit.

[0028] The linear scores of all environmental units within the study area are normalized exponentially to obtain the habitat suitability index distribution of each environmental unit, which is used as the output of the maximum entropy habitat suitability model.

[0029] By using the identified occurrence point as a constraint and maximizing information entropy as the optimization objective, the optimal weight vector is iteratively solved to determine all model parameters, thus completing the construction of the maximum entropy habitat suitability model.

[0030] In a preferred embodiment, the uncertain observation points are introduced into a pre-constructed maximum entropy habitat suitability model, and the data at each occurrence point in the model's objective function are weighted according to the equivalent weights. The weighted objective function is then iteratively optimized to obtain the habitat suitability index distribution, including:

[0031] Based on the similarity between the feature vectors of the environmental variables, a similarity constraint is constructed to bring the uncertain observation point closer to the confirmed occurrence point;

[0032] The similarity constraint is introduced into the objective function of the maximum entropy habitat suitability model in the form of a regularization term. The objective function is composed of the difference between the maximum entropy likelihood term based on the confirmed occurrence point and the regularization term.

[0033] The confirmed occurrence points and uncertain observation points in the objective function are differentially weighted according to the equivalent weights, wherein the equivalent weights of the confirmed occurrence points are used to weight their contribution to the likelihood term, and the equivalent weights of the uncertain observation points are used to weight their contribution to the regularization term.

[0034] The weighted objective function is iteratively optimized to obtain the habitat suitability index distribution of the study area.

[0035] In a preferred embodiment, the similarity constraint is introduced into the objective function of the maximum entropy habitat suitability model in the form of a regularization term, including:

[0036] Calculate the similarity function value between the eigenvectors of environmental variables corresponding to each uncertain observation point and each confirmed occurrence point;

[0037] The similarity regularization term is determined based on the equivalent weights of each uncertain observation point and each confirmed occurrence point, as well as the similarity function value.

[0038] The similarity regularization term is introduced into the objective function of the maximum entropy habitat suitability model.

[0039] As a preferred implementation, the weighted objective function is iteratively optimized to obtain the habitat suitability index distribution of the study area, including:

[0040] Calculate the weighted empirical mean of the environmental features of the confirmed occurrence points based on their equivalent weights;

[0041] Based on the parameters of the current maximum entropy habitat suitability model, calculate the expected environmental characteristics of all environmental units within the study area;

[0042] The difference between the weighted empirical mean of the environmental features and the expected value of the environmental features is used as the gradient direction of the maximum entropy likelihood term in the objective function;

[0043] The gradient of the regularization term in the weighted objective function with respect to the environmental variable weight vector is taken as the gradient direction of the regularization term in the objective function.

[0044] The difference between the gradient direction of the maximum entropy likelihood term and the gradient direction of the regularization term is taken as the total gradient of the objective function.

[0045] The environmental variable weight vector is updated based on the total gradient, and the iteration is repeated until the convergence condition is met, and the normalized habitat suitability index of each environmental unit is output.

[0046] In a preferred embodiment, the method further includes:

[0047] Based on the occurrence point data from different survey methods, separate habitat suitability models corresponding to each survey method are independently constructed, and habitat suitability estimates and their uncertainties are obtained from the output of each model.

[0048] The fusion weights of each model are determined based on the uncertainty of each model's output, and the habitat suitability estimates of each model are weighted and combined according to the fusion weights to obtain the habitat suitability index distribution after multi-model fusion.

[0049] As a preferred implementation, based on the habitat suitability index distribution results, the habitat utilization of each ecological corridor construction area is evaluated, and optimization suggestions for corridor vegetation and food source configuration are proposed, including:

[0050] Based on the habitat suitability index, each grid cell within the study area is divided into multiple suitability levels;

[0051] The activity intensity index of the Guizhou golden monkey was compared and analyzed with the actual measured data of each suitability level and each fixed monitoring plot.

[0052] Areas with habitat suitability index higher than the first preset threshold and measured activity intensity lower than the second preset threshold are identified as key areas for food source allocation optimization.

[0053] The first preset threshold is determined based on the classification results of high suitability levels, and the second preset threshold is determined based on the low-value segment in the statistical distribution of measured activity intensity.

[0054] Secondly, embodiments of the present invention also provide a habitat suitability assessment system for Guizhou golden snub-nosed monkey ecological corridors based on credibility fusion, used to execute the habitat suitability assessment method for Guizhou golden snub-nosed monkey ecological corridors based on credibility fusion described in any of the above embodiments, the system comprising:

[0055] The multi-source data acquisition module is used to acquire data on the occurrence points of the Guizhou golden monkey and the corresponding environmental variable data in the ecological corridor construction area using various survey methods, and to construct environmental attribute vectors and environmental variable feature vectors based on the environmental variable data.

[0056] The credibility assessment module is used to assess the credibility of each occurrence point data based on the environmental attribute vector, and determine the credibility weight and time decay factor of each occurrence point data.

[0057] The weight fusion and classification module is used to determine the corresponding equivalent weight of each occurrence point data according to its credibility weight and time decay factor, and to classify each occurrence point data into confirmed occurrence points and uncertain observation points according to its credibility weight.

[0058] The habitat suitability modeling module is used to introduce the uncertain observation points into the pre-constructed maximum entropy habitat suitability model, and to weight the data of each occurrence point in the objective function of the model according to the equivalent weights, and to iteratively optimize and solve the weighted objective function to obtain the habitat suitability index distribution.

[0059] The habitat utilization evaluation and optimization module is used to evaluate the habitat utilization of each ecological corridor construction area based on the habitat suitability index distribution results, and to propose optimization suggestions for corridor vegetation and food source configuration.

[0060] Thirdly, embodiments of the present invention also provide an electronic device, the electronic device comprising:

[0061] One or more processors;

[0062] Storage device for storing one or more programs;

[0063] When the one or more programs are executed by the one or more processors, the one or more processors implement the Guizhou golden monkey ecological corridor habitat suitability evaluation method based on credibility fusion as described in any embodiment of the present invention.

[0064] Fourthly, embodiments of the present invention also provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for evaluating the habitat suitability of the Guizhou golden monkey ecological corridor based on credibility fusion as described in any embodiment of the present invention.

[0065] In summary, the present invention achieves the following beneficial effects:

[0066] This invention constructs a reliability-adaptive fusion mechanism for multi-source survey data, incorporating data from site visits, transect surveys, and infrared camera monitoring into a unified evaluation framework. It differentiates the model fitting and constraints using an equivalent weight combining reliability weights and timeliness decay factors, ensuring that high-reliability, time-sensitive data dominates the model. Furthermore, a semi-supervised regularization mechanism effectively introduces uncertain observation points, previously discarded in traditional methods, into the maximum entropy model training. This causes uncertain observation points to converge towards confirmed locations with similar environments, mitigating the small-sample overfitting problem caused by scarce confirmed samples of rare species and significantly improving the robustness of habitat suitability assessment results. Finally, a spatial comparative analysis of suitability distribution and measured activity intensity provides a clear spatial decision-making basis for ecological corridor vegetation restoration and food source plant configuration optimization. Attached Figure Description

[0067] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings. The drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0068] Figure 1 This is a flowchart of a method for evaluating the habitat suitability of the Guizhou golden monkey ecological corridor based on credibility fusion, provided by an embodiment of the present invention.

[0069] Figure 2 This is a schematic diagram of the overall architecture of a method for evaluating the habitat suitability of Guizhou golden monkey ecological corridors based on credibility fusion, provided in an embodiment of the present invention.

[0070] Figure 3 This is a flowchart of the iterative solution of model parameters in semi-supervised maximum entropy habitat suitability modeling provided in this embodiment of the invention;

[0071] Figure 4 This is a schematic diagram of the structure of a Guizhou golden monkey ecological corridor habitat suitability evaluation system based on credibility fusion provided in an embodiment of the present invention;

[0072] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0073] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, the accompanying drawings show only the parts relevant to the present invention, and not all of the structures.

[0074] Before discussing the exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts describe operations (or steps) as being processed sequentially, many of these operations (or steps) may be performed in parallel, concurrently, or simultaneously. Furthermore, the order of the operations may be rearranged. The process may be terminated when its operation is completed, but may also have additional steps not included in the figures. The process may correspond to a method, function, procedure, subroutine, subroutine, etc.

[0075] Example 1

[0076] like Figure 1 As shown in the flowchart, Embodiment 1 of the present invention provides a method 100 for evaluating the habitat suitability of the Guizhou golden monkey ecological corridor based on credibility fusion. The method 100 specifically includes the following steps:

[0077] Step S110: Use various survey methods to obtain data on the occurrence points of the Guizhou golden monkey and the corresponding environmental variable data within the ecological corridor construction area, and construct environmental attribute vectors and environmental variable feature vectors based on the environmental variable data.

[0078] Specifically, in this embodiment of the invention, the selected sites in Fanjingshan, such as Walnut Bay, Niujiaodong, Changgangling, Maxia'ao, and Liujiawan, are used as ecological corridor construction areas. The following multiple survey methods are employed to simultaneously obtain site data and environmental variable data:

[0079] Interview survey method: Survey local residents, forest rangers and other people familiar with the activities of monkey troops to obtain historical sighting information, record the geographical location, time and initial credibility marker of the sighting, as a source of supplementary data.

[0080] Transect survey method: Set up 5-10 transects in the study area, with each transect being no less than 3 km long. The monitoring range is 100 m on each side of the transect. During the movement along the transect, record the actual sightings of monkeys, feeding traces, activity traces, and corresponding geographical locations and distribution habitats.

[0081] Infrared camera monitoring: Within the ecological corridor area, infrared cameras are deployed in a grid pattern to capture images of the Guizhou golden monkeys from different angles and heights. The cameras utilize infrared pyroelectric sensors and feature 24 / 7 automatic trigger monitoring; recording begins as soon as an animal enters the sensing range, ensuring effective recording of animal activity within the corridor. Based on the acquired image data, further methods can be used for identifying the number of Guizhou golden monkeys and determining their behavioral patterns (including foraging, movement, and resting). Additionally, for adult individuals exhibiting clear dimorphic sexual characteristics such as body size and coloration, sex determination can be attempted.

[0082] During the implementation of the above three survey methods, environmental variable data were collected simultaneously at each occurrence point and sample plot. These included habitat type information, behavioral characteristic information, and vegetation, topography, and disturbance factor information. Among these, habitat type information refers to the coverage level of the main food source plants (Fagaceae, Lauraceae, etc.) of the Guizhou golden monkey at each occurrence point; behavioral characteristic information refers to the category of Guizhou golden monkey behavioral patterns recorded by infrared camera monitoring or transect surveys, including foraging, movement, and resting behaviors; vegetation information includes vegetation type and canopy closure, food source plant species and richness; topography information includes slope, aspect, and altitude; and disturbance factor information includes distance from corridors, distance from roads, and distance from settlements. After processing, the above environmental variable data formed an environmental variable layer database for the study area, providing spatial environmental information support for subsequent habitat suitability modeling.

[0083] After completing the above data collection, environmental attribute vectors and environmental variable feature vectors are constructed based on the environmental variable data. The specific construction method of the environmental attribute vector is as follows:

[0084] First, the categorical variables in the environmental variable data are subjected to ecologically continuous coding. For habitat type information, the coverage of the main food source plants of the Guizhou golden monkey is sorted according to each habitat type, and a normalized habitat type code value is assigned to each habitat type. The higher the coverage ranking, the larger the code value. This code value is used as the corresponding occurrence point data in the attribute vector. The value of the habitat type component is represented by the difference between any two records in this component, which is the coding distance between them. Habitat types with adjacent coverage rankings have similar coding values, while habitat types with large ranking differences have larger coding value differences. For behavioral characteristic information, normalized behavioral characteristic coding values ​​are assigned according to the general ecological differences of each behavior on individual energy metabolism and exposure risk, such as 0.7 for foraging, 0.5 for exercise, and 0.3 for resting. This coding value only reflects the category difference of the behavior type itself and is independently determined from the habitat type component to avoid a linear correlation between the two in terms of numerical values.

[0085] Next, the normalized habitat type code value, the normalized behavioral feature code value, and the location information of the occurrence point data are combined to form the environmental attribute vector of each occurrence point data, which is used to evaluate the credibility of the occurrence point data in subsequent steps.

[0086] The construction method of environmental variable feature vectors includes: standardizing the vegetation information, topographic information and interference factor information in the environmental variable data to unify the dimensions of each environmental factor, eliminate the bias of different dimensions on subsequent model calculations, and construct the environmental variable feature vectors of each environmental unit, which are used to construct the maximum entropy habitat suitability model in subsequent steps.

[0087] Step S120: Evaluate the credibility of each occurrence point data according to the environmental attribute vector, and determine the credibility weight and time decay factor of each occurrence point data.

[0088] This step addresses the issue of inconsistent data reliability obtained through different survey methods by constructing a multi-source data reliability adaptive fusion mechanism. The core idea of ​​this mechanism is as follows: First, calculate the Mahalanobis distance between environmental attribute vectors of point data obtained from different survey methods within the same spatiotemporal neighborhood, using this distance to measure the degree of attribute consistency among the point data. Then, using the basic reliability prior of the survey method to which each point data belongs as a benchmark, correct the basic reliability prior based on the degree of attribute consistency among the point data to determine the reliability weight of that point data. Simultaneously, weighted fusion of multiple Guizhou golden monkey activity observation indicators from fixed monitoring plots in each monitoring period yields the measured value of habitat use intensity for the current period. Finally, based on the degree of matching between the measured value of habitat use intensity for the current period and the prediction results of historical models, adaptively solve for the time-decrease factor corresponding to the historical data for the current period.

[0089] Preferably, the Mahalanobis distance between the attribute vectors of occurrence point data obtained by different survey methods within the same spatiotemporal neighborhood is first calculated. The Mahalanobis distance is used to measure the degree of attribute consistency among occurrence point data, including:

[0090] First, for the attribute vector The positional components are standardized, and the covariance matrix is ​​calculated after unifying the dimensions. and Mahalanobis distance This avoids an imbalance in the contribution of different dimensional attributes to distance metrics.

[0091] Then, for occurrence points obtained by different survey methods within the same spatiotemporal neighborhood, the Mahalanobis distance consistency measure between the data attribute vectors of each pair of occurrence points is calculated according to the following formula, where the attribute vector includes components such as location coordinates, habitat type encoding values, and behavioral feature encoding values:

[0092] ;

[0093] in, The Mahalanobis distance consistency measure between two occurrence point data attribute vectors; : No. The attribute vector of the points that appear; : No. The attribute vector of the points that appear; The covariance matrix of each component of the attribute vector is estimated from the attribute vectors of all points appearing in the study area and serves as a globally unified metric.

[0094] Preferably, based on the prior credibility of the survey method to which each occurrence point data belongs, the prior credibility is modified according to the degree of attribute consistency among the occurrence point data to determine the credibility weight of the occurrence point data, including:

[0095] Let candidate records be set The basic credibility prior of the investigation method is Furthermore, the prior values ​​for each method are set in advance: infrared camera > transect witness > interview survey. For example, infrared camera is set to 0.80~0.90, transect witness is set to 0.70~0.80, and interview survey is set to 0.50~0.65.

[0096] When other candidate records exist in the same spatiotemporal neighborhood (N≥2), the credibility weight is calculated using the log-odds evidence accumulation method:

[0097] ;

[0098] in, Candidate records Investigation methods The basic credibility prior value; Candidate records The types of investigative methods include infrared cameras, transect witnessing, or interview surveys; The total number of candidate observation records within the same spatiotemporal neighborhood; Candidate records Credibility weight The logarithmic odds; Indicates that within the same spatiotemporal neighborhood, excluding the current candidate record. The numbers of other candidate records besides, For the current candidate record Attribute vectors, The attribute vector of the nth other candidate record in the neighborhood; Candidate records Credibility weight; Evidence strength coefficient: Used to control the contribution of consistent evidence to the credibility weight; The similarity threshold between supporting and contradictory evidence; : Natural exponential function; where To determine the similarity threshold between supporting and contradictory evidence, a value of 0.5 can be used in this embodiment. The strength of evidence coefficient can be set to 1-2 in this embodiment.

[0099] Preferably, the credibility weighting conversion rule based on the log-probability evidence accumulation method described above needs to meet the following design objectives:

[0100] (1) When other candidate records are Highly consistent hour, Enlarge Approaching 1, it achieves a positive cumulative increase in credibility through multi-source verification;

[0101] (2) When the candidate record is obvious contradiction ( )hour, Reduce Approaching 0 aligns with the design intent of having the weight of contradictory observations approach zero.

[0102] (3) The more candidate records involved in corroboration, the more evidence items are accumulated. The more significant the positive improvement, the more it reflects the enhancing effect of observation information from multiple independent sources on the credibility evaluation results.

[0103] (4) When there are no other candidate records in the spatiotemporal neighborhood, the summation term is empty, and the above formula naturally degenerates into: That is, the credibility weight of a single-source record is directly taken from the prior credibility of the method.

[0104] Preferably, multiple activity observation indicators of the Guizhou golden monkey in fixed monitoring plots during each monitoring period are weighted and fused to obtain the measured value of habitat use intensity for the current period, including:

[0105] Let the first Periodic monitoring covers the study area A fixed monitoring plot / transect segment, with habitat use intensity of each plot forming a vector. , among which, the The sample site at the first The habitat use intensity during this period was:

[0106] ;

[0107] in, Monitoring period number; The total number of fixed monitoring plots / segments within the study area; : No. During the periodic monitoring, the vector formed by the habitat use intensity of each fixed monitoring plot; : No. During the monitoring period, the first The scalar value of habitat use intensity for each sample plot; T is the transpose operator; Weighting coefficient for the number of valid activity records; Weighting coefficient for the number of times feeding behavior is recorded; : Weighting coefficient for the number of independent individuals; and , , The sum of the three is 1, and the preferred value range is... , , ; : No. In each sample plot, the number of effective activity records per unit area, including physical sightings and effective captures by infrared cameras; : No. The number of foraging behaviors recorded per unit area in each sample plot; : No. The number of independent individuals per unit area in each sample plot.

[0108] Preferably, based on the degree of matching between the measured current habitat use intensity and the historical model prediction results, the time-decrease factor corresponding to the current historical data is adaptively solved, including: assuming the predicted value vector based on the historical cumulative model is... The consistency statistic for the difference between the two is:

[0109] ;

[0110] in, Historical cumulative model for the first The vector of predicted habitat use intensity for each monitoring plot during the period; : No. A consistency statistic between newly added monitoring data and historical cumulative model predictions; : No. New Information Vector of Covariance matrix; The inverse matrix; the covariance matrix From historical residuals The sample covariance is obtained by recursive estimation. : No. Vector of measured habitat use intensity values ​​for each historical period; : No. The vector of model-predicted values ​​for secondary habitat use intensity during the historical period.

[0111] Robust cost function After processing, the first The time-decrease factor for historical data is adaptively determined as follows:

[0112] ;

[0113] in, : The innovation statistics after robust cost function transformation; The scaling parameter of the robust cost function is derived from historical data. The absolute deviation of the median of the sequence is estimated. : No. Adaptive time-decrease factor for historical data; : A truncation function that restricts the value of the first parameter to between the second parameter (lower bound) and the third parameter (upper bound); For reference attenuation standard, The sensitivity coefficient is (0.05~0.1). The upper and lower limits of the attenuation factor can be set based on the number of historical monitoring periods and the rate of habitat change in the study area, with optimal selection. , And the parameter settings should meet the following requirements. Scale parameters Take history The absolute deviation of the median of the sequence is 1.4826 times, used for the saturation rate of the adaptive control statistic.

[0114] When the latest measured results are highly consistent with the historical model predictions Approaching Historical data is weighted more heavily; however, when significant changes in habitat occur due to factors such as corridor vegetation restoration, and new measurements deviate significantly from historical predictions, [further action is taken]. Adaptive reduction accelerates the decay of historical data weights, highlights the role of new observational evidence, and avoids the problems of lag or oversensitivity in response to fixed decay rates.

[0115] In particular, when In the initial monitoring phase, when there is no historical cumulative model for comparison, take... That is, the initial historical weights are directly assigned based on the reference decay benchmark, without any further adjustments. Statistical correction, from period 2 ( ) Calculate according to the above recursive formula. and .

[0116] In summary, this step determined the credibility weight and time decay factor for each occurrence point data. These two factors evaluate the quality of the observation data from two dimensions: spatial verification and temporal variation, providing a basis for the subsequent fusion calculation of equivalent weights.

[0117] Step S130: Determine the corresponding equivalent weight based on the confidence weight and time decay factor of each occurrence point data, and classify each occurrence point data into confirmed occurrence points and uncertain observation points based on the confidence weight.

[0118] In a preferred embodiment, the corresponding equivalent weight is determined based on the reliability weight and time decay factor of each occurrence point data, including:

[0119] Weight the credibility of each occurrence point data With the aforementioned aging decay factor Multiply them and use the product as the equivalent weight of the data at that occurrence point, for example, where Take the monitoring period to which the occurrence point belongs. Corresponding time decay factor ,Right now , For the point of occurrence The monitoring period number to which it belongs, and specifically, the equivalent weight of each occurrence point is as follows:

[0120] ;

[0121] in, Let be the credibility weight of the i-th occurrence point data, with a value range of (0,1), reflecting the reliability of this occurrence point data after multi-source verification in the spatial dimension. The time-decrease factor for the monitoring period to which the data at the i-th occurrence point belongs, with a value range of [ This reflects the reference value of the data for this monitoring period in terms of time.

[0122] In a preferred embodiment, the occurrence point data are divided into confirmed occurrence points and uncertain observation points according to the confidence weight, including:

[0123] Credibility weight Greater than or equal to the preset confidence threshold The occurrence point data ( The points are divided into confirmed occurrence points, forming a confirmed occurrence point set P;

[0124] Data points whose confidence weight is less than the preset confidence threshold ( The points are divided into uncertain observation points, forming an uncertain observation point set U;

[0125] in, The preset confidence threshold ranges from 0.7 to 0.9. The confirmed occurrence point is a reliable occurrence record that has been mutually verified by multi-source observation data. The uncertain observation point is an occurrence record that has not been verified by multiple sources, including observation records that are suspected but cannot be confirmed, such as feeding traces, activity traces, or single visit survey information.

[0126] It should be noted that, regarding entry The equivalent weights of the observation points of the set. It is still used for subsequent semi-supervised regularization term calculation, so that uncertain observation points that would otherwise be abandoned in traditional methods can participate in model training to a certain extent, achieving an adaptive fusion effect in which high-confidence, highly timely data dominate model fitting, low-confidence or outdated data participates to a certain extent, and abnormal data participates with low weight.

[0127] Step S140: Introduce the uncertain observation points into the pre-constructed maximum entropy habitat suitability model, and weight the data of each occurrence point in the objective function of the model according to the equivalent weights, iteratively optimize and solve the weighted objective function to obtain the habitat suitability index distribution.

[0128] This step is used to combine the confirmed occurrence point set P and the uncertain observation point set U obtained in step S130 with the environmental variable feature vector constructed in step S110 and the equivalent weights determined in step S130, and introduce them into the construction and solution process of the maximum entropy habitat suitability model. Its core mechanism is as follows: First, a maximum entropy model framework is constructed with the environmental variable weight vector as the parameters to be solved; then, based on the similarity between the environmental variable feature vectors, a similarity constraint is constructed to allow uncertain observation points to converge towards confirmed occurrence points, and this constraint is introduced into the model objective function in the form of a regularization term; finally, the model parameters are solved using a gradient iteration method, and the normalized habitat suitability index distribution of each grid cell in the study area is output.

[0129] The key differences between this step and the traditional maximum entropy model are twofold: First, the regularization term is constructed only for the environmental similarity relationship between uncertain observation points and confirmed occurrence points, rather than performing indiscriminate smoothing on all samples, giving semi-supervised learning a clear ecological orientation; second, both the regularization term and the maximum entropy likelihood term are uniformly weighted using equivalent weights, achieving parameter-level connection between credibility fusion and semi-supervised modeling in the likelihood and regularization terms, thereby more fully incorporating trace points and access information that are difficult to confirm but have ecological indicative significance.

[0130] As a preferred embodiment, constructing a maximum entropy habitat suitability model includes the following steps:

[0131] First, using the environmental variable weight vector as the model parameters to be solved, the environmental variable feature vectors of each environmental unit are... With the weight vector The inner product is used as the linear score of the environmental unit. Specifically, the formula for calculating the linear score is: ;

[0132] Next, the linear scores are normalized exponentially to obtain the habitat suitability index distribution for each environmental unit. ,in, The normalization constant is the sum of the linear score exponential function values ​​of all environmental units within the study area. This ensures that the sum of the suitability indices of all environmental units is 1, satisfying the basic constraints of the probability distribution. This normalized habitat suitability index distribution... As the output of the maximum entropy habitat suitability model;

[0133] Finally, confirm the set of occurrence points. To constrain and maximize information entropy as the optimization objective, the optimal weight vector is iteratively solved to determine all model parameters, thus completing the framework of the maximum entropy habitat suitability model. The core optimization objective of this model is to find the probability distribution with the maximum information entropy under the constraint of confirmed occurrence point data, thereby predicting the potential suitable habitat distribution of species within the study area.

[0134] In a preferred embodiment, the uncertain observation points are introduced into a pre-constructed maximum entropy habitat suitability model, and the data at each occurrence point in the model's objective function are weighted according to the equivalent weights, including:

[0135] The uncertain observation points are introduced into a pre-constructed maximum entropy habitat suitability model, and the data of each occurrence point in the objective function of the model are weighted according to the equivalent weights, including the following steps:

[0136] Calculate the similarity function value between the environmental variable feature vectors corresponding to each uncertain observation point and each confirmed occurrence point; determine the similarity regularization term based on the equivalent weights of each uncertain observation point and each confirmed occurrence point and the similarity function value; introduce the similarity regularization term into the objective function of the maximum entropy habitat suitability model.

[0137] For example, firstly, based on the similarity between the feature vectors of the environmental variables, a similarity constraint is constructed to bring the uncertain observation point closer to the confirmed occurrence point. Specifically, the similarity function between the feature vectors of the environmental variables is defined as follows: Then construct pairwise similarity regularization terms. Only the linear score smoothness between uncertain observation points and confirmed occurrence points is constrained, so that the suitability prediction of uncertain observation points tends to align with that of confirmed occurrence points with similar environments. The regularization term... The expression is:

[0138]

[0139] in, : A set of uncertain observation points; : Confirm the set of points that appear; Uncertain observation points Equivalent weights; : Confirm the occurrence point Equivalent weights; Uncertain observation points Confirmed occurrence point The similarity function values ​​of the environmental variable feature vectors between them; Model for uncertain observation points The linear score of the environmental unit in which it is located; The model confirms the occurrence point. The linear score of the environmental unit in which it is located; Regularization weights are used to balance the likelihood fit at confirmed points with the utilization of information from uncertain points. The regularization term... The construction method is as follows: multiply the product of the equivalent weights of each uncertain observation point and each confirmed occurrence point by the corresponding similarity function value, and then multiply by the square of the difference between their linear scores. Take the average of the weighted sums of all combinations of uncertain observation points and confirmed occurrence points (divided by twice the sum of the products of the equivalent weights) as the similarity regularization term. .

[0140] Next, the similarity constraint is introduced into the objective function of the maximum entropy habitat suitability model in the form of a similarity regularization term. The objective function is composed of the difference between the maximum entropy likelihood term based on the confirmed occurrence point and the regularization term, i.e.:

[0141] ;

[0142] in, The regularization weight is set to 0.1 in this embodiment to balance the likelihood fitting of confirmed points and the utilization of information from uncertain points.

[0143] Based on this, the confirmed occurrence points and uncertain observation points in the objective function are differentially weighted according to the equivalent weights. The equivalent weights of the confirmed occurrence points are used to weight their contribution to the likelihood term, and the equivalent weights of the uncertain observation points are used to weight their contribution to the regularization term, thereby forming a mechanism to connect credibility fusion and semi-supervised modeling.

[0144] Through the above design, the regularization term has a clear ecological orientation, that is, the habitat suitability index of uncertain observation points should approach the confirmed occurrence points with similar environmental characteristics, which is different from the traditional semi-supervised method of smoothing all samples without difference.

[0145] As a preferred embodiment, combined with Figure 3 As shown, the habitat suitability index distribution of the study area is obtained by iteratively optimizing the weighted objective function, including the following steps:

[0146] First, based on the equivalent weights of the confirmed occurrence points, the weighted empirical mean of the environmental features of the confirmed occurrence points is calculated. :

[0147] ;

[0148] in, : Confirm the occurrence point The environmental variable feature vector at that location, For the point of occurrence The equivalent weight, the weighted empirical mean This reflects the weighted center location of the confirmed occurrence point in the environmental space, serving as the target constraint for fitting the maximum entropy likelihood term.

[0149] Next, based on the parameters of the current maximum entropy habitat suitability model, the expected environmental characteristics of all environmental units within the study area are calculated;

[0150] Then, the difference between the weighted empirical mean of the environmental features and the expected value of the environmental features is used as the gradient direction of the maximum entropy likelihood term in the objective function. This gradient direction indicates in which direction the weight vector should be adjusted so that the model prediction approximates the confirmed point distribution.

[0151] The gradient of the regularization term in the weighted objective function with respect to the environmental variable weight vector is taken as the gradient direction of the regularization term in the objective function. This gradient direction indicates in which direction the weight vector should be adjusted so that the prediction of uncertain observation points moves closer to the confirmed occurrence points with similar environments.

[0152] The difference between the gradient direction of the maximum entropy likelihood term and the gradient direction of the regularization term is taken as the total gradient of the objective function; the environmental variable weight vector is updated according to the total gradient at a preset learning rate η. In this embodiment, η is 0.01. Specifically, the model parameters... Solving through gradient iteration, the th... The update rules are as follows:

[0153]

[0154] : No. During round iteration, the current parameter The model predicts the expected environmental characteristics of each environmental unit in the study area. : No. The model parameter vector during each iteration; : No. The model parameter vector updated after each iteration; The learning rate controls the step size of parameter updates in each iteration. : No. Model parameter vector during round iteration The transpose of the matrix; Uncertain observation points The feature vector of environmental variables at that location; : Confirm the occurrence point The feature vector of environmental variables at that location; : For sets and Double summation is performed on all observation points.

[0155] After each iteration, The projection was truncated, exceeding a set threshold. Scaling proportionally Environmental variable feature vectors can be taken An empirical multiple of the upper bound of the absolute value of each component after standardization, such as 3-5 times, is used to prevent parameter divergence.

[0156] Repeat the above iterative process until the convergence condition is met. If the change in the objective function value between two adjacent iterations is less than the preset tolerance, output the normalized habitat suitability index of each grid cell in the study area.

[0157] As an optional implementation, the outputs of multiple models can be weighted and referenced based on the uncertainty of the prediction results from different data sources. Specifically, for the same environmental unit, if the transect / interview survey branch and the infrared camera monitoring branch independently provide suitability estimates and their uncertainties, the uncertainty can be represented by the variance of the predicted samples obtained from bootstrap resampling, and the branch with smaller variance can be weighted to provide multi-source mutually corroborating reference information for habitat suitability assessment.

[0158] Step S150: Based on the distribution results of the habitat suitability index, evaluate the habitat utilization of each ecological corridor construction area and propose optimization suggestions for corridor vegetation and food source configuration.

[0159] In a preferred embodiment, firstly, based on the distribution results of the normalized habitat suitability index of each grid cell in the study area output in step S140, the grid cells in the study area are sorted from high to low according to the habitat suitability index values, and divided into three levels: high suitability, medium suitability, and low suitability. For example, an equal-interval division method can be used, such as dividing the suitability index interval into three equal segments, or a quantile division method can be used to ensure that the number of grid cells in each level is approximately equal. The specific division criteria can be adjusted according to the actual ecological conditions of the study area and the corridor construction goals.

[0160] The activity intensity indicators of Guizhou golden monkeys in each suitability level zone were compared and analyzed with those measured in each fixed monitoring plot. The activity intensity indicators included the population size, behavioral types (foraging, exercise, rest, etc.), and seasonal activity intensity data of Guizhou golden monkeys in each fixed monitoring plot.

[0161] The goal of the comparative analysis is to identify the differences between measured activity intensity and theoretical expectations within each suitability level. Specifically, in highly suitable areas, environmental conditions should theoretically support the survival and activity of the Guizhou golden monkey. If the measured activity intensity is high, it indicates that the suitability assessment results for this area are consistent with actual utilization, and the corridor habitat is in a state of good utilization. If the measured activity intensity is low, it indicates that there are limiting factors restricting the actual utilization by the monkey population, and further identification of specific limiting factors is needed, such as insufficient food plants, poor habitat connectivity, and high intensity of human disturbance. Similarly, the same comparative analysis is conducted in moderately and lowly suitable areas to comprehensively assess the actual utilization level of the monkey population under different suitability levels.

[0162] For example, a spatial overlay analysis method can be used to overlay the distribution maps of each level of region with the spatial distribution maps of measured activity intensity of each sample plot, and to determine whether the spatial overlap between high suitability level sections and low activity intensity sections is statistically significant through statistical tests.

[0163] Based on the comparative analysis results above, areas with a habitat suitability index higher than the first preset threshold and measured activity intensity lower than the second preset threshold were identified as key areas for food source allocation optimization. The first preset threshold is the lower limit of the high suitability level, i.e., the lowest index value corresponding to the highest-level grid cell after sorting by habitat suitability index from highest to lowest. The second preset threshold is the mean of measured activity intensity in each monitoring plot within the same level segment minus one standard deviation. This area is the core optimization target area to be output. A high suitability index but relatively low current activity intensity indicates that the area has a good basic habitat condition, but it has not been fully and effectively utilized by the monkey population, and urgently needs optimization through artificial intervention.

[0164] Furthermore, by synthesizing the analysis results of the above sub-steps, the habitat utilization and corridor connectivity effectiveness of each ecological corridor construction area are evaluated. Specifically, the measured activity intensity of each fixed monitoring plot within the same level section is compared to identify plots with activity intensity significantly lower than the average level of the same level section. Possible causes are diagnosed based on the specific environmental characteristics of these plots, such as low food source plant richness, significant plot marginalization effect, and high intensity of human disturbance. For the identified key areas for food source configuration optimization, targeted suggestions for optimizing the configuration of native food source plants are proposed. For example, supplementing the planting of the main food source plants for the Guizhou golden monkey, which are currently lacking in the area, increases the species diversity and spatial distribution evenness of food source plants, enhances the habitat carrying capacity of the area, and promotes the actual realization of corridor connectivity. Simultaneously, for areas with low habitat suitability but high activity intensity, their limiting factors are analyzed, such as insufficient vegetation canopy closure and steep terrain, providing a reference for corridor habitat restoration.

[0165] Based on the above evaluation and optimization suggestions, we can further develop regional ecological corridor vegetation management plans, clarify the priority intervention order, recommended food source plant species and configuration patterns for each key optimization area, and provide scientific basis and technical support for optimizing vegetation and food source configuration in ecological corridor habitats.

[0166] Based on the above embodiments, the present invention achieves the following beneficial effects:

[0167] (1) This invention incorporates visit surveys, transect surveys and infrared camera monitoring data into a unified evaluation framework through adaptive fusion of the credibility of multi-source survey data, and effectively incorporates uncertain observation points (trace points, single-source access information, etc.) abandoned in traditional methods into model training using a semi-supervised mechanism, thereby alleviating the problem of scarce confirmed samples of rare and endangered species.

[0168] (2) This invention differentiates the model fitting and constraints by using equivalent weights, so that data with high credibility and timeliness dominate the model, low credibility or outdated data participates moderately, and abnormal data participates with low weight, thereby enhancing the robustness of habitat suitability evaluation results under small sample conditions.

[0169] (3) By comparing and analyzing the results of habitat suitability evaluation with the measured population activity intensity, this invention identifies suitable but underutilized areas, providing clear spatial guidance for corridor vegetation restoration and food source plant configuration optimization, and has direct production application value.

[0170] Example 2

[0171] This embodiment uses the Fanjingshan Walnut Bay Ecological Corridor construction area as an example to illustrate the specific implementation process of the method of the present invention, combined with... Figure 2 As shown, this embodiment specifically includes the following steps:

[0172] (I) Acquisition of multi-source monitoring data

[0173] This embodiment selects the Walnut Bay Ecological Corridor construction area in Fanjingshan as the research object. Two transects were set up in this area, each with a length of no less than 3 km. The monitoring range was 100 m on both sides of the transect. Three fixed monitoring plots were set up, and infrared cameras were installed in a 50 m × 50 m grid pattern for 24-hour all-weather triggered monitoring. Combined with interviews and surveys of surrounding residents, a total of 22 occurrence points were obtained, of which 12 occurrence points were confirmed by image / multi-source verification, and 10 were uncertain observation points of trace type. At the same time, environmental variables such as vegetation type, food source plant richness, slope, altitude, and distance from the interference source were extracted for each point to form an environmental variable layer database of the study area.

[0174] (II) Credibility Assessment and Occurrence Classification

[0175] First, the environmental attribute vectors of each occurrence point data are categorized and standardized based on ecological similarity, and the Mahalanobis distance between each pair of occurrence point data is calculated to measure the degree of attribute consistency between occurrence point data from different sources within the same spatiotemporal neighborhood.

[0176] Then, using the basic credibility prior of each investigation method as a benchmark (e.g., 0.85 for infrared cameras, 0.75 for transect witnesses, and 0.60 for interviews), the credibility weight is calculated using an evidence accumulation method. When other candidate records exist in the same spatiotemporal neighborhood, the basic credibility prior is adjusted by logarithmic probability evidence accumulation based on the degree of attribute consistency; the higher the degree of consistency and the more candidate records involved in corroboration, the more significant the increase in credibility weight; conversely, if there are obvious contradictions between candidate records, the credibility weight is reduced accordingly. When there are no other candidate records in the neighborhood, the credibility weight is directly taken from the basic credibility prior of that method.

[0177] Finally, with a confidence threshold of 0.8, the occurrence point data with a confidence weight greater than or equal to 0.8 are divided into a confirmed occurrence point set P; and the occurrence point data with a confidence weight less than 0.8 are divided into an uncertain observation point set U.

[0178] (III) Calculation of Time-Degradation Factor and Equivalent Weight

[0179] This embodiment involves data from two monitoring years, with different time-series decay factors for each period. For the first monitoring period (Period 1), since there is no historical cumulative model available for comparison, the time-series decay factor is directly taken as the reference decay benchmark value of 0.95.

[0180] For the second phase of monitoring data, the number of valid activity records per unit area, the number of foraging behavior records per unit area, and the number of independent individuals per unit area were first counted for two fixed monitoring plots. These three indicators were then weighted and summed according to preset weights to obtain the measured habitat use intensity for each monitoring plot in the second phase. Next, based on the historical cumulative model constructed from the first phase data, the habitat use intensity for each monitoring plot in the second phase was predicted, yielding the corresponding predicted values. The measured values ​​for each monitoring plot were compared with the predicted values, and the overall deviation between the two was calculated. After adaptive transformation, the time-degradation factor for the historical data in the second phase was dynamically determined.

[0181] The time-dependent decay factor reflects the reference value of the data in the first period when modeling in the second period. If the data in the first period is still highly consistent with the current habitat status, the decay factor will approach the upper limit, retaining more weight of historical data. If the habitat changes significantly due to reasons such as corridor vegetation restoration, the decay factor will decrease adaptively, accelerating the decay of historical data and making the modeling results more focused on the latest observational evidence.

[0182] Finally, the reliability weight of each occurrence point data is multiplied by the timeliness decay factor of the corresponding monitoring period. The occurrence point data in the first period is multiplied by the first period decay factor, and the occurrence point data in the second period is multiplied by the second period decay factor. The product is the equivalent weight of each occurrence point data. This equivalent weight comprehensively reflects the differences in reliability in the spatial dimension and the differences in timeliness in the temporal dimension of each occurrence point data, serving as a unified quantitative basis for the contribution of each occurrence point data in subsequent model construction.

[0183] (iv) Model building and iterative solution

[0184] Based on the confirmed occurrence point set P, the uncertain occurrence point set U, and the equivalent weights of each occurrence point data, a weighted semi-supervised maximum entropy habitat suitability model is constructed. In this embodiment, the model parameters are set as follows: the regularization weight is 0.1, used to balance the likelihood fitting of confirmed occurrence points and the utilization of information from uncertain observation points; the learning rate is 0.01, used to control the parameter update step size in each iteration; the truncation projection threshold is set to 4 times the upper bound of the absolute value of each component after standardization of the environmental variable feature vector, used to prevent the model parameters from diverging during the iteration process.

[0185] The objective function of this model consists of two terms: the first is the maximum entropy likelihood term based on confirmed occurrence points, used to fit high-confidence confirmed occurrence point data, making the model's output suitability distribution as close as possible to the environmental feature distribution of confirmed occurrence points; the second is the similarity regularization term based on uncertain observation points, used to include uncertain observation points in the model training, making its suitability predictions converge towards confirmed occurrence points with similar environments. These two terms are balanced by regularization weights, and the overall objective function is the difference between the two terms, that is, maximizing the likelihood fit while minimizing the regularization constraint.

[0186] The model parameters are solved using a gradient iteration method. In each iteration, two key directions are first calculated based on the current model parameters: the likelihood gradient direction, which is the difference between the weighted empirical mean of the confirmed occurrence points and the expected environmental features predicted by the current model. This direction guides the model to adjust towards a direction that better fits the distribution of confirmed occurrence points. The regularization gradient direction, which is the gradient of the regularization term with respect to the environmental variable weight vector, guides the model to maintain consistency with similar confirmed occurrence points at uncertain observation points. Subtracting the two gradient directions yields the total gradient direction for this iteration. The environmental variable weight vector is then updated along this direction, allowing the model to fully incorporate the ecological information carried by uncertain observation points while maintaining the fitting accuracy of confirmed occurrence points.

[0187] After each iteration, a truncation projection constraint is applied to the environmental variable weight vector. That is, if the absolute value of any component in the weight vector exceeds a preset truncation projection threshold, that component is proportionally reduced back to the threshold range to prevent parameter divergence and model instability. This iterative process is repeated until a convergence condition is met, such as when the change in the objective function value between two adjacent iterations is less than a preset tolerance. At this point, the model parameters have stabilized, and the normalized habitat suitability index distribution of each grid cell in the study area is output as input data for subsequent habitat utilization evaluation steps.

[0188] (V) Habitat Utilization Evaluation and Optimization Recommendations

[0189] The grid units within the study area were ranked from highest to lowest according to their habitat suitability index values, and divided into three levels: high suitability, medium suitability, and low suitability. For example, an equal-interval division method was used to divide the suitability index interval into three equal segments. The activity intensity index of the Guizhou golden monkey in each suitability level segment was compared and analyzed with the measured activity intensity index of each fixed monitoring plot.

[0190] The goal of the comparative analysis is to identify the differences between measured activity intensity and theoretical expectations within each level of area. Specifically, in highly suitable areas, if the measured activity intensity is high, it indicates that the suitability assessment results are consistent with actual utilization, and the corridor habitat is in good use. If the measured activity intensity is low, it indicates that there are limiting factors restricting the actual utilization by monkey populations in the area. Through the above comparative analysis, areas with high habitat suitability indices but relatively low actual activity intensity are identified. Based on this, it is determined that these areas may have limiting factors such as insufficient food source plants, poor habitat connectivity, or human disturbance. These areas are then identified as key areas for food source allocation optimization, providing a reference for optimizing the allocation of native food source plants. For example, supplementing the planting of the main food source plants for the Guizhou golden monkey, which are currently lacking in the area, can improve the habitat carrying capacity and actual utilization of the area.

[0191] Through the specific implementation process of this second embodiment, it was verified that the method of the present invention can effectively integrate multi-source survey data and make full use of information from uncertain observation points under the condition of small sample monitoring data of rare and endangered species, and output habitat suitability evaluation results and vegetation optimization suggestions with clear spatial orientation, providing an operable technical solution for ecological corridor habitat restoration and food source allocation optimization.

[0192] Example 3

[0193] Figure 4 This is a schematic diagram of the structure of a habitat suitability evaluation system for the Guizhou golden monkey ecological corridor based on credibility fusion, as provided in Embodiment 3 of the present invention. Figure 4 As shown, the system includes:

[0194] The multi-source data acquisition module 410 is used to acquire data on the occurrence points of the Guizhou golden monkey and the corresponding environmental variable data in the ecological corridor construction area using multiple survey methods, and to construct environmental attribute vectors and environmental variable feature vectors based on the environmental variable data.

[0195] The credibility assessment module 420 is used to assess the credibility of each occurrence point data based on the environmental attribute vector, and determine the credibility weight and time decay factor of each occurrence point data.

[0196] The weight fusion and classification module 430 is used to determine the corresponding equivalent weight of each occurrence point data according to its credibility weight and time decay factor, and to classify each occurrence point data into confirmed occurrence points and uncertain observation points according to its credibility weight.

[0197] The habitat suitability modeling module 440 is used to introduce the uncertain observation points into the pre-constructed maximum entropy habitat suitability model, and to weight the data of each occurrence point in the objective function of the model according to the equivalent weights, and to iteratively optimize and solve the weighted objective function to obtain the habitat suitability index distribution.

[0198] The habitat utilization evaluation and optimization module 450 is used to evaluate the habitat utilization of each ecological corridor construction area based on the habitat suitability index distribution results, and to propose optimization suggestions for corridor vegetation and food source configuration.

[0199] The Guizhou golden monkey ecological corridor habitat suitability assessment system based on credibility fusion provided in this embodiment of the invention can execute the Guizhou golden monkey ecological corridor habitat suitability assessment method based on credibility fusion provided in any of the embodiments of the invention above. It has the corresponding functions and beneficial effects of executing the Guizhou golden monkey ecological corridor habitat suitability assessment method based on credibility fusion. For detailed process, please refer to the relevant operations of the Guizhou golden monkey ecological corridor habitat suitability assessment method based on credibility fusion in the foregoing embodiments.

[0200] Example 4

[0201] Figure 5 This is a schematic diagram of the structure of an electronic device provided in Embodiment 4 of the present invention. The electronic device 10 is intended to represent various forms of digital computers, and may also represent various forms of mobile devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the invention described and / or claimed herein.

[0202] like Figure 5As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded into the RAM 13 from storage unit 18. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0203] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0204] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, central processing units (CPUs), graphics processing units (GPUs), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any suitable processor, controller, microcontroller, etc. Processor 11 executes the credibility fusion-based habitat suitability assessment method for the Guizhou golden snub-nosed monkey ecological corridor described above.

[0205] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0206] The above embodiments are merely illustrative examples and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A method for evaluating the habitat suitability of the Guizhou golden snub-nosed monkey ecological corridor based on credibility fusion, characterized in that, include: Multiple survey methods were used to obtain data on the occurrence points of the Guizhou golden monkey and the corresponding environmental variable data within the ecological corridor construction area, and environmental attribute vectors and environmental variable feature vectors were constructed based on the environmental variable data. Based on the environmental attribute vector, the credibility of each occurrence point data is evaluated, and the credibility weight and time decay factor of each occurrence point data are determined. The corresponding equivalent weight is determined based on the confidence weight and time decay factor of each occurrence point data, and each occurrence point data is divided into confirmed occurrence points and uncertain observation points according to the confidence weight. The uncertain observation points are introduced into a pre-constructed maximum entropy habitat suitability model, and the data of each occurrence point in the objective function of the model are weighted according to the equivalent weights. The weighted objective function is solved iteratively to obtain the habitat suitability index distribution. Based on the distribution results of the habitat suitability index, the habitat utilization of each ecological corridor construction area is evaluated, and suggestions for optimizing the configuration of corridor vegetation and food sources are proposed.

2. The method for evaluating the habitat suitability of the Guizhou golden snub-nosed monkey ecological corridor based on credibility fusion according to claim 1, characterized in that, The environmental variable data includes habitat type information, behavioral characteristic information, and vegetation, topography, and disturbance factor information. Based on this environmental variable data, environmental attribute vectors and environmental variable feature vectors are constructed, including: The habitat type information is sorted according to the coverage of the main food source plants of the Guizhou golden monkey by each habitat type and converted into normalized habitat type code values. The behavioral feature information is converted into normalized behavioral feature encoding values ​​according to the ecological differences corresponding to the behavioral type; The normalized habitat type code value, the normalized behavioral feature code value, and the location information of the observation record are used together to form the attribute vector; The vegetation, topography, and disturbance factor information are standardized to form the environmental variable feature vector.

3. The method for evaluating the habitat suitability of the Guizhou golden snub-nosed monkey ecological corridor based on credibility fusion according to claim 1, characterized in that, The credibility of each occurrence point data is evaluated based on the environmental attribute vector, and the credibility weight and timeliness decay factor of each occurrence point data are determined, including: Calculate the Mahalanobis distance between the attribute vectors of point data obtained by different survey methods within the same spatiotemporal neighborhood, and measure the degree of attribute consistency between point data based on the Mahalanobis distance. Based on the basic credibility prior of the investigation method to which each point of occurrence data belongs, the basic credibility prior is modified according to the degree of consistency of attributes among the point of occurrence data to determine the credibility weight of the point of occurrence data. By weighted and fused multiple activity observation indicators of Guizhou golden monkeys in fixed monitoring plots during each monitoring period, the measured value of habitat use intensity for the current period was obtained. Based on the degree of matching between the measured value of habitat use intensity in the current period and the prediction results of historical models, the time decay factor corresponding to the historical data in the current period is adaptively solved.

4. The method for evaluating the habitat suitability of the Guizhou golden snub-nosed monkey ecological corridor based on credibility fusion according to claim 3, characterized in that, The equivalent weights of each occurrence point data point are determined based on their reliability weights and time decay factors. Then, based on these reliability weights, each occurrence point data point is categorized into confirmed occurrence points and uncertain observation points, including: Multiply the credibility weight of each occurrence point data by the time decay factor, and use the product as the equivalent weight of that occurrence point data; Data points with a confidence weight greater than or equal to a preset confidence threshold are classified as confirmed occurrence points, while data points with a confidence weight less than the preset confidence threshold are classified as uncertain observation points.

5. The method for evaluating the habitat suitability of the Guizhou golden snub-nosed monkey ecological corridor based on credibility fusion according to claim 2, characterized in that, The maximum entropy habitat suitability model is constructed through the following steps: Using the environmental variable weight vector as the model parameters to be solved, the inner product of the environmental variable feature vector and the weight vector of each environmental unit is used as the linear score of that environmental unit. The linear scores of all environmental units within the study area are normalized exponentially to obtain the habitat suitability index distribution of each environmental unit, which is used as the output of the maximum entropy habitat suitability model. By using the identified occurrence point as a constraint and maximizing information entropy as the optimization objective, the optimal weight vector is iteratively solved to determine all model parameters, thus completing the construction of the maximum entropy habitat suitability model.

6. The method for evaluating the habitat suitability of the Guizhou golden snub-nosed monkey ecological corridor based on credibility fusion according to claim 5, characterized in that, The uncertain observation points are introduced into a pre-constructed maximum entropy habitat suitability model, and the data at each occurrence point in the model's objective function are weighted according to the equivalent weights. The weighted objective function is then iteratively optimized to obtain the habitat suitability index distribution, including: Based on the similarity between the feature vectors of the environmental variables, a similarity constraint is constructed to bring the uncertain observation point closer to the confirmed occurrence point; The similarity constraint is introduced into the objective function of the maximum entropy habitat suitability model in the form of a regularization term. The objective function is composed of the difference between the maximum entropy likelihood term based on the confirmed occurrence point and the regularization term. The confirmed occurrence points and uncertain observation points in the objective function are differentially weighted according to the equivalent weights, wherein the equivalent weights of the confirmed occurrence points are used to weight their contribution to the likelihood term, and the equivalent weights of the uncertain observation points are used to weight their contribution to the regularization term. The weighted objective function is iteratively optimized to obtain the habitat suitability index distribution of the study area.

7. The method for evaluating the habitat suitability of the Guizhou golden snub-nosed monkey ecological corridor based on credibility fusion according to claim 6, characterized in that, The similarity constraint is introduced into the objective function of the maximum entropy habitat suitability model in the form of a regularization term, including: Calculate the similarity function value between the eigenvectors of environmental variables corresponding to each uncertain observation point and each confirmed occurrence point; The similarity regularization term is determined based on the equivalent weights of each uncertain observation point and each confirmed occurrence point, as well as the similarity function value. The similarity regularization term is introduced into the objective function of the maximum entropy habitat suitability model.

8. The method for evaluating the habitat suitability of the Guizhou golden snub-nosed monkey ecological corridor based on credibility fusion according to claim 6, characterized in that, Iterative optimization of the weighted objective function yields the habitat suitability index distribution of the study area, including: Calculate the weighted empirical mean of the environmental features of the confirmed occurrence points based on their equivalent weights; Based on the parameters of the current maximum entropy habitat suitability model, calculate the expected environmental characteristics of all environmental units within the study area; The difference between the weighted empirical mean of the environmental features and the expected value of the environmental features is used as the gradient direction of the maximum entropy likelihood term in the objective function; The gradient of the regularization term in the weighted objective function with respect to the environmental variable weight vector is taken as the gradient direction of the regularization term in the objective function. The difference between the gradient direction of the maximum entropy likelihood term and the gradient direction of the regularization term is taken as the total gradient of the objective function. The environmental variable weight vector is updated based on the total gradient, and the iteration is repeated until the convergence condition is met, and the normalized habitat suitability index of each environmental unit is output.

9. The method for evaluating the habitat suitability of the Guizhou golden snub-nosed monkey ecological corridor based on credibility fusion according to claim 1, characterized in that, The method further includes: Based on the occurrence point data from different survey methods, separate habitat suitability models corresponding to each survey method are independently constructed, and habitat suitability estimates and their uncertainties are obtained from the output of each model. The fusion weights of each model are determined based on the uncertainty of each model's output, and the habitat suitability estimates of each model are weighted and combined according to the fusion weights to obtain the habitat suitability index distribution after multi-model fusion.

10. The method for evaluating the habitat suitability of the Guizhou golden snub-nosed monkey ecological corridor based on credibility fusion according to claim 1, characterized in that, Based on the habitat suitability index distribution results, the habitat utilization of each ecological corridor construction area is evaluated, and suggestions for optimizing the vegetation and food source configuration of the corridors are proposed, including: Based on the habitat suitability index, each grid cell within the study area is divided into multiple suitability levels; The activity intensity index of the Guizhou golden monkey was compared and analyzed with the actual measured data of each suitability level and each fixed monitoring plot. Areas with habitat suitability index higher than the first preset threshold and measured activity intensity lower than the second preset threshold are identified as key areas for food source allocation optimization. The first preset threshold is determined based on the classification results of high suitability levels, and the second preset threshold is determined based on the low-value segment in the statistical distribution of measured activity intensity.

Images