Indoor POI saliency evaluation method based on fisher discriminant

Abstract

The application discloses a kind of indoor POI's saliency evaluation method based on Fisher discriminant method, by determining POI saliency influence factor, and to the sample data collected are semantically quantified and standardized processing, generate sample dataset;Then the Fisher discriminant analysis method is used to solve POI saliency sample data, obtain the weight coefficient of the spatial position, the popularity of POI, relative scale and the influence factor of uniqueness, to determine the evaluation discriminant function of POI saliency, and using evaluation discriminant function to evaluate the sample to be measured discriminant.The application can be adapted to different indoor scenes, and the evaluation accuracy is high, and it has important significance to promote the development of indoor positioning and navigation service.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of indoor space, and in particular to a method for evaluating the saliency of indoor POI based on Fisher discriminant method. BACKGROUND

[0002] Indoor space is a real three-dimensional artificial enclosed space, and the similarity of each floor is great. The three-dimensional coordinate information provided by the positioning device (such as GPS, WIFI, Bluetooth, etc.) cannot completely meet the needs of users to determine the position and distinguish the direction. As the most basic and important space element in the cognition of indoor space, indoor landmarks can help users establish a structured understanding of the indoor environment. In the process of wayfinding, users recognize the landmarks in the map, which is an important clue to connect the real world space and the cognitive map space, and plays the role of "anchoring point".

[0003] Currently, the method of extracting landmarks is usually based on the saliency model to calculate the saliency of the point of interest (POI), and select the POI with higher saliency as the landmark, so the saliency of the POI is the main basis for extracting the landmark. However, the research on the landmark is mostly limited to the outdoor space, and the extraction of the landmark in the indoor space is rarely involved. Compared with the outdoor navigation, the indoor landmark plays a more important role in the navigation location service, and the research on the extraction of the indoor landmark based on the landmark saliency model has attracted more and more attention of researchers. In 2007, Millonig et al. proposed a system for pedestrian navigation based on landmarks, which uses a smart phone as a terminal and uses indoor landmarks to guide pedestrians to navigate in a train station, and the main problem concerned is to identify the visual attributes of the indoor landmark. In 2014, Russo et al. proposed an algorithm that can automatically generate an indoor navigation route containing visible landmarks (doors, windows, passages). Subsequently, Lyu proposed the indexes for evaluating the saliency of the indoor landmark: visibility and perceptible shape features (based on visual saliency), accessibility and location saliency (based on structural saliency) and functional saliency (based on semantic saliency), and these indexes are used to quantitatively extract the indoor landmark. Hund proved through a cognitive experiment that the visual spatial memory helps the wayfinding and orientation in the indoor space. Fellner et al. proposed a classification-based method that scores the saliency factors of the landmark in the indoor navigation, generates a navigation route based on the indoor landmark to help people find their way in an unfamiliar indoor environment. Li et al. proposed a cognitive strength grid model based on landmarks for improving indoor path planning and guidance. Shen et al. used a questionnaire survey method to study the factors affecting the saliency of the indoor landmark, and believed that the factors affecting the saliency evaluation of the indoor landmark include: utility, functional saliency, reference rate and entrance and exit. Shen et al. proposed an instance-based scoring system, analyzed the indexes affecting the saliency of the indoor space object from the aspects of vision, semantics and structure, calculated the weight coefficients of the indoor landmark saliency evaluation indexes by using the analytic hierarchy process (AHP), constructed a hierarchical structure model of the indoor landmark saliency, and verified the hierarchical structure model of the indoor landmark saliency by using the eye movement experiment method and the questionnaire survey method. Xuke et al. proposed a nonlinear model of the indoor landmark saliency constructed by using the genetic algorithm (GP).

[0004] However, most of the current indoor landmark saliency evaluation models do not fully consider the characteristics of the indoor environment and the subjective experience of the user, the evaluation index system usually adopts the expert scoring or follows the index system of the outdoor space, and the weight calculation of each index is also based too much on the past experience, so that the landmarks selected by using these methods do not match the cognition of the user. SUMMARY

[0005] Therefore, the present application aims to overcome the defects in the prior art, and provide a Fisher discriminant method-based indoor POI saliency evaluation method, which can adapt to different indoor scenes, has high evaluation accuracy, can provide a scientific basis for accurate extraction of indoor landmarks, and has important significance for promoting the development of indoor positioning and navigation services.

[0006] The Fisher discriminant method-based indoor POI saliency evaluation method of the present application comprises the following steps:

[0007] S1. Determine POI saliency influencing factors; the influencing factors include popularity, spatial position, uniqueness, and relative scale;

[0008] S2. Collect a plurality of indoor POIs, and grade the plurality of indoor POIs to obtain indoor POIs of different grades;

[0009] S3. Quantitatively process the saliency influencing factors of the plurality of indoor POIs to obtain quantitatively processed influencing factors;

[0010] S4. Standardize the quantitatively processed influencing factors to obtain standardized influencing factors;

[0011] S5. Construct a discriminant model based on the Fisher discriminant method, use indoor POIs of different grades as the initial classification results of the discriminant model, use POI saliency influencing factors as discriminant factors of the discriminant model, and use the standardized influencing factors as sample data to solve the discriminant model to obtain a discriminant function of the discriminant model and a plurality of centroid values of the discriminant function;

[0012] S6. Quantitatively and standardize the saliency influencing factors of a to-be-tested indoor POI in turn to obtain processed POI saliency influencing factors;

[0013] S7. Substitute the processed POI saliency influencing factors into the discriminant function to calculate a function value of the to-be-tested indoor POI;

[0014] S8. Calculate the distance between the function value of the to-be-tested indoor POI and each group of centroid values to obtain a plurality of distance values; and use the grade of the group corresponding to the smallest distance value in the plurality of distance values as the saliency grade of the to-be-tested indoor POI.

[0015] Further, the plurality of indoor POIs are graded to obtain indoor POIs of different grades, which specifically comprises:

[0016] S21. Collect the number of times of recalling the plurality of indoor POIs;

[0017] S22. Arrange the number of times of recalling the plurality of indoor POIs in descending order to obtain a sequence of the number of times of recalling.

[0018] S23. The recall sequence is hierarchically divided by using natural break clustering, and a classification result of the indoor POIs from high to low is obtained.

[0019] Further, the influence factors of the prominence of the indoor POIs are quantitatively processed, and the quantitatively processed influence factors are obtained, which specifically include:

[0020] The popularity of the indoor POIs is scored in a set score range, and the average value of the score is taken as the quantified value of the popularity;

[0021] The level of the indoor POIs in a set spatial position level is determined, and the score value corresponding to the level is taken as the quantified value of the spatial position;

[0022] The level of the indoor POIs in a set uniqueness level is determined, and the score value corresponding to the level is taken as the quantified value of the uniqueness;

[0023] The level of the indoor POIs in a set relative scale level is determined, and the score value corresponding to the level is taken as the quantified value of the relative scale.

[0024] Further, the step S4 specifically includes:

[0025] S41. The arithmetic mean of the quantitatively processed influence factors is calculated;

[0026] S42. The standard deviation of the quantitatively processed influence factors is calculated;

[0027] S43. The difference between the quantitatively processed influence factors and the corresponding arithmetic mean is calculated, and the ratio of the difference to the corresponding standard deviation is calculated, and the calculation result is taken as the standard value of the quantitatively processed influence factors;

[0028] S44. The standard value of all the quantitatively processed influence factors is obtained by analogy with step S43.

[0029] Further, the discriminant function of the discriminant model is determined according to the following method:

[0030] S51. A plurality of discriminant functions are obtained by solving the discriminant model;

[0031] S52. The discriminant function with the highest variance percentage is selected from the plurality of discriminant functions as a candidate discriminant function;

[0032] S53. The discriminant function with the highest regular correlation is selected from the candidate discriminant function as a target discriminant function; and the target discriminant function is taken as the discriminant function of the discriminant model.

[0033] Further, the discriminant function y1 of the discriminant model is determined according to the following formula:

[0034] y1 = γ1x1 + γ2x2 + γ3x3 + γ4x4 - δ

[0035] wherein x1 is the popularity; γ1 is the popularity coefficient; x2 is the spatial position; γ2 is the spatial position coefficient; x3 is the uniqueness; γ3 is the uniqueness coefficient; x4 is the relative scale; γ4 is the relative scale coefficient; and δ is the constant term.

[0036] The indoor POI significance evaluation method based on the Fisher discriminant method disclosed by the application has the advantages that the POI significance influence factors are determined, the collected sample data are subjected to semantic quantization and standardization processing to generate a sample data set, then the Fisher discriminant analysis method is used to solve the POI significance sample data to obtain the weight coefficients of the spatial position, the popularity, the relative scale and the uniqueness of the POI, so as to determine the evaluation discriminant function of the POI significance, and the evaluation discriminant function is used to evaluate and discriminate the to-be-tested sample. The application has wide adaptability, high evaluation and discrimination accuracy, can provide a scientific basis for the accurate extraction of indoor landmarks, and has important significance for promoting the development of indoor positioning and navigation services. BRIEF DESCRIPTION OF DRAWINGS

[0037] The application will be further described below in combination with the drawings and embodiments:

[0038] Figure 1 Fig. 1 is a method flowchart of the application;

[0039] Figure 2 Fig. 2 is a test flowchart for exploring the indoor POI significance influence factors of the application;

[0040] Figure 3 Fig. 3 is a recall person-time natural breakpoint grading histogram of the application;

[0041] Figure 4 Fig. 4 is a refinement system diagram of three significance influence factors of the application;

[0042] Figure 5 Fig. 5 is a classification diagram represented by two typical discriminant functions of the application. DETAILED DESCRIPTION

[0043] The application will be further described below in combination with the drawings and embodiments:

[0044] The indoor POI significance evaluation method based on the Fisher discriminant method disclosed by the application has the advantages that the POI significance influence factors are determined, the collected sample data are subjected to semantic quantization and standardization processing to generate a sample data set, then the Fisher discriminant analysis method is used to solve the POI significance sample data to obtain the weight coefficients of the spatial position, the popularity, the relative scale and the uniqueness of the POI, so as to determine the evaluation discriminant function of the POI significance, and the evaluation discriminant function is used to evaluate and discriminate the to-be-tested sample. The application has wide adaptability, high evaluation and discrimination accuracy, can provide a scientific basis for the accurate extraction of indoor landmarks, and has important significance for promoting the development of indoor positioning and navigation services.

[0045] S1. Determine the POI prominence influencing factors; the influencing factors include popularity, spatial location, uniqueness, and relative scale;

[0046] S2. Collect a plurality of indoor POIs, and grade the plurality of indoor POIs to obtain indoor POIs of different grades;

[0047] S3. Quantitatively process the prominence influencing factors of the plurality of indoor POIs respectively to obtain quantitatively processed influencing factors;

[0048] S4. Standardize the quantitatively processed influencing factors to obtain standardized influencing factors;

[0049] S5. Construct a discriminant model based on Fisher discriminant method, take indoor POIs of different grades as initial classification results of the discriminant model, take the POI prominence influencing factors as discriminant factors of the discriminant model, and take the standardized influencing factors as sample data to solve the discriminant model to obtain a discriminant function of the discriminant model and each group of centroid values of the discriminant function;

[0050] S6. Quantitatively and standardize the prominence influencing factors of the to-be-tested indoor POI in sequence to obtain processed POI prominence influencing factors; wherein the quantitatively processing is the same as the processing manner in step S3, and the standardization processing is the same as the processing manner in step S4, which will not be described herein again;

[0051] S7. Substitute the processed POI prominence influencing factors into the discriminant function to calculate a function value of the to-be-tested indoor POI;

[0052] S8. Calculate distances between the function value of the to-be-tested indoor POI and each group of centroid values to obtain a plurality of distance values; and take a group grade corresponding to a smallest distance value in the plurality of distance values as a prominence grade of the to-be-tested indoor POI.

[0053] In this embodiment, in step S1, in an indoor space, an indoor POI has obvious recognition and can help a user determine direction and position in an unfamiliar indoor environment, and is a "spiritual anchor" of indoor location service. However, in a high information density indoor space, the identification of an indoor POI is not fixed, and is affected by many factors, such as characteristics of the indoor POI itself (such as scale, popularity, location, etc.), characteristics of the user (knowledge structure, gender, age, demand, etc.), and external environmental factors (information carrier, light, noise). In order to study an indoor POI prominence model consistent with user cognition, the characteristics of the indoor POI itself are selected as the research object, and a research test of indoor POI prominence influencing factors in a shopping mall environment is carried out from the user behavior demand, such as Figure 2

[0054] ​Firstly, the spatial cognitive ability test questions (MCT) are obtained, and the test personnel are homogenized into 4 groups. Through homogenization processing, it is ensured that the spatial cognitive ability of each group is equal, the male and female distribution is uniform, and the age distribution is similar. Secondly, two different indoor shopping malls near a certain place are selected as test sites. Then the test personnel are in the state of "shopping", walking along the designed route from the starting point of the corresponding mall to the end point. After reaching the end point, the test individual recalls and records the POIs seen in the walking process independently. Finally, the test data is cleaned and sorted, and the POI influencing factors are determined through the offline questionnaire survey of the test personnel.

[0055] In the test, the more the number of people who recall, the higher the significance of the POI. In the analysis process, the number of people who recall is also used as the significance of the POI. To further understand the importance of each influencing factor, a field survey is conducted on each POI to obtain the popularity, spatial position, size of each POI, and whether it has uniqueness. The identification results of indoor POIs of different groups in each mall are shown in Table 1. Based on this, it is analyzed that the significance of POI is mainly influenced by popularity, and is also influenced by spatial position, uniqueness and relative size.

[0056] Table 1

[0057]

[0058]

[0059] In this embodiment, the number of POI recalls in Table 1 represents the significance division gradient of the POI, and also contains the attribute information of each POI influencing factor. Most of these sample data are semantic information. To effectively analyze them mathematically and reduce the differences between the influencing factors and the orders of magnitude as much as possible, the number of recalls is classified, the influencing factors are classified and quantified, and the standardization processing is performed.

[0060] Among them, a number of indoor POIs are classified to obtain indoor POIs of different levels, which specifically include:

[0061] S21. Collect the number of recalls of a number of indoor POIs;

[0062] S22. Arrange the number of recalls of a number of indoor POIs in descending order to obtain a recall sequence;

[0063] S23. Classify the recall sequence by natural break clustering to obtain a classification result of indoor POIs from high to low.

[0064] That is, the number of times of recalling people as the prominence of POI elements, by scientific classification of the number of times of recalling people can intuitively show the regional distribution characteristics and the relationship between various elements, so that the subsequent production of indoor map has better information transmission effect. In this embodiment, the natural break clustering is used to process the number of times of recalling people of the POI elements in table 1, that is, the POI elements of different levels can be obtained, and the natural break point histogram is shown in Figure 3 The figure shows that 10 POIs with the number of times of recalling people of 42 or more in table 1 are classified as the first level, 18 POIs with the number of times of recalling people between 15 and 42 are classified as the second level, and 23 POIs with the number of times of recalling people below 15 are classified as the third level. The more the number of times of recalling people, the higher the corresponding level.

[0065] In this embodiment, the prominence of POI is the result of the comprehensive influence of different factors. After exploring the general rule of the comprehensive influence of different factors on the prominence of POI, the attribute information data needs to be quantitatively expressed, so that the sample data can be applied to mathematical analysis. On the basis of fully considering the feedback information of the test personnel, the expert scoring method can be used to refine the spatial position, uniqueness and relative scale, and finally form the classification system of influencing factors, as shown in Figure 4 The spatial position, uniqueness and relative scale are based on the "normal" category.

[0066] Then the influencing factors of the prominence of indoor POI are quantitatively processed to obtain the quantized influencing factors, which specifically include:

[0067] The popularity of indoor POI is scored in the set score range, and the average value of the score is taken as the quantized value of the popularity. The popularity of POI is divided by quantization, all test personnel are asked to evaluate the popularity of each POI by 1 to 10 points through network questionnaire survey, and the average value is taken as the quantized value of the popularity;

[0068] The level of indoor POI in the set spatial position level is determined, and the corresponding score value of the level is taken as the quantized value of the spatial position. The spatial position level from low to high is normal, channel corner, multi-channel port, elevator port and entrance, and the corresponding score value is 1-5, and the corresponding quantized value is also 1-5. The elevator port is the vertical connecting channel between floors in the mall, including the elevator, escalator and common staircase. The entrance refers to the channel port of the mall, which is mostly two categories of space or connecting channel between buildings in the plane horizontal direction, and follows the cognitive rule from the whole to the part. The external connecting entrance has greater weight than the internal channel port, and the internal multi-channel port includes the corner.

[0069] determining the level of the indoor POI in the set uniqueness level, and taking the score value corresponding to the level as the quantized value of the uniqueness; wherein the uniqueness of the POI is represented by the uniqueness of the name, decoration and content, and the uniqueness of the decoration is considered to include the uniqueness of the name and the uniqueness of the content; that is, the uniqueness level is in turn normal, name unique, decoration unique and content unique from low to high, and the corresponding score values are in turn 1-4, and the corresponding quantized values are also in turn 1-4;

[0070] determining the level of the indoor POI in the set relative scale level, and taking the score value corresponding to the level as the quantized value of the relative scale; wherein the relative scale here includes two meanings, "relative" is relative to other POIs in the mall, and the scale refers to the size of the POI space measured on site; the relative scale level is in turn normal, larger and large from low to high, and the corresponding score values are in turn 1-3, and the corresponding quantized values are also in turn 1-3.

[0071] After rounding the quantized value of the popularity of the POI, and quantizing the spatial position, uniqueness and relative scale of the POI according to the attribute refinement system shown in Figure 4 , the partial POI classification and quantized information of the significance influencing factors in Table 1 can be obtained, as shown in Table 2.

[0072] Table 2

[0073]

[0074] In this embodiment, in order to avoid large differences between the influencing factors and the order of magnitude, the quantized influencing factors are processed by the Z-score standardization method. The step S4 specifically includes:

[0075] S41. calculating the arithmetic mean of the quantized influencing factors;

[0076] S42. calculating the standard deviation of the quantized influencing factors;

[0077] S43. calculating the difference between the quantized influencing factors and the corresponding arithmetic mean, and performing ratio operation on the difference and the corresponding standard deviation, and taking the ratio operation result as the standard value of the quantized influencing factors;

[0078] S44. according to step S43, obtaining the standard value of all quantized influencing factors.

[0079] The four influencing factors in Table 2 are processed according to the above standardization processing method, and the processed results are shown in Table 3.

[0080] Table 3

[0081]

[0082] In the embodiment, the generalized square discrimination method, the Fisher discriminant analysis method, the Mahalanobis distance discriminant method, the maximum posterior probability discriminant method and the Bayes discriminant method are common methods for classifying discriminant categories, and the Fisher discriminant analysis method is used as the category division analysis method in the application. The Fisher discriminant analysis method has no restriction on the distribution of the sample population, as long as the mean and covariance matrix of the population exist and the covariance matrix of the population is invertible. Therefore, it is not necessary to test the distribution of the sample data, which improves the prediction efficiency and has a wide range of applications.

[0083] In step S5, the FDA discriminant model for POI significance evaluation is established by using SPSS 25.0 software; wherein, the Fisher discriminant analysis method (Fisher Discriminant Analysis, FDA); the three levels of POI are taken as the initial classification results for training the Fisher discriminant model, and the four influencing factors of the POI, namely, the popularity, the spatial position, the uniqueness and the relative scale, are taken as the discriminant factors of the FDA; after the 51 POI samples in Table 1 are quantified and standardized, all of them are used to establish the FDA discriminant model for POI significance evaluation.

[0084] The discriminant function of the discriminant model is determined according to the following method:

[0085] S51. The discriminant functions are obtained by solving the discriminant model based on the FDA algorithm; in the embodiment, two discriminant functions are obtained, as shown in Table 4:

[0086] Table 4

[0087]

[0088] The eigenvalues and group centroid values corresponding to the two discriminant functions in Table 4 are shown in Table 5:

[0089] Table 5

[0090]

[0091] S52. The discriminant function with the highest variance percentage is selected from the plurality of discriminant functions as a candidate discriminant function.

[0092] S53. The discriminant function with the highest regular correlation is selected from the candidate discriminant function as a target discriminant function; and the target discriminant function is taken as the discriminant function of the discriminant model.

[0093] As shown in Table 5, in the two discriminant functions, the variance percentage of the first discriminant function is 100.0%, i.e., the first discriminant function can explain all the information of the sample, and the canonical correlation is 0.896, the variance percentage of the second discriminant function is about 0.0%, and the canonical correlation is only 0.036. Therefore, the first discriminant function is selected as the target discriminant function, and is used as the final discriminant function of the discriminant model. The group centroid values in Table 5 are the critical values of different POI level divisions.

[0094] In addition, the Fisher discriminant function is used to group the samples, and the result is shown in Table 5. Figure 5 Figure 5 In Table 5, it can be seen from the projection of the POI sample data on the discriminant function that only the first discriminant function (function 1) can obviously distinguish three different categories, which further confirms the reliability and accuracy of selecting the first discriminant function as the target discriminant function.

[0095] In this embodiment, the discriminant function y1 of the discriminant model is determined according to the following formula:

[0096] y1 = γ1x1 + γ2x2 + γ3x3 + γ4x4 - δ;

[0097] wherein x1 is the popularity, γ1 is the popularity coefficient, x2 is the spatial position, γ2 is the spatial position coefficient, x3 is the uniqueness, γ3 is the uniqueness coefficient, x4 is the relative scale, γ4 is the relative scale coefficient, and δ is a constant term. In this embodiment, the discriminant function y1 = 2.038x1 + 1.091x2 + 0.788x3 + 0.330x4 - 0.004.

[0098] In this embodiment, by substituting the processed landmark prominence influencing factors into the discriminant function y1, the function value y of the to-be-tested indoor space landmark can be calculated. By calculating the distance between the function value y of the to-be-tested indoor POI and the three group centroids, three distance values are obtained, which are |y-0.629|, |y+1.900| and |y-3.237| respectively. The group level corresponding to the smallest distance value in the three distance values is taken as the prominence level of the to-be-tested indoor POI. Wherein, one centroid group corresponds to one POI level (as shown in Table 5), after the centroid group corresponding to the smallest distance value is determined, the level corresponding to the centroid group is taken as the prominence level of the to-be-tested indoor POI.

[0099] ​In order to further explore the adaptability of the indoor POI salience evaluation method of the present application in other shopping malls, verification tests were carried out according to the method of the exploration test, other shopping malls were selected as test points, and 40 volunteers were recruited as subjects. In the test, 18 POIs were recalled by more than 5 volunteers, with a total of 211 recall times. The popularity of some brands in the mall had not been statistically analyzed before, so the questionnaire was released again and reordered, and finally 53 valid questionnaires were collected.

[0100] The obtained sample data were quantified and standardized, and the natural discontinuity method was used to divide the POIs into grades, the first 5 POIs were divided into grade 1, the 6th to 11th POIs were divided into grade 2, and the 12th to 18th POIs were divided into grade 3; the verification test data results are shown in Table 6.

[0101] Table 6

[0102]

[0103] The values of each influencing factor in Table 6 were substituted into the discriminant function y1 to obtain the Fisher discriminant model function value, the difference between the model function value and the group centroid value was compared, and then the POI salience was judged, as shown in Table 7.

[0104] Table 7

[0105]

[0106]

[0107] As can be seen from Table 7, all POIs of the first level were correctly judged, 5 of the 6 POIs of the second level were correctly judged, and 6 of the 7 POIs of the third level were correctly judged. Of the 18 POIs, 16 were correctly judged and 2 were incorrectly judged. The overall correct judgment rate was 88.9%. Through this test, it can be concluded that the salience evaluation method of the present application has a high discrimination rate in evaluating the salience of indoor POIs and is suitable for other large indoor shopping malls.

[0108] Finally, it should be pointed out that the above examples are only used to illustrate the technical solutions of the present application and are not limiting, although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application, and all should be covered in the scope of the claims of the present application.

Claims

1. A method for evaluating the significance of indoor points of interest (POIs) based on Fisher's discriminant method, characterized in that: Includes the following steps: S1. Determine the factors influencing the significance of POIs; these factors include popularity, spatial location, uniqueness, and relative size. S2. Collect several indoor POIs and classify them into different levels to obtain indoor POIs of different levels; S3. Quantify the significance of several indoor POIs to obtain the quantified influencing factors; S4. Standardize the quantified influencing factors to obtain standardized influencing factors, which include: S41. Calculate the arithmetic mean of the quantified influencing factors; S42. Calculate the standard deviation of the quantified influencing factors; S43. Calculate the difference between the quantified influencing factor and its corresponding arithmetic mean, and perform a ratio operation between the difference and its corresponding standard deviation, and use the result of the ratio operation as the standard value of the quantified influencing factor; S44. Following step S43, obtain the standard values ​​of all quantified influencing factors; S5. Construct a discriminant model based on Fisher's discriminant method, use indoor POIs of different levels as the initial classification results of the discriminant model, use the POI significance influencing factors as the discriminant factors of the discriminant model, and use the standardized influencing factors as sample data to solve the discriminant model, so as to obtain the discriminant function of the discriminant model and the centroid values ​​of each group of the discriminant function. The discriminant function of the discriminant model is determined using the following method: S51. By solving the discriminant model, several discriminant functions are obtained; S52. Select the discriminant function with the highest percentage of variance from several discriminant functions as the candidate discriminant function; S53. Select the discriminant function with the highest regularity correlation from the candidate discriminant functions as the target discriminant function; and use the target discriminant function as the discriminant function of the discriminant model; S6. The factors influencing the significance of the indoor POI to be tested are quantified and standardized in sequence to obtain the factors influencing the significance of the processed POI. S7. Substitute the factors affecting the significance of the processed POI into the discriminant function to calculate the function value of the indoor POI to be tested; S8. Calculate the distance between the function value of the indoor POI to be tested and the centroid values ​​of each group to obtain several distance values; and take the group level corresponding to the smallest distance value among the several distance values ​​as the significance level of the indoor POI to be tested.

2. The method for evaluating the significance of indoor POIs based on Fisher's discriminant method according to claim 1, characterized in that: Several indoor POIs are classified into different levels, including: S21. Collect recall data from several indoor POIs; S22. Arrange the number of recalls of several indoor POIs in descending order to obtain the recall sequence; S23. The natural discrete clustering method is used to classify the sequence of recallers into levels, and the indoor POI classification results are obtained from high to low level.

3. The method for evaluating the significance of indoor POIs based on Fisher's discriminant method according to claim 1, characterized in that: The factors influencing the significance of indoor POIs were quantified to obtain the quantified influencing factors, which specifically include: The popularity of indoor POIs is scored within a set score range, and the average score is used as the quantitative value of popularity. Determine the level of the indoor POI in the set spatial location level, and use the score value corresponding to the level as the quantitative value of the spatial location; Determine the level of the indoor POI in the set uniqueness level, and use the score value corresponding to the level as the quantitative value of uniqueness; Determine the level of the indoor POI within a set relative size hierarchy, and use the score corresponding to the level as the quantitative value of the relative size.

4. The method for evaluating the significance of indoor POIs based on Fisher's discriminant method according to claim 1, characterized in that: The discriminant function of the discriminant model is determined according to the following formula. : ； in, For brand recognition; This refers to the brand awareness coefficient. Spatial location; Spatial location coefficient; For uniqueness; The uniqueness coefficient; For relative size; This is a relative scale coefficient; This is a constant term.

Images