Campus emotion risk individual assessment method and system based on expression recognition

Abstract

The invention provides a campus emotion risk individual assessment method and system based on expression recognition, and the method comprises the steps: collecting monitoring image data for recognition, and obtaining recognized individual expression data and related information data; performing processing according to the individual expression data to obtain group emotion data; correcting the individual expression data according to the group emotion data to obtain individual negative emotion data; and evaluating an emotion risk target individual according to historical statistics of the individual negative emotion data. According to the method, group emotional abnormity can be accurately eliminated in a group scene, the individual emotional risk degree is objectively judged, and the accuracy of identifying the campus emotional risk target is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of artificial intelligence emotion computing, and in particular relates to a method and system for evaluating individual campus emotion risks based on expression recognition. Background Art

[0002] Campus emotional risk refers to the various emotional problems faced by students on campus, and the risk of negative behaviors due to emotional changes. For example, students may suffer from emotional changes due to stress or bullying, and may engage in behaviors such as self-harm or harming others.

[0003] With the widespread application of artificial intelligence pattern recognition technology, many monitoring devices have been able to automatically identify individual expressions in image data. Therefore, on campus, surveillance cameras can collect and identify students' expressions, including sadness, disgust, fear, anger, calmness, happiness, surprise, confusion, etc., and then judge whether there is a risk based on the students' expressions and issue alarm information in time. Management personnel can obtain the alarm information in time for manual tracking and processing.

[0004] The essence of this processing solution is to use individual expressions as individual emotions for evaluation and judgment. However, individual expressions cannot fully represent individual emotions. Using individual expressions as individual emotions can only simply and qualitatively identify the types of individual emotions with low accuracy.

[0005] In order to meet the requirements of actual work for the accuracy of emotion analysis, the existing technology uses machine learning to obtain relatively accurate individual emotions from individual expressions through a large amount of data annotation, model training and testing. For example, the patent "A Risk Student Prediction Method Based on Expression Recognition" with publication number CN113792627A divides the expression into positive and negative expressions through machine training according to the results of student expression recognition, thereby obtaining the student's emotional state to determine whether the student is a risk student. However, this method has a large training workload, a large amount of data storage, a high implementation cost, and a large difference in recognition accuracy for different groups of people, making it difficult to promote widely. Summary of the invention

[0006] The present invention provides a method and system for assessing individual campus emotional risks based on facial expression recognition, which can accurately eliminate group emotional anomalies in group scenarios and objectively judge individual emotional risks, thereby improving the accuracy of identifying campus emotional risk targets.

[0007] Based on the laws of nature, individual expressions are the result of individual emotions being affected by group emotions. In most cases, it can be simply expressed as: individual emotions + group emotional influence = individual expressions, for example, individual anger + group happy emotional influence = individual sad expression. Therefore, individual emotions can be obtained from the relationship between individual expressions and group emotions. The design idea of ​​the present invention is to use the influence of group emotions to modify individual expressions and obtain true individual emotions.

[0008] The technical solution of the present invention is achieved in this way:

[0009] A method for assessing individual campus emotional risk based on facial expression recognition, comprising:

[0010] S1, collect monitoring image data for recognition, and obtain the recognized individual expression data and related information data;

[0011] S2, obtaining group emotion data based on individual expression data;

[0012] S3, correcting individual expression data according to group emotion data to obtain individual negative emotion data;

[0013] S4. Evaluate emotional risk target individuals based on historical statistics of individual negative emotion data.

[0014] Furthermore, the method of correcting individual expression data according to group emotion data in step S3 includes:

[0015] Distinguish whether individual expression data belongs to risk-free data or risky data according to group negative emotion data in group emotion data, and obtain individual negative emotion data according to individual expression data belonging to risky data;

[0016] The method for distinguishing individual expression data includes: determining a target threshold of the number of people in a group, and when the number of people corresponding to a certain negative emotion in the group emotion data is greater than or equal to the target threshold of the number of people in the group emotion data, the group emotion is abnormal and belongs to risk-free data;

[0017] When the number of people with a certain negative emotion is less than the target threshold of the group, it is considered risky data.

[0018] Furthermore, the method of correcting individual expression data according to group emotion data in step S3 includes:

[0019] The individual negative expression data is corrected according to the group non-negative emotion data in the group emotion data to obtain the individual negative emotion data, where individual negative emotion = individual negative expression - group non-negative emotion impact.

[0020] Furthermore, step S3 also includes: after obtaining the individual negative emotion data, calculating the emotion intensity to obtain individual negative emotion quantitative data including the individual emotion intensity; the emotion intensity is evaluated based on the group non-negative emotion data.

[0021] Furthermore, in step S3, the number of times of collecting monitoring image data is additionally increased for the individuals corresponding to the identified risky data.

[0022] Furthermore, the method for evaluating the emotional risk target individual described in step S4 includes: setting a risk threshold for the individual's negative emotional data, and identifying the individual as an emotional risk individual if the individual exceeds the risk threshold based on the statistical number and frequency of the individual's negative emotional data.

[0023] Furthermore, step S4 also includes: evaluating emotional risk target individuals based on historical statistics of individual negative emotion data combined with multi-physiological signal data analysis.

[0024] Furthermore, the method for identifying emotional risk target individuals in step S4 includes:

[0025] Calculate an individual emotional risk value of an individual, where the individual emotional risk value is the sum of all emotional intensities of the individual within a set time period;

[0026] Determine whether the individual emotional risk value meets a preset strategy, and if so, identify the individual as an emotional risk individual; the preset strategy is a preset threshold or a ratio of a preset individual emotional risk value to other values.

[0027] On the other hand, the present invention also proposes a method for correcting individual emotion data applied to campus emotion risk assessment, comprising:

[0028] The group emotion data is obtained by processing the individual emotion data;

[0029] Determine the target threshold of the number of people in the group. When the number of people with a certain negative emotion in the group emotion data is greater than or equal to the target threshold of the number of people in the group, it is considered risk-free data; when the number of people with a certain negative emotion is less than the target threshold of the number of people in the group, it is considered risky data.

[0030] Determine the intensity value of the individual emotion data corresponding to the risky data according to the proportion of the group's non-negative emotion data to the total number of people;

[0031] The intensity value is used as an individual emotion data correction value for campus emotion risk assessment;

[0032] The target threshold of the number of people in the group varies in a step-by-step manner according to the total number of people.

[0033] This method is mainly used to assess individual emotional risks on campus using one-way observation methods such as physiological signal information. If the technical solution does not consider the impact of group emotions on individual emotions, this method can be used to correct individual emotional data.

[0034] The present invention also proposes a campus emotional risk individual assessment system based on expression recognition, comprising:

[0035] Recognition module: collects monitoring image data for recognition, and obtains the recognized individual expression data and related information data;

[0036] Processing module: obtain group emotion data based on individual expression data;

[0037] Correction module: Correct individual expression data according to group emotion data to obtain individual negative emotion data;

[0038] Evaluation module: Evaluate emotional risk target individuals based on historical statistics of individual negative emotion data.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] The present invention overcomes the inherent technical defect of the prior art that a single micro-expression data source cannot quantify emotions. It corrects individual emotions into quantitative data based on group non-negative emotion data, realizes the conversion of two homologous qualitative data into quantitative evaluation, and creates a precedent for quantitative evaluation of emotion intensity in the field of emotion recognition technology based on expression. Although this method is only applicable to special scenarios such as campuses and the accuracy of individual emotion quantification is not high, it has good practical value because it does not require contact for identification and realizes automatic standardized processing by the system, and can complete campus emotion risk assessment with high quality. At the same time, the results of emotion intensity quantification data can also be used for subsequent actual needs of emotional depth analysis such as emotional counseling, emotional adjustment tracking, personality analysis, and learning attention assessment.

[0041] The present invention can accurately eliminate group emotional anomalies and objectively judge individual emotional risks in group scenarios, thereby improving the accuracy of the system in identifying risky targets, thereby reducing invalid data and alarm information processing, which can effectively improve computing efficiency, and can effectively delete the original image data of group emotional anomalies and risk-free data to reduce data storage, thereby achieving the technical goal of improving hardware computing efficiency and execution effect.

[0042] In the field of emotion recognition technology based on facial expressions, the present invention overcomes the inherent technical defects of the single data source of the prior art, overcomes technical bias and introduces group emotion data into individual emotion recognition, thereby improving the accuracy of individual emotion recognition, reducing the storage volume of image data and improving system efficiency. At the same time, the technical solution is transformed from single image feature recognition to big data relationship analysis, which solves the key technical problem that emotion recognition in this technical field cannot be objectively quantified, creates a precedent for automatic quantitative assessment of emotion intensity, and achieves significant progress from qualitative to quantitative assessment. At the same time, the above data can be used to train a big data model for identifying individual emotional risks of students. The big data model can be automatically verified by humans using the test results after the application of this system, and can learn and evolve rapidly, fundamentally solving the two major problems of the source of training image data and the verification of recognition results, thereby quickly and at low cost realizing the improvement of the accuracy of individual emotion assessment in this technical field. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a schematic diagram of the process of Example 1 of the present invention. DETAILED DESCRIPTION

[0044] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0045] The design concept of the present invention is to utilize group emotions, individual expressions and the logical relationship between individual emotions to process individual expression data, thereby achieving an objective quantitative assessment of individual emotional risks.

[0046] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

[0047] Embodiment 1:

[0048] Taking "collecting student image data through surveillance cameras in the classroom to identify emotional risk individuals" as an example, this embodiment 1 is as follows Figure 1 As shown, the following steps are included:

[0049] S1, collect monitoring image data for recognition, and obtain the recognized individual expression data and related information data;

[0050] This step specifically includes:

[0051] S11: receiving image data from a surveillance camera and storing it in a pre-processing area P;

[0052] In this embodiment, the system receives image data containing human faces captured by one or more surveillance cameras in the same classroom, and the system is required to process image data of the same scene and time period each time.

[0053] The same scene is usually the same classroom. If the activities are divided into groups, the same scene is the same group.

[0054] All targets should be captured as much as possible during the same period, and the length of the period is determined based on factors such as the scene, total number of people, number of cameras and capture speed. The shorter the period, the higher the accuracy of the evaluation.

[0055] In order to ensure the integrity and timeliness of data collection, a surveillance camera patrol and capture plan is usually set in advance. For example, the monitoring equipment automatically completes the task of capturing faces of the entire classroom every 10 minutes and transmits them to the system in a centralized manner.

[0056] S12: Enter the information required for the assessment into the early warning statistical data;

[0057] The individual expression data, individual identity recognition information, captured photo name information, and camera number are entered into the early warning statistical database F.

[0058] Individual expression data can be obtained by identifying the image data through third-party software, or through the functions provided by the surveillance camera manufacturer. For example, Dahua’s face capture camera DH-IPC-HDBW4643R1-ZYL-PV-SA / AS can recognize eight expressions: anger, calm, happiness, sadness, disgust, surprise, confusion, and fear, and automatically send the expression recognition result data; at the same time, it can compare the face capture image with the face library information entered in advance, and send the recognized personal identity information data such as name and class.

[0059] Establish an early warning statistical database F to input the information required for evaluation. In the later stages of individual and group screening, evaluation, and alarm, the data in the early warning statistical database F can be directly called without reprocessing the image data, thereby reducing repeated data processing and improving system operation efficiency.

[0060] S2, obtaining group emotion data based on individual expression data;

[0061] The group emotion data in this step includes emotions and corresponding numbers of people, and the emotions include negative emotions and non-negative emotions;

[0062] This step specifically includes:

[0063] S21: Statistical analysis of group expression data;

[0064] Taking the eight individual expressions of anger, calmness, happiness, sadness, disgust, surprise, confusion, and fear captured as an example, the group expression data is a collection of the eight expressions of anger, calmness, happiness, sadness, disgust, surprise, confusion, and fear and the number of people involved. After completing step S12, the group expression data can be directly extracted from the early warning statistical database F.

[0065] If there is no data of the four negative expressions of anger, sadness, fear, and disgust, it is considered that there is no emotional risk, and the data captured this time does not need to be evaluated. All image data can be deleted and this work is ended.

[0066] S22: Processing the group expression data to obtain group emotion data;

[0067] The group expression data is processed to obtain the group emotion data, and the group emotion data is entered into the early warning statistical database F.

[0068] In this step, the processing method can adopt the common method in the industry, directly use the group expression data as the group emotion data, and the group emotion data is a collection of each type of emotion and the number of people in this type of data. For example, if the group expression data is 8 types of expressions and 3 people for each expression, then the group emotion data is also 3 angry people, 3 sad people, 3 afraid people, 3 calm people, 3 happy people, 3 disgusted people, 3 surprised people, and 3 confused people.

[0069] Of course, machine learning and other methods can also be used to complete the work of processing and identifying individual expressions into group emotions. For example, the collected individual expression data can be organized into a data set, and a neural network model can be used for data training to obtain an individual expression to group emotion recognition model, thereby making the group emotion data recognition results more accurate.

[0070] S23: Delete the four types of individual facial expression image data: calm, happy, surprised, and confused.

[0071] Data was sorted and image data of four individual expressions: calm, happy, surprised, and confused were deleted.

[0072] The purpose of this step is to reduce the amount of data that needs to be processed later. Under normal circumstances, the four individual expressions of calm, happy, surprised, and confused will not cause emotional risks and do not need to be analyzed and processed later. Therefore, the image data corresponding to these four individual expressions are deleted, and only the image data of the four individual expressions of sadness, disgust, fear, and anger are retained. This operation only deletes the image data of individual expressions and does not affect the individual expression data and the processed group emotion data.

[0073] S3, correcting individual expression data according to group emotion data to obtain individual negative emotion data;

[0074] According to the group negative emotion data in the group emotion data, it is distinguished whether the individual expression data belongs to risk-free or risky. The group non-negative emotion data is used to correct the individual expression data belonging to risk to obtain the individual negative emotion data; and the individual negative emotion quantitative data including the individual emotion intensity can be further obtained.

[0075] The eight group emotions affect each individual emotion in different ways and degrees. It is necessary to analyze and correct individual expression data one by one according to the different types of group emotion data, so as to obtain more objective and accurate individual emotion data.

[0076] When using group emotion data to correct the individual expression data, simplifying the correction method to classify and cluster group emotions can effectively improve the correction efficiency.

[0077] In this embodiment, since the emotional consistency of the student group is high, the group emotions are divided into two categories: group negative emotions (sadness, disgust, fear, anger) and group non-negative emotions (calmness, happiness, surprise, confusion).

[0078] Therefore, this step S3 specifically includes the following:

[0079] S31: Correcting the individual expression data of this time according to the group negative emotion data;

[0080] The group emotion data obtained in step S22 is used to correct the individual expression data obtained in step S12, and only the main influencing factors are considered to improve efficiency.

[0081] The emergence of negative emotions in some specific scenarios is a normal group reaction, which is harmless and does not require warning or disposal. Therefore, it is necessary to eliminate risk-free group negative emotions.

[0082] For example, when studying the lesson about heroes and martyrs’ sacrifice in class, the system will recognize negative emotions such as sadness and fear among many people in step S22.

[0083] Generally, a group is considered to be a group if the number of people is greater than or equal to 2. Taking into account the possibility that individuals may have the same emotions as the total number of people increases, it is necessary to determine the group judgment criteria based on the total number of people.

[0084] A target threshold N of the number of people in a group is set for each of the four negative emotions of sadness, disgust, fear, and anger. The target threshold N of the number of people in a group changes in a step-by-step manner according to the total number of people collected En. The more the total number of people collected En is, the larger the target threshold N of the number of people in a group is.

[0085] When the total number of collected people En is 2 to 9, the target threshold of group number N is 3;

[0086] When the total number of people collected En is between 10 and 19, the target threshold of the number of people in the group N is 4;

[0087] When the total number of collected persons En is 20 or above, the target threshold of the number of people in the group N is 5.

[0088] In addition, the target threshold N of the number of people in the group can also be adjusted according to the level of emotional consistency of the specific target group. For example, middle school students have stronger personalities and weaker emotional consistency than elementary school students, so the value of N needs to be lower.

[0089] When the system detects that the number of people with a certain negative emotion in the group emotion data is greater than or equal to the target threshold N of the group number, it determines that this emotion is risk-free data, changes all individual expression data corresponding to this type of negative emotion in the early warning statistical database F to group negative emotions, deletes the preprocessing area P this type of emotion image data, and ends the processing of this type of emotion data; when the number of people with a certain negative emotion is less than the target threshold N of the group number, it determines that all individual expression data corresponding to this emotion are risky data, triggering the early warning mechanism.

[0090] Executing the early warning mechanism includes adding the current early warning information in the early warning statistical database F and transferring the image data of the pre-processing area P to the early warning area W.

[0091] The execution of the early warning mechanism can increase the number of additional image acquisition and detection times for the individuals in the early warning information to increase the evaluation accuracy. While improving the detection effect, it also greatly reduces the overall data collection, processing and storage volume to meet the requirements of multi-target real-time monitoring and response.

[0092] S32: Correcting the individual expression data corresponding to the risky data according to the group non-negative emotion data to obtain individual negative emotion data;

[0093] In this embodiment, the risky data obtained in S31 can be directly used as individual negative emotion data, or it can be further modified. Based on the laws of nature, the group's non-negative emotions have a weakening or no effect on individual negative emotions. For example, individual anger + group happiness = individual sad expression, individual sad emotion + group calm emotion = individual sad expression, so the correction is based on individual emotion = individual expression - group non-negative emotion, and the real individual emotion is obtained. Because the individual expressions corresponding to the risky data are all negative expressions, the individual emotion data obtained are also individual negative emotion data.

[0094] After obtaining the individual negative emotion data, we can proceed to step S4 to evaluate the emotional risk individuals.

[0095] However, in order to more accurately assess the degree of emotional risk in the future, individual negative emotion quantitative data including individual emotion intensity can be further obtained. The individual negative emotion quantitative data is also obtained by evaluating the group non-negative emotions. For example, the effect of various types of group non-negative emotions on individual negative emotions can be evaluated, or the proportion of the group non-negative emotion data to the total number of people collected can be evaluated.

[0096] Since happy emotions have the greatest impact on negative emotions, in order to improve efficiency, this embodiment only uses single happy emotion data among the group's non-negative emotions, and evaluates the individual negative emotion intensity value by the proportion of happy emotions in the total number of collected people.

[0097] When the number of people with happy emotions accounts for 80% or more of the total number of valid collected emotions, the individual negative emotion intensity value is equal to 2; when the number of people with happy emotions accounts for 50% or more and less than 80% of the total number of valid collected emotions, the individual negative emotion intensity value is equal to 1.5; when the number of people with happy emotions accounts for less than 50% of the total number of valid collected emotions, the individual negative emotion intensity value is equal to 1. The total number of valid collected emotions can be directly extracted from the early warning statistical database F.

[0098] S33: Processing individual negative emotion data.

[0099] After completing the correction of individual expression data to obtain individual emotion data, the individual negative emotion data and intensity value are entered into the early warning statistical database F.

[0100] S4. Evaluate emotional risk target individuals based on historical statistics of individual negative emotion data.

[0101] This step S4 specifically includes:

[0102] S41: Assess individual targets for emotional risk;

[0103] In this embodiment, the emotional risk assessment is mainly based on the data such as the type, frequency, and intensity of individual negative emotions in the early warning statistical database F. Then, the preset identification strategy is called to determine whether the individual emotional risk value meets the preset strategy. If it does, it is identified as an emotional risk individual.

[0104] If the correction in step S32 obtains individual negative emotion data rather than individual negative emotion quantitative data, the preset strategy can be a set risk threshold for individual negative emotion data, such as 5 times a day or 4 times within 2 hours, etc. If the statistical number of individual negative emotion data of an individual exceeds 5 times a day or exceeds 4 times within 2 hours, the individual can be identified as an emotional risk individual target.

[0105] If the correction in step S32 obtains quantitative data of individual negative emotions, the preset strategy can be that when the intensity value of an individual's negative emotions is higher than or equal to 50% of the total valid expression collection times, it is assessed as an emotional risk. In the 10th valid expression collection of the individual on that day, one new fear (emotion intensity value is 2), and the individual's negative emotion history data on that day also has 2 sadness (emotion intensity values ​​are 1 and 2), then the individual emotional risk value is 1+2+2=5, and the risk value 5 has reached 50% of the valid expression collection times 10. Therefore, according to the preset strategy, the individual is identified as an emotional risk individual target.

[0106] Multi-physiological signal data can also be combined when evaluating and identifying individual targets with emotional risks. Multi-physiological signal emotion recognition technology mainly includes data analysis of common physiological signals such as electrocardiogram and respiratory signals, pulse waves, and skin conductance. Since multi-physiological signals are objective and not easily concealed, the recognition accuracy can reach 95%.

[0107] As a preference, in the process of identifying emotional risk individuals, multiple physiological signal data related to emotional judgment can be combined to improve the accuracy of target identification of emotional risk individuals;

[0108] S42: Alarm mechanism is triggered after assessing emotional risk individual targets;

[0109] The alarm mechanism includes sending the alarm information to the relevant person in charge for processing and transferring the image data and evaluation results of the warning area of ​​the risk person to the alarm area; sending the alarm information and relevant data of the alarm area to the responsible person for processing.

[0110] Preferably, the alarm mechanism also includes feedback from the person in charge on the alarm verification results so as to optimize and adjust the alarm strategy later. It can also be used as training data for the individual emotional risk big data model.

[0111] S43: After completing the emotional risk assessment of all individuals in the newly added warning data, complete the individual emotional risk assessment work and regularly delete the data in the warning area, alarm area and warning statistical database.

[0112] Preferably, one-way observation methods such as physiological signal information are used to assess individual emotional risks on campus. If the technical solution does not consider the impact of group emotions on individual emotions, the method of using group emotional data to correct individual emotional data can also be used to improve the accuracy of identifying individual emotional risk targets.

[0113] This embodiment can overcome the inherent technical defect of the existing technology that a single micro-expression data source cannot quantify emotions, and corrects individual emotions into quantitative data based on group non-negative emotion data, realizing the conversion of two homologous qualitative data into quantitative evaluation, and creating a precedent for quantitative evaluation of emotion intensity in the field of emotion recognition technology based on expression. Although this method is only applicable to special scenarios such as campuses and the accuracy of individual emotion quantification is not high, it has good practical value because it does not require contact for identification and realizes automatic standardized processing by the system, and can complete campus emotion risk assessment with high quality. At the same time, the results of emotion intensity quantification data can also be used for subsequent actual needs of emotional depth analysis such as emotional counseling, emotional adjustment tracking, personality analysis, and learning attention assessment, and the overall technical effect of business rules and method features is significantly improved.

[0114] This embodiment can accurately eliminate group emotional anomalies and objectively judge individual emotional risks in group scenarios, thereby improving the accuracy of the system in identifying risk targets, thereby reducing invalid data and alarm information processing, which can effectively improve computing efficiency, and can effectively delete original image data of group emotional anomalies to reduce data storage, thereby achieving the technical goal of improving hardware computing efficiency and execution effect.

[0115] In the field of emotion recognition technology based on facial expressions, this embodiment can overcome the inherent technical defects of the single data source of the existing technology, overcome technical bias and introduce group emotion data into individual emotion recognition, thereby improving the accuracy of individual emotion recognition, and using classification and clustering and other technologies to reduce the storage of image data and improve system efficiency. At the same time, the technical solution is transformed from single image feature recognition to big data relationship analysis, which solves the key technical problem that emotion recognition in this technical field cannot be objectively quantified, creates a precedent for automatic quantitative assessment of emotion intensity, and achieves significant progress from qualitative to quantitative assessment. At the same time, the above data can be used to train a big data model for identifying individual emotional risks of students. The test results of the big data model after application in this system are automatically verified by humans and can learn and evolve rapidly, fundamentally solving the two major problems of the source of training image data and the verification of recognition results, thereby quickly and at low cost realizing the improvement of the accuracy of individual emotion assessment in this technical field, which has extremely high social value.

[0116] Embodiment 2:

[0117] This embodiment 2 proposes a campus emotional risk individual assessment system based on expression recognition, including:

[0118] Recognition module: collects monitoring image data for recognition, and obtains the recognized individual expression data and related information data;

[0119] Processing module: obtain group emotion data based on individual expression data;

[0120] Correction module: Correct individual expression data according to group emotion data to obtain individual negative emotion data;

[0121] Evaluation module: Evaluate emotional risk target individuals based on historical statistics of individual negative emotion data.

[0122] In this system, the processing module includes:

[0123] Count the individual expression data to obtain the group expression data of each expression and the corresponding number of people, and use the group expression data as the group emotion data;

[0124] or:

[0125] Use machine learning to complete the processing and recognition of individual expression data into group emotion data.

[0126] In this system, the correction module includes:

[0127] Distinguish whether individual expression data belongs to risk-free data or risky data according to group negative emotion data in group emotion data, and obtain individual negative emotion data according to individual expression data belonging to risky data;

[0128] The method for distinguishing individual expression data includes: determining a target threshold of the number of people in a group, and when the number of people corresponding to a certain negative emotion in the group emotion data is greater than or equal to the target threshold of the number of people in the group emotion data, the group emotion is abnormal and belongs to risk-free data;

[0129] When the number of people with a certain negative emotion is less than the target threshold of the group, it is considered risky data;

[0130] and / or:

[0131] The correction modules include:

[0132] include:

[0133] The individual negative expression data is corrected according to the group non-negative emotion data in the group emotion data to obtain the individual negative emotion data, where individual negative emotion = individual negative expression - group non-negative emotion impact.

[0134] The correction module further comprises: after obtaining the individual negative emotion data, calculating the emotion intensity to obtain individual negative emotion quantitative data including the individual emotion intensity; the emotion intensity is evaluated according to the group non-negative emotion data.

[0135] In the correction module, the number of times of collecting monitoring image data is additionally increased for the individuals corresponding to the identified risky data.

[0136] In this system, the evaluation module includes: setting a risk threshold for individual negative emotion data, and identifying an individual as an emotional risk individual if the individual exceeds the risk threshold based on the statistical times and frequency of the individual negative emotion data.

[0137] The assessment module also includes: assessing emotional risk target individuals based on historical statistics of individual negative emotion data combined with multi-physiological signal data analysis.

[0138] Alternatively: Assessment modules include:

[0139] Calculate an individual emotional risk value of an individual, where the individual emotional risk value is the sum of all emotional intensities of the individual within a set time period;

[0140] Determine whether the individual emotional risk value meets a preset strategy, and if so, identify the individual as an emotional risk individual; the preset strategy is a preset threshold or a ratio of a preset individual emotional risk value to other values.

[0141] The campus emotional risk individual assessment system based on facial expression recognition proposed in this Example 2 can implement the campus emotional risk individual assessment method based on facial expression recognition described in Example 1, and has the same beneficial technical effects as the campus emotional risk individual assessment method based on facial expression recognition described in Example 1.

[0142] The above-mentioned embodiments are only preferred implementations of the present invention, and are only used to help understand the method and core ideas of the present application. The protection scope of the present invention is not limited to the above-mentioned embodiments. All technical solutions under the idea of ​​the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A method for assessing individual campus emotional risk based on facial expression recognition, characterized in that: include: S1, collect monitoring image data for recognition, and obtain the recognized individual expression data and related information data; S2, obtaining group emotion data based on individual expression data; S3, correcting individual expression data according to group emotion data to obtain individual negative emotion data; S4. Evaluate emotional risk target individuals based on historical statistics of individual negative emotion data.

2. The method for assessing individual campus emotional risk based on facial expression recognition according to claim 1, characterized in that: The method of correcting individual expression data according to group emotion data in step S3 includes: Distinguish whether individual expression data belongs to risk-free data or risky data according to group negative emotion data in group emotion data, and obtain individual negative emotion data according to individual expression data belonging to risky data; The method for distinguishing individual expression data includes: determining a target threshold of the number of people in a group, and when the number of people corresponding to a certain negative emotion in the group emotion data is greater than or equal to the target threshold of the number of people in the group emotion data, the group emotion is abnormal and belongs to risk-free data; When the number of people with a certain negative emotion is less than the target threshold of the group, it is considered risky data.

3. The method for assessing individual campus emotional risk based on facial expression recognition according to claim 1, characterized in that: The method of correcting individual expression data according to group emotion data in step S3 includes: The individual negative expression data is corrected according to the group non-negative emotion data in the group emotion data to obtain the individual negative emotion data, where individual negative emotion = individual negative expression - group non-negative emotion impact.

4. The method for assessing individual campus emotional risk based on facial expression recognition according to claim 1, characterized in that: Step S3 also includes: after obtaining the individual negative emotion data, calculating the emotion intensity to obtain individual negative emotion quantitative data including the individual emotion intensity; the emotion intensity is evaluated according to the group non-negative emotion data.

5. The method for assessing individual campus emotional risk based on facial expression recognition according to claim 2, characterized in that: In step S3, the number of times of collecting monitoring image data is additionally increased for the individuals corresponding to the identified risky data.

6. The method for assessing individual campus emotional risk based on facial expression recognition according to claim 1, characterized in that: The method for evaluating emotional risk target individuals described in step S4 includes: setting a risk threshold for individual negative emotional data, and identifying an individual as an emotional risk individual if the individual exceeds the risk threshold based on the statistical number and frequency of the individual negative emotional data.

7. The method for assessing individual campus emotional risk based on facial expression recognition according to claim 1, characterized in that: Step S4 also includes: evaluating the emotional risk target individual based on historical statistics of individual negative emotion data combined with multi-physiological signal data analysis.

8. The method for assessing individual campus emotional risk based on facial expression recognition according to claim 4, characterized in that: The method for identifying emotional risk target individuals in step S4 includes: Calculate an individual emotional risk value of an individual, where the individual emotional risk value is the sum of all emotional intensities of the individual within a set time period; Determine whether the individual emotional risk value meets a preset strategy, and if so, identify the individual as an emotional risk individual; the preset strategy is a preset threshold or a ratio of a preset individual emotional risk value to other values.

9. A method for correcting individual emotion data applied to campus emotion risk assessment, characterized in that: include: The group emotion data is obtained by processing the individual emotion data; Determine the target threshold of the number of people in the group. When the number of people corresponding to a certain negative emotion in the group emotion data is greater than or equal to the target threshold of the number of people in the group, it is considered risk-free data. When the number of people with a certain negative emotion is less than the target threshold of the group, it is considered risky data; Determine the intensity value of the individual emotion data corresponding to the risky data according to the proportion of the group's non-negative emotion data to the total number of people; The intensity value is used as an individual emotion data correction value for campus emotion risk assessment; The target threshold of the number of people in the group varies in a step-by-step manner according to the total number of people.

10. A campus emotional risk individual assessment system based on expression recognition, characterized in that: include: Recognition module: collects monitoring image data for recognition, and obtains the recognized individual expression data and related information data; Processing module: obtain group emotion data based on individual expression data; Correction module: Correct individual expression data according to group emotion data to obtain individual negative emotion data; Evaluation module: Evaluate emotional risk target individuals based on historical statistics of individual negative emotion data.

Images