A cleaning management system

Abstract

The present invention relates to a smart public toilet cleaning management system, which records the cleaning time and cleaning frequency of the cleaning staff in the cleaning state through the cleaning status in the smart public toilet, and transmits the cleaning status and the cleaning time and the cleaning frequency to the cloud platform for coordinated management, and ensures that the cleaning management method is intelligently managed through the cleaning time AI detection model. It is more convenient to judge the next cleaning time under the current conditions, reduce the hard work of the cleaning staff, and improve the sanitary conditions of the smart public toilets. Only relying on the interval of the punch-in time, it is inferred that the effective punch-in of the discrete punch-in data of the point data of the radio frequency punch-in machine in a spatial distribution area and a time distribution area is distinguished. The division of the cleaning work time period and the time period without cleaning work of the cleaner is automatically inferred.

Description

Technical Field

[0001] The present invention relates to the technical field of AI intelligent management, and in particular to a cleaning management system. Background Art

[0002] With the continuous development of urban construction, urban sanitation has become a key factor in the overall appearance of a city. Public toilets are an indispensable infrastructure in urban public places, especially in areas with large traffic such as cultural attractions and parks. The density of public toilets is relatively high. The sanitation of public toilets is directly related to the cultural appearance of the street area and even the entire city, as well as the level of civilization of the city.

[0003] Public toilets are an important assessment and identification indicator of a city's sanitation. The cleaning of existing public toilets is maintained by dedicated cleaning staff, and the management of cleaning staff is mainly manual, that is, cleaning staff are arranged to clean and inspect the sanitation of public toilets at regular intervals, such as cleaning every three hours. At the same time, the public toilets are divided into areas according to the number of cleaning staff, and the cleaning staff will maintain and clean the cleaning areas they are responsible for. At the same time, all public toilets in the cleaning area are inspected and monitored at regular intervals, and then judged whether the public toilets need cleaning and maintenance based on their own experience.

[0004] If there is a sudden flow of people within the specified time interval, such as in scenic spots and highway service areas, several buses stopping at the same time may cause the public toilets to be full. Cleaning at the current frequency cannot guarantee the cleanliness of the public toilets. And with the increasing number of public toilets year by year, the management of public toilet cleaning staff has become more and more prominent. Cleaning staff often cannot clean public toilets in the first time. They cannot use corresponding cleaning methods to clean the cleaning status of public toilets based on experience. For example, cleaning staff cannot make accurate judgments on the cleaning status indicators of public toilets such as the concentration of ammonia and H2S in the toilets based on their own experience, resulting in incomplete cleaning of public toilets. Cleaning staff cannot deal with the sanitation and cleaning status of public toilets in a targeted manner, resulting in generally low labor efficiency and prominent problems of dirty and messy toilets. These problems seriously restrict the development of the urban environment and increase the maintenance cost of public toilets. It is also impossible to verify whether the cleaning staff has cleaned the toilets. At the same time, special treatment for some special situations cannot be efficient. Summary of the invention

[0005] In order to solve the problem that the above-mentioned existing cleaning methods cannot be effectively supervised, the present invention provides an intelligent public toilet cleaning management system, including a cloud platform and a data acquisition module, wherein the cloud platform is communicatively connected with the data acquisition module;

[0006] The data acquisition module includes a multi-cooperative data terminal, which is used to obtain the cleaning status information in the smart public toilet and record the cleaning time and cleaning frequency of the cleaning staff, and transmit the cleaning status information and the cleaning time and cleaning frequency to the cloud platform;

[0007] The cloud platform includes an AI processing module and a communication module. The AI ​​processing module is used to receive the cleaning status information, input the cleaning status information into a pre-trained cleaning time AI detection model, obtain the cleaning information of the smart public toilet, and transmit the cleaning information to the multi-cooperative data terminal;

[0008] The communication module is communicatively connected to the AI ​​processing module, and the communication module is used to receive the cleaning information and push the cleaning information to the cleaning staff.

[0009] Based on the above technical solution of the present invention, the present invention can also make the following improvements.

[0010] Furthermore, the cleaning status is input into a pre-trained cleaning time AI detection model to obtain the cleaning staff scheduling information of the smart public toilet:

[0011] S1. Acquire the cleaning status of the public toilet to be inspected, wherein the cleaning status includes a plurality of cleaning status parameters, and perform elimination processing on the plurality of cleaning status parameters, eliminate cleaning status parameter values ​​in an abnormal state, and obtain a cleaning status data set;

[0012] S2. Input the cleaning status data set into a pre-trained cleaning time AI detection model for processing to obtain the cleaning staff scheduling information of the smart public toilet.

[0013] Further, training the cleaning time AI detection model includes the following steps:

[0014] S201, obtaining historical cleaning information of the public toilet to be detected, and preprocessing the historical cleaning information to obtain a matrix of weights and bias values ​​of multiple input parameters;

[0015] S202, constructing an initial cleaning time Al detection model based on a back propagation algorithm, and setting initial expected target parameters;

[0016] S203, iteratively training the initial cleaning time AI detection model according to the matrix of weights and bias values ​​of multiple input parameters, and obtaining the cleaning status classification probability of the public toilet according to multiple preset cleaning status classifications;

[0017] S204, calculating the standard deviation according to the initial expected target parameter and the cleaning state classification probability, when the standard deviation abs=(r 2 -x2 ) is less than the first threshold, the training of the initial cleaning time AI detection model is completed to obtain the cleaning time AI detection model, where r is the cleaning state classification probability, and x is the initial expected target parameter.

[0018] Further, in step S201, the historical cleaning status information is preprocessed to obtain a matrix of weights and bias values ​​of multiple input parameters, including: normalizing the cleaning status information according to history, and presetting the initial values ​​of all the input parameters to random values, and then inputting the matrix of weights and bias values ​​of all the input parameters.

[0019] Furthermore, the multiple cleaning status classifications are preset with cleaning times corresponding thereto, and the cleaning staff scheduling information of the smart public toilet is obtained according to the cleaning status classification probability.

[0020] Furthermore, the multi-cooperative data terminal records the cleaning time and cleaning frequency of the cleaning staff including the following steps:

[0021] S001: Obtain the tag card information of the cleaning staff within the preset punch-in range, the preset range is 1m to 5m, the tag card information includes the tag card number, time information and location information; record the time information as T1, T2, ...T n , the location information is recorded as D1, D2, ...D n , where n is the number of times the tag card information of the same tag card number is obtained, and the process goes to step S002;

[0022] S002: If the current time information T n Equal to the next time information T n+1 When, and the current position information D n The next time the position information D n+1 If the distance does not exceed the preset punch-in range, exit; if 30S≤Tn +1 -T n ≤10800S, and the current position information D n The last position information D n-1 If the distance does not exceed the preset clock-in range, the process proceeds to step S003;

[0023] S003: The multi-cooperative data terminal records the increase in the number of effective cleaning times and sends the next time information T n+1 Synchronize to the current time information T n , the next time the position information D n+1 Update to the current location information D n .

[0024] Furthermore, the S002 further includes: if 0<T n+1-T n When the time is ≤30S, it is determined that the cleaning staff has continuously clocked in within the preset range, and the process ends.

[0025] Furthermore, the cleaning staff scheduling information is pushed to the cleaning staff as follows: after the communication module receives the cleaning staff scheduling information, the cleaning staff scheduling information is pushed to the cleaning staff's handheld terminal and the multi-cooperative data terminal, the multi-cooperative data terminal includes a display and a broadcast module, the multi-cooperative data terminal displays the received pushed cleaning staff scheduling information on the display, and prompts the cleaning staff of the cleaning information through the broadcast module.

[0026] Furthermore, the multiple cleaning status parameters include ammonia concentration value, temperature value, humidity value, PM2.5 value, hydrogen sulfide concentration value and cleaning punch-in time. The multiple cleaning parameters are eliminated as follows: any of the cleaning status parameters is respectively provided with a target expected value corresponding to it. When the cleaning status parameter is significantly different from the corresponding target expected value, it is determined that the cleaning status parameter is in an abnormal state, and the cleaning status parameter value in the abnormal state is eliminated to obtain a cleaning status data set.

[0027] Furthermore, the communication module also includes an output control unit, which is used to screen and judge the received cleaning information based on data history records and cleaning status information, and after inputting the abnormal cleaning status information into a pre-trained cleaning time AI detection model, the cleaning information of the smart public toilet is deleted.

[0028] The beneficial effect of the present invention is that after a certain period of accumulation, AI calculations are performed on the cleaning history data and the environmental sensor history data of the cleaning module. As the data accumulates, it will become more and more accurate, and it will be easy to judge the next cleaning time under the current conditions, reduce the hard work of cleaning staff, and improve the sanitary conditions of smart public toilets. Only relying on the interval of the punch-in time, it is inferred that the effective punch-in of the discrete punch-in data of the point data of the radio frequency punch-in machine in a spatial distribution area and a time distribution area is distinguished. The division of the cleaning work time period and the non-cleaning work time period of the cleaner is automatically inferred. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 This is a system block diagram of an intelligent public toilet cleaning management system of the present invention;

[0030] Figure 2 It is a flow chart of a specific method for determining the effective cleaning times of the present invention. DETAILED DESCRIPTION

[0031] The principles and features of the present invention are described below in conjunction with the accompanying drawings. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0032] like Figure 1 As shown, the present invention provides an intelligent public toilet cleaning management system, characterized in that it includes a cloud platform and a data acquisition module, and the cloud platform is respectively communicated with the data acquisition module and a plurality of intelligent public toilets; the data acquisition module includes a multi-cooperative data terminal, and the multi-cooperative data terminal is used to obtain the cleaning status information in the intelligent public toilet and record the cleaning time and cleaning frequency of the cleaning staff, and transmit the cleaning status information and the cleaning time and cleaning frequency to the cloud platform; the cloud platform includes an AI processing module and a communication module, and the AI ​​processing module is used to receive the cleaning status information, and input the cleaning status information into a pre-trained cleaning time AI detection model to obtain the cleaning information of the intelligent public toilet, and transmit the cleaning information to the multi-cooperative data terminal; the communication module is communicatively connected to the AI ​​processing module, and the communication module is used to receive the cleaning information and push the cleaning information to the cleaning staff.

[0033] The beneficial effect of the present invention is that after a certain period of accumulation, AI calculations are performed on the cleaning history data and the environmental sensor history data of the cleaning module. As the data accumulates, it will become more and more accurate, and it will be easy to judge the next cleaning time under the current conditions, reduce the hard work of cleaning staff, and improve the sanitary conditions of smart public toilets. Only relying on the interval of the punch-in time, it is inferred that the effective punch-in of the discrete punch-in data of the point data of the radio frequency punch-in machine in a spatial distribution area and a time distribution area is distinguished. The division of the cleaning work time period and the non-cleaning work time period of the cleaner is automatically inferred.

[0034] Specifically in this embodiment, during the back propagation algorithm, the gradient of the loss function with respect to each parameter can be obtained by the chain rule, and in practical applications, the gradient is obtained by the automatic differentiation algorithm. Based on the characteristics of the error back propagation in the network, an optimizer is selected to update the model parameters so that the gap between the predicted value and the preset target parameter is continuously approached, and finally the optimal value of the objective function is obtained to complete the network training. Through the continuous accumulation of data by the trained cleaning time AI detection model, the parameters in the model are continuously optimized and iterated, so that the cleaning time obtained will become more and more accurate, and it is easy to judge the next cleaning time under the current conditions.

[0035] First, the cleaning status information of the public toilet to be inspected is obtained through the data input by the comprehensive gas sensor (Xiamen Zhongwu Zhilian's all-in-one Iora sensor or distributed independent gas sensor), such as ammonia concentration, temperature, humidity, PM2.5, human flow sensor data, weather information, cleaning records, and cleaning time data for each time. The cleaning status information includes multiple cleaning status parameter values. The system extracts sensor data from the database of the cloud platform to obtain historical data values ​​such as ammonia concentration, temperature, humidity, PM2.5, H2S, human flow sensor data, weather information, and clock-in data for a period of time, and performs multiple cleaning status parameter values ​​according to the data history records. Elimination processing, eliminating the cleaning status parameter values ​​that are in an abnormal state, obtaining the cleaning status parameter values, comprehensively judging the data and sensor status of the data history records extracted from the database, the multiple cleaning status parameters include ammonia concentration value, temperature value, humidity value, PM2.5 value, hydrogen sulfide concentration value and cleaning punch-in time, and the elimination processing of the multiple cleaning parameters is specifically that any of the cleaning status parameters is respectively provided with a target expected value corresponding to it, when the cleaning status parameter differs greatly from the corresponding target expected value, it is determined that the cleaning status parameter is in an abnormal state, and the cleaning status parameter values ​​in an abnormal state are eliminated to obtain a cleaning status data set. The specific setting of the target expected value in the present invention is not used to limit the scope of the present invention, and any technical solution that determines whether the cleaning status parameter is in an abnormal state by setting the target expected value and screens the cleaning status parameters in the abnormal state should fall within the protection scope of the present invention.

[0036] Specifically in this embodiment, when the flow sensor obtains flow data of more than 6,000 people per hour, which is more than ten times the target expected value of 600 people per hour corresponding to the flow data, it is determined that the flow data in the current public toilet is in an abnormal state. After the flow data is removed, when the cleaning state parameter predicts the cleaning time of the current public toilet through the cleaning time AI detection model, the flow data in the abnormal state does not affect the cleaning time, so that the cleaning time AI detection model predicts the cleaning time more accurately. When the temperature and humidity values ​​obtained by the temperature and humidity sensor continuously change and exceed the preset temperature and humidity target expected values, it is determined that the temperature and humidity data in the current public toilet is in an abnormal state. For example, when the temperature exceeds 80 degrees Celsius or is lower than -50 degrees, the temperature value data is considered to be abnormal temperature data. When the PM2.5 sensor and the H2S sensor obtain the concentration of fine particles and the concentration of H2S in the air close to the full scale, it is determined that the PM2.5 data and H2S data of the current public toilet are in an abnormal state. When the weather information acquisition fails or other reasons cause the uploaded weather data to be abnormal, the weather information in abnormal state is eliminated. For example: during the time period when the human flow sensor is normal or the squatting position sensor is normal, and the ammonia sensor and water meter usage are also normal, the extracted data is considered valid and put into the corresponding array as input data. For abnormal data, such as data of abnormal state of sensors with particularly large human flow, or data that cannot match each other after combining data of other normal sensors, they are eliminated. Obtaining effective cleaning state parameter values ​​can effectively avoid the abnormal state of cleaning state information obtained by the multi-cooperative data terminal in actual work due to force majeure or other factors. For example, when obtaining human flow data, mosquitoes may be active within the sensing range of the multi-cooperative data terminal, constantly triggering the module for obtaining human flow data in the multi-cooperative data terminal, resulting in the obtained human flow data value being too large and abnormal. This leads to inaccurate calculation of recognition time.

[0037] Specifically, in this embodiment, a data set is constructed based on normal cleaning status parameter values, and the data set constructed by the cleaning status parameter values ​​is input into the cleaning time AI detection model for processing to obtain the cleaning time required for the intelligent public toilet. Training the cleaning time AI detection model includes the following steps: obtaining the historical cleaning status information of the public toilet to be detected, preprocessing the historical cleaning status information to obtain a matrix of weights and bias values ​​of multiple input parameters, normalizing the cleaning status information according to history, and after the initial values ​​of all the input parameters are preset to random values, inputting the matrix of weights and bias values ​​of all the input parameters. Considering the characteristics of local longitudinal correlation of data, the weight parameter is modified to longitudinal correlation, and a random matrix of weights and bias values ​​of each input parameter is used. For example, [w00, w10, w20, w30, ..., wM0, b0] (random weights for severe ammonia pollution), [w01, w11, w21, w31, ..., wM1, b1] (random weights for mild ammonia pollution), [w02, w12, w22, w32, ..., wM2, b2] (random weights for severe H2S pollution), and [w03, w13, w23, w33, ..., wM3, b3] (random weights for mild H2S pollution).

[0038] Since it is necessary to comprehensively consider the values ​​of various cleaning status parameters to determine whether a public toilet currently needs cleaning work and the duration of the cleaning work, after constructing the initial cleaning time AI detection model based on the back propagation algorithm, the expected target parameters that need to be achieved by the various cleaning status parameter values ​​are set, such as the standard that the average concentration of ammonia does not exceed 1.5 mg / m3. The average concentration of H2S does not exceed 0.06 mg / m3. However, the determination of the sanitary status of a public toilet does not determine whether the sanitary status of the public toilet needs cleaning or not when any one or more cleaning status parameter values ​​reach the preset expected target parameters. Instead, it is necessary to comprehensively consider all cleaning status parameter values ​​to determine the sanitary status of the public toilet and the required cleaning time. For this reason, the present invention will modify the weight parameters to vertical correlation, and use a random matrix of weights and bias values ​​of each input parameter to achieve the above purpose.

[0039] Finally, the initial cleaning time AI detection model is iteratively trained according to the matrix of weights and bias values ​​of multiple input parameters, and the cleaning status classification probability of the public toilet is calculated according to the preset multiple cleaning status classifications, and the input data and random weight parameters and bias results are obtained: c[i, j] = a[i, j] * w[k, j], i∈(0, 7), j∈(0, M), k∈(0, 3).

[0040] Weight sharing is used vertically for the 7 sets of input data. The results are normalized by sigmoid((∑i∑j(c[i,j]+b[k,j+1])) / 7), i∈(0,7), j∈(0,M), k∈(0,3).

[0041] Sigmoid(x)=1 / (1+e^-x) is to convert any value into the range of 0 to 1. Then the cross entropy calculation softMax(y) is performed for the weight combination of the four categories (random weight for severe ammonia pollution), (random weight for mild ammonia pollution), (random weight for severe H2S pollution), and (random weight for mild H2S pollution). SoftMax=e^y / ∑i(e^yi), i∈(0,3), where yi is the probability calculated by the contribution of the output of all categories to the parameters of all classes. The sum of the probabilities calculated by the information entropy of the parameters of all classes is 1. The Softmax layer is a special fully connected layer that realizes the classification prediction of the network, and the activation function used is the softmax function. The activity value of the previous neural network layer is used as the input of the softmax layer, and the output is the class score. The normalized exponential function is used to represent the probability distribution of different categories to which the current batch of samples belongs, and the output with the highest probability is taken as the final prediction. The Softmax layer achieves the purpose of mapping the learned feature representation to the sample label space. The Softmax layer is a special fully connected layer that realizes the classification prediction of the network. The activation function used is the softmax function. The activity value of the previous neural network layer is used as the input of the softmax layer, and the output is the class score. The normalized exponential function is used to represent the probability distribution of different categories of the current batch of samples, and the output with the highest probability is taken as the final prediction. The Softmax layer realizes the purpose of mapping the learned feature representation to the sample label space.

[0042] The multiple cleaning status classifications are preset with cleaning times corresponding to them one by one, and the cleaning staff scheduling information of the smart public toilet is obtained according to the cleaning status classification probability. According to the weight parameters of the longitudinal association corresponding to the above cleaning status classification, the cleaning staff scheduling information is obtained using a random matrix of weights and bias values ​​of each input parameter. For example, when obtaining the cleaning time point of the cleaning staff, the cleaning time and cleaning times of the cleaning staff in the cleaning status obtained by the multi-cooperative data terminal are the records of the cleaning staff's punch-in, which is the time point sequence of the cleaning time. After repeatedly tuning the parameters of the cleaning time AI detection model and adding other existing parameters, the scheduling information of the cleaning staff is calculated. For example, when calculating the probability range of the current classification belonging to severe pollution, the current cleaning time point distribution is [0, 1, 0, 0, 0, 1, 1, 0, 0, 0, 0…], which is divided according to the actual working time period (starting the calculation from 0:00 to 8:00 p.m. every day), and a point in the time point sequence is obtained every 15 minutes. The number 0 in the time point sequence indicates that the cleaning status of the public toilet at the current time point does not require cleaning maintenance, and the number 1 indicates that the cleaning status of the public toilet at the current time point requires cleaning maintenance. The scheduling information of the cleaning staff is obtained by processing the matrix of the input weights and bias values ​​through sigmoid().

[0043] The classification result is compared with the target test value, and the standard deviation is calculated according to the initial expected target parameter and the cleaning state classification probability. When the standard deviation abs=(r 2 -X 2 ) is less than the first threshold (0.05), the initial cleaning time AI detection model training is completed to obtain the cleaning time AI detection model, where r is the cleaning state classification probability and x is the initial expected target parameter. Calculate the square of the difference between the classification result and the test target value. When the value is close to 1, the weights and biases of the corresponding larger items are modified. And the modification is made from the last layer forward. The modification relies on the principle of obtaining the expected variance gradient descent (derivative descent or dy / dx descent) (dErr / dw = (Err2-Err1) / (w2-w1) = expected variance change rate / parameter change rate, so that it becomes the smallest, that is, to find the ratio of the expected difference to the relevant parameter difference, and select the cross entropy loss function. The cross entropy between the true value and the predicted value represents the distance between the actual output and the expected output. The smaller the cross entropy value, the closer the two are. The objective function is used to guide network training. The purpose is to minimize the ratio of the error change rate to the parameter change rate, which means the closest to the expected target value.

[0044] Specifically, in this embodiment, the weight parameters and biases recorded and the current sensor data are used for calculation, and updated to the cleaning check-in time point of 0 or 1. For example, the current average concentration trend of ammonia or H2S at the next moment can be obtained to know whether the conditions for cleaning are met. If H2S or ammonia exceeds the national secondary emission standard or the set value after calculation under the current other parameters, the message of needing cleaning will be pushed to the APP, WeChat applet, or sent to the mobile phone by text message, and displayed on the local large screen and cloud work order display.

[0045] In this embodiment, if Figure 2 As shown, the system records the cleaning time and cleaning frequency of the cleaning staff through the multi-cooperative data terminal, obtains the tag card information of the cleaning staff within the preset punch-in range of 1m to 5m, and the tag card information of the cleaning staff includes the tag card number, time information and location information; and records the time information as T1, T2, ...T n , the location information is recorded as D1, D2, ...D n , where n is the number of times the tag card information of the same tag card number is obtained. If the current time information T n Equal to the next time information T n+1 When, and the current position information D n The next time the position information D n+1 If the distance does not exceed the preset punch-in range, exit; 30S≤T n+1 -T n ≤10800S, and the current position information D n The last position information D n-1 When the distance does not exceed the preset clock-in range, the multi-cooperative data terminal records it as the effective cleaning times and sends the next time information T n+1 Synchronize to the current time information T n , the next time the position information D n+1 Update to the current location information D n , if 0<T n+1 -T n When the number of times of cleaning is less than or equal to 30 seconds, it is determined that the cleaning staff has continuously clocked in within the preset range, and the process ends. The present invention can accurately monitor whether the cleaning staff has effectively cleaned the public toilets through the above-mentioned method for determining the number of effective cleaning times, thereby accurately coordinating and managing the cleaning status of the public toilets, reducing the maintenance cost of the sanitation status of the public toilets, and improving labor efficiency.

[0046] Specifically in this embodiment, after the communication module receives the cleaning information, the cleaning information is pushed to the handheld terminal of the cleaning staff and the multi-cooperative data terminal. The multi-cooperative data terminal includes a display and a broadcast module. The multi-cooperative data terminal displays the received pushed cleaning information on the display, and prompts the cleaning staff of the cleaning information through the broadcast module. The communication module also includes an output control unit, which is used to judge the received cleaning information based on the data history records and the cleaning status information, and after the abnormal cleaning status information is input into the pre-trained cleaning time AI detection model, the cleaning information of the smart public toilet is deleted. Set the controllable mode of the target to be achieved, and output the parameters that need to be considered to be controlled by integrating the current other uncontrollable sensor parameter data. That is, output the parameters of cleaning (punch-in time) to avoid the multi-cooperative data terminal from pushing inaccurate calculation and recognition time to the cleaning staff in actual work due to force majeure or other factors.

[0047] The reader should understand that in the description of this specification, the reference terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" and the like are intended to mean that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine different embodiments or examples described in this specification and the features of different embodiments or examples, without contradiction.

[0048] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the method embodiments described above are only illustrative, for example, the division of steps is only a logical function division, and there may be other division methods in actual implementation, such as multiple steps can be combined or integrated into another step, or some features can be ignored or not executed.

[0049] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. An intelligent public toilet cleaning management system, characterized in that: It includes a cloud platform, a plurality of intelligent public toilets and a multi-cooperative data terminal, wherein the cloud platform is communicatively connected with the multi-cooperative data terminal; The multi-cooperative data terminal is used to obtain the cleaning status of the smart public toilet and record the cleaning time and cleaning frequency of the cleaning staff in the cleaning status, and transmit the cleaning status, the cleaning time and the cleaning frequency to the cloud platform for coordinated management; The cloud platform includes an AI processing module and a communication module. The AI ​​processing module is used to receive the cleaning status and input the cleaning status into a pre-trained cleaning time AI detection model to obtain the cleaning staff scheduling information of the smart public toilet, and transmit the cleaning staff scheduling information to the communication module; The communication module is in communication connection with the AI ​​processing module, and the communication module is used to receive the cleaning staff scheduling information and push the cleaning staff scheduling information to the cleaning staff; Among them, the cleaning status is input into the pre-trained cleaning time AI detection model, and the cleaning staff scheduling information of the smart public toilet is obtained as follows: S1. Acquire the cleaning status of the public toilet to be inspected, wherein the cleaning status includes a plurality of cleaning status parameters, and perform elimination processing on the plurality of cleaning status parameters, eliminate cleaning status parameter values ​​in an abnormal state, and obtain a cleaning status data set; S2. Input the cleaning status data set into a pre-trained cleaning time AI detection model for processing to obtain the cleaning staff scheduling information of the smart public toilet; Wherein, training the cleaning time AI detection model in S2 comprises the following steps: S201, obtaining historical cleaning information of the public toilet to be detected, and preprocessing the historical cleaning information to obtain a matrix of weights and bias values ​​of multiple input parameters; S202, constructing an initial cleaning time AI detection model based on the back propagation algorithm, and setting initial expected target parameters; S203, iteratively training the initial cleaning time AI detection model according to the matrix of weights and bias values ​​of multiple input parameters, and obtaining the cleaning status classification probability of the public toilet according to multiple preset cleaning status classifications; S204, calculating the standard deviation according to the initial expected target parameter and the cleaning state classification probability, when the standard deviation abs=(r 2 –x 2 ) is less than the first threshold, the initial cleaning time AI detection model training is completed to obtain the cleaning time AI detection model, where r is the cleaning state classification probability, and x is the initial expected target parameter; Wherein, in step S201, preprocessing the historical cleaning status information to obtain a matrix of weights and bias values ​​of multiple input parameters includes: normalizing the cleaning status information according to history, and after presetting the initial values ​​of all the input parameters to random values, inputting the matrix of weights and bias values ​​of all the input parameters; Among them, the multiple cleaning status classifications include severe ammonia pollution, mild ammonia pollution, severe H2S pollution and mild H2S pollution. The multiple cleaning status classifications are provided with corresponding cleaning times, and the cleaning staff scheduling information of the smart public toilet is obtained according to the cleaning status classification probability.

2. The intelligent public toilet cleaning management system according to claim 1 is characterized in that: The multi-cooperative data terminal records the cleaning time and cleaning frequency of the cleaning staff, including the following steps: S001: Obtain the tag card information of the cleaning staff within the preset punch-in range, the preset range is 1m to 5m, the tag card information includes the tag card number, time information and location information; record the time information as T1, T2, ...Tn, and record the location information as D1, D2, ...Dn, where n is the number of times the tag card information of the same tag card number is obtained, and enter step S002; S002: If the current time information Tn is equal to the next time information Tn+1, and the distance between the current location information Dn and the next location information Dn+1 does not exceed the preset punch-in range, then exit; if 30S≤Tn+1-Tn≤10800S, and the distance between the current location information Dn and the previous location information Dn-1 does not exceed the preset punch-in range, then enter step S003; S003: The multi-cooperative data terminal records the increase in the number of effective cleaning times, and synchronizes the next time information Tn+1 to the current time information Tn, and updates the next location information Dn+1 to the current location information Dn.

3. According to claim 2, the intelligent public toilet cleaning management system is characterized in that: The S002 also includes: if 0<Tn+1-Tn≤30S, it is determined that the cleaning staff continuously clocks in within the preset range, and the process ends.

4. The intelligent public toilet cleaning management system according to any one of claims 1 to 3, characterized in that: Pushing the cleaning staff scheduling information to the cleaning staff is specifically as follows: after the communication module receives the cleaning staff scheduling information, the cleaning staff scheduling information is pushed to the cleaning staff's handheld terminal and the multi-cooperative data terminal, the multi-cooperative data terminal includes a display and a broadcast module, the multi-cooperative data terminal displays the received pushed cleaning staff scheduling information on the display, and prompts the cleaning staff scheduling information through the broadcast module.

Images