An image recognition method for operating specifications of catering industry staff

Abstract

The present invention discloses an image recognition method for the work specifications of catering industry staff, which relates to the field of image recognition technology. It solves the technical problem that the existing method cannot identify the identities of different staff members, which leads to excessive false detections. The present invention includes the steps of: human body detection and tracking, human body attribute recognition, judgment of work specification constraint objects, and recognition of non-standard behaviors of objects judged to be work specification constraint objects. This method simultaneously recognizes the identities of staff members and work specification behaviors. The work specification behavior recognition can not only identify non-standard clothing, but also identify non-standard behaviors such as overtime rest. It has a high recognition accuracy, effectively avoids false detections, and has a wide range of applications and market promotion value.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition in the catering industry, and in particular to an image recognition method for operating specifications of catering industry workers. Background Art

[0002] In the restaurant industry, companies often need to regulate and monitor employee behavior to improve service quality and operational compliance. For example, dress codes in kitchens and dining areas require kitchen staff to wear aprons, sanitary caps, and masks; and staff are not allowed to take excessive breaks in rest areas. In recent years, with the widespread implementation of artificial intelligence algorithms, surveillance cameras have been installed in all areas of restaurants. Using image recognition to detect non-compliant behavior by restaurant personnel has become widespread and widely used.

[0003] However, existing methods fail to address the problem of identifying different staff members, such as kitchen staff. Because different staff members have different supervisory and management requirements, a unified standard of management cannot be applied. Therefore, a method is urgently needed to identify restaurant staff and accurately determine the operating standards of the identified staff members. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide an image recognition method for the working specifications of catering industry staff in view of the above-mentioned defects of the prior art.

[0005] The technical solution adopted by the present invention to solve the technical problem is: an image recognition method for the working specifications of catering industry staff, characterized by comprising the following steps:

[0006] S10. In the dining space, perform human body detection and tracking on the acquired multiple frames of images to be identified, and obtain a number corresponding to each person detected and tracked;

[0007] S20. During the process from the appearance to the disappearance of the person corresponding to each number in the dining space, record the number of people corresponding to each number in each frame, the length of stay corresponding to each number, the human attribute status corresponding to each number in each frame, and the number of frames corresponding to the human attribute status;

[0008] S30, when the person corresponding to each number disappears from the dining space, judging whether the person corresponding to each number is an operation-constrained object based on the number of frames corresponding to the human attribute state of the number, the length of stay corresponding to the number, and the total number of frames of the image to be identified;

[0009] S40. When each operation-constrained object is judged to have disappeared from the dining space, the irregular operation behavior of each operation-constrained object corresponding to the number is judged and recorded according to the number of people corresponding to the number in each frame, the length of stay corresponding to the number, the number of frames corresponding to the human attribute state of the number, and the total number of frames of the image to be identified.

[0010] Furthermore, step S30 includes the following steps:

[0011] When each person corresponding to the number disappears from the dining space, the length of stay of each numbered non-operational specification constraint object is calculated; all persons corresponding to the numbers whose length of stay of the non-operational specification constraint object is not less than a first threshold are determined as the operation specification constraint objects; wherein the operation specification constraint object is a chef, and the dining space is a kitchen.

[0012] Furthermore, the calculation formula for the stay time T of each numbered non-operational specification constraint object is:

[0013] T=(t2-t1)×(count-count1) / count;

[0014] Among them, t2-t1 is the dwell time corresponding to each number; count is the total number of frames of the image to be identified; count1 is the number of image frames corresponding to the first attribute state of the human body attribute state for each number; the first attribute state is wearing an apron or a sanitary cap.

[0015] Furthermore, when the person corresponding to each number passes through the non-operation specification constraint object and is not confirmed as the operation specification constraint object, step S30 further includes the following steps:

[0016] When the person corresponding to each number disappears from the dining space, the ratio of the number of image frames corresponding to the first attribute state of each number to the total number of frames of the image to be identified is calculated; and the people corresponding to all the numbers whose ratio is not less than a preset ratio are confirmed as the objects of the operation specification constraints.

[0017] Furthermore, step S40 includes the following steps:

[0018] When it is determined that the operation constraint object disappears from the dining space, the cumulative stay time of each operation specification constraint object is calculated; each operation specification constraint object whose cumulative stay time is not less than the second threshold is confirmed as having non-standard operation behavior and recorded, and the process returns to step S10 to perform image recognition for the next batch.

[0019] Furthermore, the calculation formula for the cumulative stay time T1 of each operation specification constraint object is as follows:

[0020] T1 = (t4 - t3)×(count - count2) / count;

[0021] Where, t4 - t3 is the residence duration corresponding to the number of each object constrained by the operation specification; count is the total number of frames of the image to be recognized; count2 is the number of frames of the image in which the human body attribute status corresponding to the number of each object constrained by the operation specification simultaneously meets the first attribute status and the second attribute status; the second attribute status includes not wearing an apron, not wearing a sanitary cap, and not wearing a mask.

[0022] Further, step S40 further includes the following steps:

[0023] When it is determined that the operation constraint object disappears from the dining space, calculate the non-compliance residence duration of each object constrained by the operation specification; determine each object constrained by the operation specification whose non-compliance residence duration meets the preset conditions as operation non-compliance and record it, return to step S10, and perform image recognition for the next batch; where, the dining space is the rest area.

[0024] Further, the calculation formula of the non-compliance residence duration DT of each object constrained by the operation specification is as follows:

[0025] DT = (t6 - t5)×(1 - r);

[0026] r = num / count3;

[0027] Where, num is the number of frames of the image in which the number of people corresponding to the number of each object constrained by the operation specification in each frame is greater than C; count3 is the number of frames of the image corresponding to the number of each operation constraint object; C is the minimum number threshold for normal residence in the dining space; t6 - t5 is the residence duration corresponding to the number of each operation constraint object.

[0028] Further, the preset condition is:

[0029] The non-compliance residence duration is not less than the third threshold, and r < R; where, R is the threshold of the ratio of the total residence time of the operation constraint object with normal residence in the dining space to the collective gathering time.

[0030] Implementing the technical solution of the image recognition method for the operation specification of catering industry staff has the following advantages or beneficial effects:

[0031] This invention proposes a method for identifying different staff members and for efficiently and accurately identifying the standardized work behaviors of identified staff members. This method simultaneously identifies both the identities of staff members and their standardized work behaviors, achieving high recognition accuracy and effectively preventing false positives. In particular, it effectively avoids false positives caused by ID number swapping in human tracking algorithms. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive work. In the drawings:

[0033] Figure 1 The present invention is a flowchart of an image recognition method for the working standards of catering industry staff according to an embodiment of the present invention. DETAILED DESCRIPTION

[0034] In order to make the objects, technical solutions and advantages of the present invention clearer, the various exemplary embodiments to be described below will refer to the corresponding drawings, which constitute a part of the exemplary embodiments, in which various exemplary embodiments that may be used to implement the present invention are described. Unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementation methods described in the following exemplary embodiments do not represent all implementation methods consistent with the present disclosure. It should be understood that they are only examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims, and other embodiments may be used, or structural and functional modifications may be made to the embodiments listed herein without departing from the scope and spirit of the present invention. In other cases, detailed descriptions of well-known systems, devices, circuits and methods are omitted to avoid unnecessary details that hinder the description of the present invention.

[0035] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "thickness", "up and down, front and back, left and right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise" and "counterclockwise" indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings. They are only for the convenience of describing the present invention and simplifying the description, and are not intended to indicate or imply that the components or plug-ins referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, they cannot be understood as limitations on the present invention. In addition, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, the defined "first" and "second" features may explicitly or implicitly include one or more of the said features. In the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined. It should be noted that, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood broadly. For example, they may refer to fixed, removable, or integral connections; mechanical, electrical, or communicative connections; direct or through an intermediate medium; and internal communication between two components or interaction between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0036] Example 1:

[0037] like Figure 1 As shown, an embodiment of the present invention provides an image recognition method for catering industry staff work specifications, including the following steps:

[0038] S10. In the dining space, perform human body detection and tracking on the acquired multiple frames of images to be identified, and obtain a number corresponding to each person detected and tracked;

[0039] In this embodiment, the catering space includes but is not limited to the kitchen, dining room, and staff rest area. Video frames can be captured from cameras installed in the catering space, with each frame corresponding to a still image to be identified. Human body detection and tracking are then performed on each of the multiple frames of imagery to be identified. Human body detection obtains all human body frames in each image and can be implemented using any target detection algorithm, such as YOLO, SSD, or CenterNet. Human body tracking associates the human body frames detected in the preceding and subsequent frames, maintaining a corresponding ID number for each individual, from their appearance in the catering space until their disappearance. Common human body tracking algorithms, such as sort and deepsort, can be used. It should be noted that this step extracts human body images from each frame of the image to be identified based on the human body frames. Attribute state recognition is performed using a binary classification model for different human body attributes, such as aprons, sanitary caps, and masks (e.g., if a staff member is wearing an apron, the binary classification model outputs a 1, otherwise a 0). It should be noted that human body attribute state classification models require the pre-training of a large number of positive and negative human body image samples with corresponding attributes. Any deep learning classification method, such as ResNet18 or loss, can be used.

[0040] S20. As each person corresponding to a number appears and disappears from the dining space, record the number of people corresponding to each number in each frame, the length of stay corresponding to each number, the human attribute status corresponding to each number in each frame, and the number of frames corresponding to the human attribute status;

[0041] It should be noted that in the implementation of step S20, each time the human tracking algorithm generates a new ID number, it creates an identification status statistics table with four indicators for it. These indicators record the attribute identification of the relevant staff type of the person corresponding to the ID number throughout the entire identification cycle (the time from appearance to disappearance), and obtain statistical data on the status under different attributes. For example, in the chef identity and standard operation identification process below, the four indicators are: chef (wearing an apron or sanitary cap), chef without an apron (wearing a sanitary cap and not wearing an apron), chef without a sanitary cap (wearing an apron and not wearing a sanitary cap), and chef without a mask (wearing an apron or sanitary cap and not wearing a mask). The statistical content of these indicators includes: the number of people detected and tracked, the number of people corresponding to the number in each frame (the total number of people detected and tracked in each frame), the duration of stay corresponding to the number, the human attribute state corresponding to the number in each frame, and the number of frames corresponding to the human attribute state. In addition, when calculating the dwell time of all detected and tracked person IDs, if the ID appears for the first time, a headcount cache table is initialized for that ID, and the start time t1 of the ID's first appearance is recorded. If the ID is removed in the current frame, the end time t2 of the ID's removal is recorded. When an ID is removed in the current frame, the detected and tracked person corresponding to that ID is considered to have disappeared from the dining space.

[0042] S30: When the person corresponding to each number disappears from the dining space, whether the person corresponding to each number is an operation-constrained object is determined based on the number of frames corresponding to the human attribute state of the number, the length of stay corresponding to the number, and the total number of frames of the image to be recognized;

[0043] S40. When each object judged as an operation constraint disappears from the dining space, the irregular operation behavior of each operation constraint object corresponding to the number is judged and recorded based on the number of people corresponding to the number in each frame, the length of stay corresponding to the number, the number of frames corresponding to the human attribute state of the number, and the total number of frames of the image to be identified.

[0044] It should be noted that the existing methods for identifying irregular work behaviors of staff in the catering industry generally use the time period from the time each staff member enters the catering space until they leave as a time period, and separately count the relevant human attributes of the staff member during this time period, such as the cumulative time of not wearing an apron, not wearing a sanitary cap, not wearing a mask, etc. If it is not less than a specified threshold, it means that this type of irregular behavior has occurred, and a corresponding event is recorded. However, the reality is that there are many types of staff members in the catering industry, such as chefs, waiters, front desk staff, lobby managers, cleaning staff, etc. Using a unified standard to identify their irregular work behaviors does not meet specific actual needs and may result in misidentification; and using a single time period to judge irregular behavior may result in insufficient recognition accuracy. The present invention provides a high-accuracy method that can effectively identify the identities of different staff members based on the standardized behavior of chefs in the catering industry. This method can achieve effective identity recognition and high-accuracy recognition of irregular behavior for chefs, waiters, service staff, cleaners, etc. The present invention provides an embodiment for automatically identifying the identities of staff in the catering industry, using two schemes: staying in the corresponding work area for a long time, and in most cases wearing clothing or wearing identification (such as wearing work clothes, an apron, or a sanitary cap, etc.). The two schemes complement each other, ensuring that the corresponding category of staff can be accurately identified, thereby greatly reducing the error recognition rate. This method can be applied to the recognition of standardized work behaviors of different types of staff in the catering industry, and can also be applied to the recognition of standardized work behaviors of staff in other industries, and has a wide range of applicability.

[0045] In addition, this embodiment will label different types of employees in the catering industry with an ID number. In a monitoring scenario such as catering, the occlusion of the front and back personnel is relatively serious. When their trajectories intersect, the human tracking algorithm is prone to ID number exchange. In this case, the same ID number actually includes two different people. The existing technology directly identifies non-compliant behaviors for each frame of the ID number regardless of whether it is a related staff member, which will cause misidentification of staff members who are not in specific occasions. The present invention only identifies the dress code or the wearing of identification marks for frames of the human attribute status of different types of employees, effectively avoiding the shortcomings of the existing technology.

[0046] As an optional implementation, step S30 includes the following steps:

[0047] When the person corresponding to each number disappears from the catering space, the length of time that each numbered non-operational specification constraint object stays is calculated; the persons corresponding to all numbers whose non-operational specification constraint object stays for a length of time not less than a first threshold are determined as operational specification constraint objects; wherein the operational specification constraint object is the chef, and the catering space is the kitchen.

[0048] As an optional implementation, the calculation formula for the stay time T of each numbered non-operation specification constraint object is:

[0049] T=(t2-t1)×(count-count1) / count (1);

[0050] Among them, t2-t1 is the dwell time corresponding to each number; count is the total number of frames of the image to be recognized; count1 is the number of image frames corresponding to the first attribute state of the human body attribute of each number; the first attribute state is wearing an apron or a sanitary cap.

[0051] Furthermore, when the person corresponding to each number passes the non-operation specification constraint object and is not confirmed as the operation specification constraint object by the length of stay, step S30 further includes the following steps:

[0052] When the person corresponding to each number disappears from the dining space, the ratio of the number of image frames corresponding to the first attribute state of each number to the total number of frames of the image to be identified is calculated; all persons corresponding to numbers whose ratio is not less than the preset ratio are confirmed as objects subject to operational specification constraints. Of course, all persons corresponding to numbers whose ratio is less than the preset ratio are determined to have no non-compliant behavior. The calculation formula for the above ratio B is as follows:

[0053] B = count1 / count (2);

[0054] It should be noted that, generally, only chefs are required to wear aprons, sanitary caps, and masks at the same time, while other personnel only need to wear masks because they do not stay in the kitchen for a long time. Therefore, the screening criteria of this step are: wearing an apron or a sanitary cap, which can exclude other personnel to a certain extent. It should be further explained that this step provides two implementation methods for chef identification in the kitchen. For related types of staff in other areas of the catering industry, such as restaurants, customer waiting areas, etc., human attributes can be replaced by clothing, work badges, etc., and the first attribute status can be replaced by wearing work clothes or wearing work badges, etc. Of course, for other industries outside the catering industry, such as shopping stores, production plants, etc., the above-mentioned human attributes and first attribute status can be set according to actual conditions.

[0055] As an optional implementation, step S40 includes the following steps:

[0056] When it is determined that the operation constraint object disappears from the catering space, the cumulative stay time of each operation specification constraint object is calculated; each operation specification constraint object whose cumulative stay time is not less than the second threshold is confirmed as having non-standard operation behavior and recorded, and the process returns to step S10 to perform image recognition for the next batch.

[0057] As an optional implementation, the calculation formula for the cumulative stay time T1 of each operation specification constraint object is as follows:

[0058] T1=(t4-t3)×(count-count2) / count (3);

[0059] Where t4-t3 are the dwell times corresponding to the number of each object constrained by the work specification; count is the total number of frames in the image to be recognized; and count2 is the number of frames in the image where the human attribute state corresponding to each object constrained by the work specification satisfies both the first attribute state and the second attribute state. Simultaneous first and second attribute states include wearing a sanitary cap and no apron, wearing an apron and no sanitary cap, or wearing an apron or sanitary cap and no mask. The second attribute state includes not wearing an apron, not wearing a sanitary cap, or not wearing a mask.

[0060] It should be noted that this step provides an implementation method for identifying the standardized operating behavior of chefs whose operating standard constraints are objects. During the identification, it is necessary to separately identify each of the three specific non-standard states of the second attribute state, namely, not wearing an apron, not wearing a sanitary cap, and not wearing a mask, and then obtain the result of whether the chef is wearing an apron, a sanitary cap, or a mask. Of course, the second attribute state is not limited to these three and can be determined according to actual conditions. In addition, for relevant types of staff in other areas of the catering industry, such as restaurants, customer waiting areas, etc., the second attribute state can be replaced by wearing work clothes, wearing work badges, etc. Of course, for other industries outside the catering industry, such as shopping stores, production plants, etc., the above-mentioned human attributes and second attribute states can be set according to actual conditions.

[0061] In summary, considering that people other than chefs generally don't stay in the kitchen for long periods of time, chefs can be identified by using a longer duration of stay. For chefs with shorter durations, considering that even chefs who are not properly dressed will usually not be without both an apron and a sanitary cap, they can be identified by wearing either apron or sanitary cap. Thus, this embodiment automatically identifies chefs using two schemes: a longer duration of stay and wearing either an apron or a sanitary cap. These two schemes complement each other, ensuring accurate identification of all chefs and significantly reducing the false recognition rate. In summary, this embodiment uses duration of stay and a specific human attribute state (first attribute state) to identify the identity of objects subject to operational specification constraints. For identified objects subject to operational specification constraints, a second attribute state is used to identify non-standard operations. This method can identify the identities of different types of staff in the catering industry, effectively avoiding misidentification of different types of staff, significantly improving the efficiency and accuracy of identifying operational specifications for catering industry staff, and effectively preventing false positives caused by ID number swapping in human tracking algorithms.

[0062] Example 2:

[0063] This embodiment also provides an image recognition method for determining whether rest time of catering industry staff is irregular. Specifically, step S40 further includes the following steps:

[0064] When it is determined that the operation-constrained object disappears from the catering space, the non-compliant stay time of each operation-specification-constrained object is calculated; each operation-specification-constrained object whose non-compliant stay time meets the preset conditions is determined as an operation non-standard and recorded, and the process returns to step S10 to perform image recognition for the next batch; wherein the catering space is a rest area.

[0065] As an optional implementation, the calculation formula for the non-compliant stay time DT of each operation specification constraint object is as follows:

[0066] DT = (t6 - t5) × (1 - r) (4);

[0067] r = num / count3 (5);

[0068] Among them, num is the number of image frames corresponding to the number of people in each frame of each operation constraint object that is greater than C; count3 is the number of image frames corresponding to the number of each operation constraint object; C is the minimum number of people that can normally stay in the catering space; t6-t5 is the length of stay corresponding to the number of each operation constraint object.

[0069] As an optional implementation manner, the preset conditions of this embodiment are:

[0070] The duration of non-compliant stay is not less than the third threshold, and r <R;

[0071] Among them, R is the threshold value of the ratio of the total time that the normally staying work-constrained objects stay in the catering space to the collective gathering time.

[0072] It should be noted that the operational regulations of this implementation can apply to chefs, all restaurant staff, and, of course, other industries as described in Example 1. Rest periods can be standardized for all staff or for specific job categories. For the identification of all employees' identity attributes, the relevant attribute state can be the attire or related accessories worn by each type of employee. For example, in Example 1, the chef can be identified by wearing a sanitary cap or apron; for front-of-house staff, the attire or identification badges worn can be identified. It can also be identified by common employee identification, such as chefs, dishwashers, and cleaners in the kitchen, who all wear uniform work clothes, aprons, or sanitary caps. Another example is that all staff members wear badges, or all employees' clothing has a prominent company logo. Of course, in some cases, this identification may not be performed for each employee category, such as when no one else is allowed into the employee rest area. This can be determined based on actual circumstances.

[0073] It's important to note that during normal rest periods, large numbers of people often gather in the dining area's lounge, so the number of people present can be used to roughly determine whether a person is resting. However, because each person enters the lounge at a different time, there may not be many people present at the beginning or end of a break, making it easy to mistakenly detect those who enter the lounge first or leave last as overstaying. However, these situations all share a common characteristic: the entire time a person spends in the lounge overlaps with the time period of the group gathering. Therefore, this overlap can be used to filter out these situations, avoiding false positives. Based on the above situation, on the one hand, this embodiment maintains a cache data list consisting of a number, the number corresponding to the number in each frame, the stay time corresponding to the number, and the number of image frames corresponding to the number for each tracked and detected person, and sets a maximum number of non-compliant stay thresholds, which is equivalent to automatically discovering normal rest time through this method and not counting the stay time for these times, which effectively avoids a large number of false detections of overtime stays in cases where people gather together; on the other hand, calculating the proportion of the stay time corresponding to the number that overlaps with the collective gathering time, which is equivalent to automatically discovering the person who enters the lounge first or leaves the lounge last during the rest time through this method, making up for the defect of judging the rest time simply by the number of people, which effectively avoids false detections of overtime stays at the start and end of the rest, and can even infer the actual complete rest time period based on this.

[0074] In addition, in some application scenarios, it is possible to identify irregular rest time not only for the task-constrained objects, but also for all employees. The relevant steps are as follows:

[0075] According to the number of each person obtained by human body detection and tracking, the number corresponding to each detected and tracked person from the time he appears to the time he disappears in the dining space, the number corresponds to the number of people in each frame, the length of stay corresponding to the number, and the number of image frames corresponding to the number are recorded; according to the number of each detected and tracked person, the number of people in each frame corresponding to the number, the length of stay corresponding to the number, and the number of image frames corresponding to the number, the non-compliant length of stay of each detected and tracked person is calculated respectively; the detected and tracked person whose non-compliant length of stay meets the preset conditions is manually determined as having non-standard operation, and the process returns to step S10 to perform image recognition for the next batch.

[0076] To sum up, the embodiments of the present invention provide an image recognition method for the work specifications of catering industry staff, which can identify the irregular wear of specific types of staff, and can also identify the irregular behavior of staff taking overtime breaks. It has a high recognition accuracy rate, effectively avoids false detection, and has a wide range of applications and market promotion value.

[0077] The above are merely preferred embodiments of the present invention. Those skilled in the art will appreciate that various changes or equivalent substitutions may be made to these features and embodiments without departing from the spirit and scope of the present invention. Furthermore, under the guidance of the present invention, these features and embodiments may be modified to adapt to specific circumstances and materials without departing from the spirit and scope of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are intended to be within the scope of the present invention.

Claims

1. A method for image recognition of work specifications of catering industry staff, characterized by: The following steps are involved: S10. In the dining space, perform human body detection and tracking on the acquired multiple frames of images to be identified, and obtain a number corresponding to each person detected and tracked; S20. During the process from the appearance to the disappearance of the person corresponding to each number in the dining space, record the number of people corresponding to each number in each frame, the length of stay corresponding to each number, the human attribute status corresponding to each number in each frame, and the number of frames corresponding to the human attribute status; S30, when the person corresponding to each number disappears from the dining space, judging whether the person corresponding to each number is an operation-constrained object based on the number of frames corresponding to the human attribute state of the number, the length of stay corresponding to the number, and the total number of frames of the image to be identified; S40: When each object determined to be constrained by the work disappears from the dining space, the irregular work behavior of each object corresponding to the number is determined and recorded based on the number of people in each frame, the length of stay corresponding to the number, the number of frames corresponding to the human attribute state of the number, and the total number of frames of the image to be identified; Step S30 includes the following steps: When each person corresponding to the number disappears from the dining space, the length of time that each number of the non-operational specification constraint object stays is calculated; and all persons corresponding to the numbers whose length of time that the non-operational specification constraint object stays is not less than a first threshold are determined as the operation specification constraint object; The calculation formula for the length of time T of stay of each non-operational specification constraint object is: T=(t2-t1)×(count-count1) / count; Among them, t2-t1 is the dwell time corresponding to each number; count is the total number of frames of the image to be identified; count1 is the number of image frames corresponding to the first attribute state of the human body attribute state for each number.

2. The image recognition method for the working standards of catering industry staff according to claim 1 is characterized in that: The object constrained by the operation specification is the chef, and the dining space is the kitchen.

3. The image recognition method for the working standards of catering industry staff according to claim 2 is characterized in that: The first attribute state is wearing an apron or a sanitary cap.

4. The image recognition method for the working standards of catering industry staff according to claim 3 is characterized in that: When the person corresponding to each number passes the non-operation specification constraint object and the length of stay is not confirmed as the operation specification constraint object, step S30 further includes the following steps: When the person corresponding to each number disappears from the dining space, the ratio of the number of image frames corresponding to the first attribute state of each numbered human body to the total number of frames of the image to be identified is calculated; The persons corresponding to all the numbers whose ratios are not less than a preset ratio are confirmed as the objects constrained by the operation specification.

5. The image recognition method for the working standards of catering industry staff according to claim 3 is characterized in that: Step S40 includes the following steps: When it is determined that the operation constraint object disappears from the dining space, the cumulative stay time of each operation constraint object is calculated; Each of the operation specification constraint objects whose cumulative stay time is not less than the second threshold is confirmed as having non-standard operation behavior and recorded, and the process returns to step S10 to perform image recognition for the next batch.

6. The image recognition method for the working standards of catering industry staff according to claim 5 is characterized in that: The calculation formula for the cumulative stay time T1 of each object constrained by the operation specification is as follows: T1=(t4-t3)×(count-count2) / count; Wherein, t4-t3 is the dwell time corresponding to the number of each operation specification constraint object; count is the total number of frames of the image to be identified; count2 is the number of image frames in which the human attribute state corresponding to the number of each operation specification constraint object satisfies both the first attribute state and the second attribute state; The second attribute state includes not wearing an apron, not wearing a sanitary cap, and not wearing a mask.

7. The image recognition method for the working standards of catering industry staff according to claim 1 is characterized in that: Step S40 also includes the following steps: When it is determined that the operation constraint object disappears from the dining space, calculating the non-compliant stay time of each operation constraint object; Determine each of the operation specification constraint objects whose non-compliant stay duration meets the preset conditions as an operation non-standard and record the result, and return to step S10 to perform image recognition for the next batch; Wherein, the dining space is a rest area.

8. The image recognition method for the working standards of catering industry staff according to claim 7 is characterized in that: The calculation formula of the non-compliant stay time DT for each object constrained by the operation specification is as follows: DT = (t6 - t5) × (1 - r); r = num / count3; Among them, num is the number of people corresponding to each operation constraint object in each frame is greater than the number of image frames corresponding to C, count3 is the number of image frames corresponding to the number of each operation constraint object; C is the minimum number threshold for normal stay in the catering space; t6-t5 is the length of stay corresponding to the number of each operation constraint object.

9. The image recognition method for the working standards of catering industry staff according to claim 8 is characterized in that: The preset conditions are: The non-compliant stay duration is not less than the third threshold, and r <R; Among them, R is the threshold value of the ratio of the total time that the normally staying operation-constrained objects stay in the catering space to the collective gathering time.

Images