A method and system for improving the management efficiency of a specific hotel

Abstract

This invention provides a method and system for improving the management efficiency of specific hotels. It utilizes pedestrian re-identification to determine whether people in the scene are wearing protective clothing; it links hotel room information with camera monitoring to verify whether the number of people in the room matches the registered number, and to check whether people have left without authorization or whether outsiders have entered; it utilizes existing equipment without the need to add additional electronic devices, thereby improving the management efficiency of specific hotels.

Description

Technical Field

[0001] This invention relates to the field of epidemic prevention and control, and in particular to the management of specific hotels. Background Technology

[0002] In certain hotels, there have been instances of specific individuals leaving without authorization, family members and other unspecified individuals mixing into the hotel and sharing rooms with specific individuals, and hotel staff not operating their services properly and not wearing protective clothing as required.

[0003] Currently, in certain hotels, manual intervention is required to monitor whether the number of guests in each room matches the front desk registration, whether personnel in corridors are wearing protective clothing as required, and whether specific personnel have left their rooms without authorization. This results in a significant waste of human and material resources. This invention addresses these three issues by proposing an intelligent solution to improve the management efficiency of specific hotels. Summary of the Invention

[0004] (1) Technical problems solved

[0005] To address the aforementioned technical problems, this invention provides a method and system for improving the management efficiency of specific hotels. It utilizes pedestrian re-identification to determine whether individuals in the scene are wearing protective clothing; it links hotel room information with camera monitoring to verify whether the number of people in the rooms matches the registered number and to check whether personnel have left without authorization; and it utilizes existing equipment without the need for additional electronic devices, thereby improving the management efficiency of specific hotels.

[0006] (2) Technical solution

[0007] To solve the aforementioned technical problems and achieve the invention's objective, the present invention is implemented through the following technical solution:

[0008] A method for improving the management efficiency of a specific hotel includes the following steps:

[0009] S1: The intelligent analysis service accesses the video stream of the monitoring screen, performs motion detection on the screen, and determines whether there are people or objects in the screen;

[0010] S2: Enable the head / body-based target detection algorithm to detect whether there is a person in the scene. If the head / body target detection has a result, it means that there is a person in the scene. Then cache the frame image and proceed to S3. Otherwise, return to S1.

[0011] S3: Personnel tracking between consecutive frames, and the fitted movement trajectory;

[0012] S4: Based on the movement trajectory of the personnel, determine the number of people entering and leaving each room, thereby determining the personnel's entry and exit locations and the number of people in the room; compare the number of people in the room with the number of people registered in the hotel for that room. If the number does not match, proceed to S7; otherwise, return to S1.

[0013] S5: Determine if the target person is the target person. If yes, return to step S1; otherwise, proceed to the next step.

[0014] S6: When a target is determined to be a non-target person, further determine whether the target person is present at this moment.

[0015] If the target is determined to be a non-target person and no target person is traveling with them, proceed to the next step; otherwise, return to step S1.

[0016] S7: Linked Warning. Uploads abnormal room information and images of target personnel to the front-end server and issues a warning message.

[0017] Further,

[0018] Step S2 also includes: using the cross-union ratio result of the image processing of the head and body bounding boxes in the current frame, the head and body can be matched to the same identity.

[0019] In step S3, the method for determining whether the person targets in the previous and next frames have the same identity is optical flow or IOU matching.

[0020] Furthermore, step S4 determines the locations of personnel entering and exiting the room and the number of people in the room, specifically as follows:

[0021] A detection frame is set at each door location. When a target appears, it is determined whether the target overlaps with the detection frame area of ​​that door, and the following cases are considered:

[0022] (1) If the target is detected in the detection frame of door A in the first frame and disappears in the detection frame of door B, it is determined that the person exits from door A and enters from door B, and the number of people in the corresponding room is added or subtracted respectively.

[0023] (2) If the target is detected in the door detection frame of room A in the first frame and disappears in a position outside the door detection frame, it is determined that the person left from door A and left from other positions, and the number of people in the corresponding room is added or subtracted respectively.

[0024] (3) If the target is detected in the non-door detection frame area in the first frame and disappears in the door detection frame of room A, it is determined that the person entered from other locations and entered room A. The number of people in the corresponding room is then added or subtracted.

[0025] Furthermore, step S5, determining whether a target person is the target person, specifically includes:

[0026] Establish a database of images of people wearing protective suits as a feature base for pedestrian re-identification (reID);

[0027] In the movement trajectory of people, the candidate image with the best quality continuous target bounding boxes for the same identity ID is selected. Further, it is determined whether the aspect ratio of the human body in the candidate image is greater than the human aspect ratio threshold for pedestrian re-identification. If the threshold is not met, it indicates that the target is too small or occluded, and is an incomplete human body image; if the threshold is met, it indicates that it is a complete human body image.

[0028] Pedestrian re-identification features are extracted from the highest quality complete human body image obtained above, compared with the protective clothing feature database, and the best comparison result is selected. The comparison result is then compared with two thresholds extracted by the pedestrian re-identification algorithm, namely the minimum identification threshold and the maximum identification threshold.

[0029] When the comparison result is less than the minimum threshold, it indicates that the target is not a target person. When the comparison result is greater than the maximum identification threshold, it indicates that the target is a target person. When the comparison result falls between the minimum and maximum identification thresholds, further determination is needed to determine whether the target is a target person.

[0030] When the comparison result falls between the minimum and maximum recognition thresholds, the method for further determining whether the target is a person is as follows: Convert the human image of the target to the HSV color space. Calculate whether the proportion of white pixels is greater than the threshold. If it is greater than the threshold, it indicates that the target is a person. The definition of white pixels can be chosen within an appropriate range, and its value is related to the specific image.

[0031] Step S6 also includes further detecting whether there are other people in the current frame using cached frame images, and determining whether these people are the target people based on step S5.

[0032] The present invention also provides a system for improving the management efficiency of a specific hotel, applied to the management of a specific hotel, the system comprising:

[0033] Detection module: Used to first perform motion detection on the screen to determine whether there are people or objects in the screen; when there are moving objects in the screen, the target detection algorithm based on human head / body is activated to detect whether there are people in the screen.

[0034] Image caching module: When the detection module detects that someone appears in the frame, it caches that frame of the image.

[0035] The trajectory fitting module is used to fit the movement trajectory. Specifically, it includes: when the detection module detects a person in the frame, it uses optical flow or IOU matching to determine whether the person target in the previous and next frames has the same ID. This results in the movement trajectory of the person from appearance to disappearance in the frame, fitted using the head and body outlines.

[0036] Counting module: Used to determine the number of people in each room; specifically, it includes: obtaining the movement trajectory of people based on the trajectory fitting module, judging the number of people entering and leaving each room, thereby determining the location of people entering and leaving and the number of people in the room, and comparing the number of people in the room with the number of people registered in the hotel for that room. If the number does not match, the alarm module is connected to issue a warning message.

[0037] Personnel detection module: Determines whether the target person is the target person. If the target person is not the target person, it retrieves the cached frame image from the image caching module and determines whether the target person is on the screen. If not, it connects to the alarm module to issue a warning message.

[0038] The process of identifying the target person specifically includes: establishing a database of images of people wearing protective clothing as the feature base database for pedestrian re-identification (reID); using the reID algorithm to extract features with an M-dimensional feature matrix, thereby obtaining an N*M feature matrix, which is referred to as G.

[0039] In the movement trajectory of people, the candidate image with the best quality continuous target bounding boxes for the same identity ID is selected. Further, it is determined whether the aspect ratio of the human body in the candidate image is greater than the human aspect ratio threshold HW for pedestrian re-identification. If the threshold is not met, it indicates that the target is too small or occluded, and is an incomplete human body image; if the threshold is met, it indicates that it is a complete human body image.

[0040] Pedestrian re-identification features are extracted from the highest quality complete human image obtained above, and compared with the protective clothing feature database. The best comparison result is taken, and this result is denoted as R. R is compared with two thresholds extracted by the reID algorithm, namely the minimum recognition threshold Tmin and the maximum recognition threshold Tmax.

[0041] When the result R is less than the minimum threshold Tmin, it indicates that the target is not a target person. When the result R is greater than the maximum threshold Tmax, it indicates that the target is a target person. When R falls within the interval [Tmin, Tmax], in order to minimize false detections caused by external factors such as lighting, clothing, and obstruction, it is necessary to further determine whether the target is a target person.

[0042] When the result R is in the range [Tmin, Tmax], the method to further determine whether the target is a person is as follows: convert the human image of the target to the HSV color space. Calculate whether the proportion of white pixels is greater than the threshold TW. If it is greater than the threshold TW, it indicates that the target is a person. The definition of white pixels can be a suitable range [Wmin, Wmax], and its value is related to the specific image.

[0043] Alarm module: It includes linked alarm devices that upload abnormal room information and images of target personnel to the front desk server and issue warning messages.

[0044] (3) Beneficial effects

[0045] (1) By intelligently monitoring whether the number of guests in each room is consistent with the front desk registration, whether the people in the corridor are wearing protective clothing as required, whether specific personnel leave the room without authorization, and whether external personnel enter, the management efficiency of specific hotels can be improved without the need to add additional electronic devices.

[0046] (2) Introducing the human head as the detection target can reduce the situation where the lower body is obscured due to people crossing in front and behind, which leads to inaccurate human body detection frame;

[0047] (3) A complete human body image is obtained by combining a pedestrian re-identification algorithm with a target box quality screening algorithm, thereby improving the accuracy of the algorithm for identifying target personnel. Attached Figure Description

[0048] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0049] Figure 1 A flowchart illustrating a method for improving the management efficiency of a specific hotel, as provided in an embodiment of the present invention.

[0050] Figure 2 A flowchart of a method for determining whether a target is a target person, provided in an embodiment of the present invention;

[0051] Figure 3 This is a system structure diagram provided for an embodiment of the present invention. Detailed Implementation

[0052] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0053] To explain the technical content, objectives, and technical effects of the present invention in detail, the following description is provided in conjunction with the embodiments and accompanying drawings.

[0054] See Figure 1 The flowchart below shows a method for improving the management efficiency of a specific hotel according to the present invention. The method specifically includes:

[0055] S1: The intelligent analysis service accesses the video stream from the monitoring screen, performs motion detection on the screen, and determines whether there are people or objects in the screen.

[0056] Compared to directly performing full-screen object and person detection, introducing motion detection can significantly reduce computational power consumption. If movement is detected in the image, subsequent processing is performed; otherwise, streaming continues.

[0057] S2: Enable a head / body-based target detection algorithm to detect if anyone is present in the scene. Using the Intersection over Union (IOU) result of the current frame's head and body bounding boxes, the head and body can be matched to have the same identity (ID). If the head / body target detection is successful, it indicates that someone is present in the scene; the frame image is cached, and the process proceeds to the next step; otherwise, the process returns to the previous step.

[0058] Introducing a human head as an additional detection target can reduce the situation where the lower body is obscured due to people crossing in front and behind, which would lead to inaccurate human body detection frames.

[0059] S3: Person tracking between consecutive frames, fitting the movement trajectory. Specifically, using methods such as optical flow and IOU matching, it determines whether the person in the consecutive frames is the same person. This results in the movement trajectory of the person from appearance to disappearance in the frame, fitted using the head and body outlines.

[0060] To prevent cross-infection caused by specific individuals visiting each other's rooms, or by outsiders entering specific rooms, or by specific individuals leaving their rooms or even leaving the hotel without authorization, which could affect the effectiveness of the measures and lead to the spread of the epidemic, it is necessary to strictly control the number of people entering and leaving the rooms.

[0061] S4: Based on the movement trajectory of personnel, determine the number of people entering and exiting each room, thereby determining the location of personnel entering and exiting and the number of people in each room. Specifically:

[0062] A detection frame is set at each door location. When a target appears, it is determined whether the target overlaps with the detection frame area of ​​that door, and the following cases are considered:

[0063] (1) If the target is detected in the detection frame of door A in the first frame and disappears in the detection frame of door B, it is determined that the person exits from door A and enters from door B. Then the number of people in room A is reduced by 1 and the number of people in room B is increased by 1.

[0064] (2) If the target is detected in the door detection frame of room A in the first frame and disappears in a position outside the door detection frame, it is determined that the person left from door A and left from other positions, and the number of people in room A is reduced by 1.

[0065] (3) If the target is detected in the non-door detection frame area in the first frame and disappears in the door detection frame of room A, it is determined that the person entered from other locations and entered room A, and the number of people in room A is increased by 1.

[0066] Compare the number of people in the room with the number of people registered in the hotel for that room. If the number of people does not match, proceed to S7; otherwise, return to S1.

[0067] S5: Determine if the target person is the target person. If yes, return to step S1; otherwise, proceed to the next step.

[0068] like Figure 2 As shown, determining whether someone is a target person involves the following steps:

[0069] S51: Establish a database of images of people wearing protective suits as the feature base for pedestrian re-identification (reID). Specifically, use N images. The larger N is, the more protective suit poses can be included, but at the same time, the computational power consumption increases. Use the reID algorithm to extract features with an M-dimensional feature matrix, and then obtain an N*M feature matrix, which is referred to as G.

[0070] S52: In the movement trajectory of a person, select the candidate image with the best quality continuous target bounding boxes for the same identity ID. Further determine whether the aspect ratio of the human body in the candidate image is greater than the human body aspect ratio threshold HW for pedestrian re-identification. If it does not meet the threshold, it indicates that the target is too small or occluded, and is an incomplete human body image; if it meets the threshold, it indicates that it is a complete human body image.

[0071] Quality screening method: Determine the trajectory of each target object. For each frame in the trajectory: Calculate the aspect ratio of the detection box and determine the deviation between the aspect ratio and a first average value to obtain an aspect ratio deviation value. The first average value is the average aspect ratio of the detection boxes of the same target object in the entire trajectory. The detection box includes a first box or a second box. Calculate the height ratio between the first box and the second box and determine the deviation between the height ratio and a second average value to obtain a height ratio deviation value. The second average value is the average height ratio between the first box and the second box of the same target object in the entire trajectory. Based on the aspect ratio deviation value, the height ratio deviation value, and the IOU score, determine the quality score result and judge the quality of the target object based on the quality score result.

[0072] S53: Extract pedestrian re-identification features from the highest quality complete human image obtained above, compare them with the protective clothing feature database, and take the best comparison result, which is denoted as R. R is compared with two thresholds extracted by the reID algorithm, namely the minimum recognition threshold Tmin and the maximum recognition threshold Tmax.

[0073] When the result R is less than the minimum threshold Tmin, it indicates that the target is not a target person. When the result R is greater than the maximum threshold Tmax, it indicates that the target is a target person. When R falls within the interval [Tmin, Tmax], in order to minimize false detections caused by external factors such as lighting, clothing, and obstruction, it is necessary to further determine whether the target is a target person.

[0074] S54: When the result R is in the range [Tmin, Tmax], the method to further determine whether the target is a person is as follows: Convert the human image of the target to the HSV color space. Calculate whether the proportion of white pixels is greater than the threshold TW. If it is greater than the threshold TW, it indicates that the target is a person. The definition of white pixels can be a suitable range [Wmin, Wmax], and its value is related to the specific image.

[0075] S6: When a target is determined to be a non-target person, further determine whether the target person is present at this moment.

[0076] For example, when being routinely inspected by specific personnel, non-target personnel are allowed to move around, but they must be accompanied by the target personnel.

[0077] The cached frame images are used to further detect whether there are other people in the current scene, and to determine whether these people are the target people according to step S5.

[0078] If the target is determined to be a non-target person and no target person is traveling with them, proceed to the next step; otherwise, return to step S1.

[0079] S7: Linked Warning. Uploads abnormal room information and images of target personnel to the front-end server and issues a warning message.

[0080] In this embodiment, pedestrian re-identification is used to determine whether people in the scene are wearing protective clothing; by linking hotel room information with camera monitoring, the number of people in the room is checked to see if it matches the number of people registered, and to check whether people have left without permission or whether outsiders have entered; the management efficiency of a specific hotel is improved by using existing surveillance cameras without the need to add additional electronic equipment.

[0081] This invention also proposes a system for improving the management efficiency of a specific hotel, the system structure of which is as follows: Figure 3 As shown, it specifically includes:

[0082] Detection module: Used to first perform motion detection on the screen to determine whether there are people or objects in the screen; when there are moving objects in the screen, the target detection algorithm based on human head / body is activated to detect whether there are people in the screen.

[0083] Image caching module: When the detection module detects that someone appears in the frame, it caches that frame of the image.

[0084] The trajectory fitting module is used to fit the movement trajectory. Specifically, it includes: when the detection module detects a person in the frame, it uses optical flow or IOU matching to determine whether the person target in the previous and next frames has the same ID. This results in the movement trajectory of the person from appearance to disappearance in the frame, fitted using the head and body outlines.

[0085] Counting module: Used to determine the number of people in each room; specifically, it includes: obtaining the movement trajectory of people based on the trajectory fitting module, judging the number of people entering and leaving each room, thereby determining the location of people entering and leaving and the number of people in the room, and comparing the number of people in the room with the number of people registered in the hotel for that room. If the number does not match, the alarm module is connected to issue a warning message.

[0086] The specific counting process of the counting module is as follows:

[0087] A detection frame is set at each door location. When a target appears, it is determined whether the target overlaps with the detection frame area of ​​that door, and the following cases are considered:

[0088] (1) If the target is detected in the detection frame of door A in the first frame and disappears in the detection frame of door B, it is determined that the person exits from door A and enters from door B. Then the number of people in room A is reduced by 1 and the number of people in room B is increased by 1.

[0089] (2) If the target is detected in the door detection frame of room A in the first frame and disappears in a position outside the door detection frame, it is determined that the person left from door A and left from other positions, and the number of people in room A is reduced by 1.

[0090] (3) If the target is detected in the non-door detection frame area in the first frame and disappears in the door detection frame of room A, it is determined that the person entered from other locations and entered room A, and the number of people in room A is increased by 1.

[0091] Personnel detection module: Determines whether the target person is the target person. If the target person is not the target person, it retrieves the cached frame image from the image caching module and determines whether the target person is on the screen. If not, it connects to the alarm module to issue a warning message.

[0092] The process of identifying the target person specifically includes: establishing a database of images of people wearing protective clothing as the feature base database for pedestrian re-identification (reID); using the reID algorithm to extract features with an M-dimensional feature matrix, thereby obtaining an N*M feature matrix, which is referred to as G.

[0093] In the movement trajectory of people, the candidate image with the best quality continuous target bounding boxes for the same identity ID is selected. Further, it is determined whether the aspect ratio of the human body in the candidate image is greater than the human aspect ratio threshold HW for pedestrian re-identification. If the threshold is not met, it indicates that the target is too small or occluded, and is an incomplete human body image; if the threshold is met, it indicates that it is a complete human body image.

[0094] Pedestrian re-identification features are extracted from the highest quality complete human image obtained above, and compared with the protective clothing feature database. The best comparison result is taken, and this result is denoted as R. R is compared with two thresholds extracted by the reID algorithm, namely the minimum recognition threshold Tmin and the maximum recognition threshold Tmax.

[0095] When the result R is less than the minimum threshold Tmin, it indicates that the target is not a target person. When the result R is greater than the maximum threshold Tmax, it indicates that the target is a target person. When R falls within the interval [Tmin, Tmax], in order to minimize false detections caused by external factors such as lighting, clothing, and obstruction, it is necessary to further determine whether the target is a target person.

[0096] When the result R is in the range [Tmin, Tmax], the method to further determine whether the target is a person is as follows: convert the human image of the target to the HSV color space. Calculate whether the proportion of white pixels is greater than the threshold TW. If it is greater than the threshold TW, it indicates that the target is a person. The definition of white pixels can be a suitable range [Wmin, Wmax], and its value is related to the specific image.

[0097] Alarm module: It includes linked alarm devices that upload abnormal room information and images of target personnel to the front desk server and issue warning messages.

[0098] The computer program of this invention can be deployed on a server or AI box, such as NVIDIA NX, without the need for additional electronic devices, thereby improving the management efficiency of specific hotels.

[0099] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for improving the management efficiency of a specific hotel, characterized in that, The method includes: S1: The intelligent analysis service accesses the video stream of the monitoring screen, performs motion detection on the screen, and determines whether there are people or objects in the screen; S2: Activate the target detection algorithm based on human head and human body to detect whether there is a person in the picture. If there is a result for human head and human body target detection, it means that there is a person in the picture. Then cache the frame image and proceed to S3. Otherwise, return to S1. S3: Track people in consecutive frames and fit their movement trajectories; determine whether the people in consecutive frames are the same person; and then obtain the movement trajectory of people from the moment they appear in the frame to the moment they disappear, fitted by the head and body outlines. S4: Based on the movement trajectory of personnel, determine the number of people entering and exiting each room, thereby determining the location of personnel entering and exiting and the number of people in the room; compare the number of people in the room with the number of people registered in the hotel for that room. If the number does not match, proceed to S7; otherwise, return to S1. S5: Determine if the personnel target is the target personnel. If yes, return to step S1; otherwise, proceed to the next step. S6: When it is determined that a person is not the target person, further determine whether the target person is present based on the cached frame image in step S2; repeat step S5 to determine whether other people in the cached frame image are the target person; If it is determined that the person is not the target person and there is no target person traveling with them, proceed to the next step; otherwise, return to step S1. S7: Linked warning, uploads abnormal room information and target images of personnel to the front desk server, and issues a warning message; The determination method in step S5 is as follows: Establish a database of images of people wearing protective suits as a feature base for pedestrian re-identification; In the movement trajectory of the person, the candidate image with the best quality of continuous target boxes of the same identity is selected; further, it is determined whether the aspect ratio of the human body in the candidate image is greater than the human body aspect ratio threshold for pedestrian re-identification; if the threshold is not met, it indicates that the target is too small or is occluded and is an incomplete human body image; if the threshold is met, it indicates that it is a complete human body image. Pedestrian re-identification features are extracted from the highest quality complete human body image obtained above, compared with the protective clothing feature database, and the best comparison result is selected. The best comparison result is then compared with the minimum and maximum identification thresholds extracted by the pedestrian re-identification algorithm. When the optimal comparison result is less than the minimum threshold, it indicates that the person's target is not the target person; when the optimal comparison result is greater than the maximum identification threshold, it indicates that the person's target is the target person; when the optimal comparison result falls between the minimum identification threshold and the maximum identification threshold, it is necessary to further determine whether the person's target is the target person.

2. The method for improving the management efficiency of a specific hotel according to claim 1, characterized in that, Step S2 specifically includes: using the intersection-union ratio (IUU) results of the image processing of the head and body bounding boxes in the current frame to match whether the head and body belong to the same person.

3. The method for improving the management efficiency of a specific hotel according to claim 1, characterized in that, In step S3, the method for determining whether the person targets in the previous and next frames have the same identity is optical flow or IOU matching.

4. The method for improving the management efficiency of a specific hotel according to claim 1, characterized in that, Step S4 further includes: setting a detection frame at each door position, and when a target appears, determining whether the target overlaps with the detection frame area of ​​the door, which is divided into the following cases: a. If a target is detected in the detection frame of door A in the first frame and disappears in the detection frame of door B, then it is determined that the person exited through door A and entered through door B, and the number of people in the corresponding room is added or subtracted accordingly. b. If a target is detected in the door detection frame of room A in the first frame and disappears in a position outside the door detection frame, it is determined that the person left from door A and left from other positions, and the number of people in the corresponding room is added or subtracted accordingly. c. If a target is detected in the non-door detection frame area in the first frame and disappears in the door detection frame of room A, it is determined that the person entered room A from another location, and the number of people in the corresponding room is added or subtracted accordingly.

5. The method for improving the management efficiency of a specific hotel according to claim 1, characterized in that, The further determination of whether the person is the target person includes: when the result comparison result is between the minimum recognition threshold and the maximum recognition threshold, converting the human image of the person to the HSV color space; calculating whether the proportion of white pixels is greater than the threshold; if it is greater than the threshold, it indicates that the person is the target person.

6. A system for improving the management efficiency of a specific hotel, characterized in that, The system includes: Detection module: Used to first perform motion detection on the image to determine whether there are people or objects in the image; when there are moving objects in the image, the target detection algorithm based on human head and body is activated to detect whether there are people in the image; Image caching module: used to cache the image frame when the detection module detects that someone is in the frame; Trajectory Fitting Module: Used to fit the movement trajectory; specifically, it includes: when the detection module detects that a person appears in the picture, it uses optical flow or IOU matching to determine whether the person target in the previous and next frames has the same ID; and then obtains the movement trajectory of the person from the appearance to the disappearance in the picture, which is fitted by the head and body outline. Counting module: used to determine the number of people in each room; specifically, it includes: obtaining the movement trajectory of people based on the trajectory fitting module, judging the number of people entering and leaving each room, thereby determining the location of people entering and leaving and the number of people in the room, and comparing the number of people in the room with the number of people registered in the hotel for that room. If the number does not match, the alarm module is connected to issue a warning message. Personnel detection module: Determines whether the personnel target is the target person. If the personnel target is not the target person, it retrieves the cached frame image from the image caching module and determines whether there is a target person on the screen. If not, it connects to the alarm module to issue a warning message. The process of identifying target personnel specifically includes: establishing a database of images of people wearing protective clothing as the feature base for pedestrian re-identification, and using pedestrian re-identification algorithms to extract features; In the movement trajectory of the person, the candidate image with the best quality of continuous target boxes of the same identity is selected; further, it is determined whether the aspect ratio of the human body in the candidate image is greater than the human body aspect ratio threshold for pedestrian re-identification; if the threshold is not met, it indicates that the target is too small or is occluded and is an incomplete human body image; if the threshold is met, it indicates that it is a complete human body image. The pedestrian re-identification features are extracted from the highest quality complete human body image obtained above, compared with the protective clothing feature database, and the best comparison result is selected. The best comparison result is then compared with the minimum and maximum identification thresholds extracted by the pedestrian re-identification algorithm. When the optimal comparison result is less than the minimum threshold, it indicates that the person's target is not the target person; when the optimal comparison result is greater than the maximum identification threshold, it indicates that the person's target is the target person; when the optimal comparison result falls between the minimum identification threshold and the maximum identification threshold, it is necessary to further determine whether the person's target is the target person. Alarm module: It includes linked alarm devices that upload abnormal room information and images of personnel to the front-end server and issue warning messages.

7. The system for improving the management efficiency of a specific hotel according to claim 6, characterized in that, The counting module further includes: setting a detection frame at each door position; when a target appears, determining whether the target overlaps with the detection frame area of ​​that door, and classifying the following cases: a. If a target is detected in the detection frame of door A in the first frame and disappears in the detection frame of door B, then it is determined that the person exited through door A and entered through door B, and the number of people in the corresponding room is added or subtracted accordingly. b. If a target is detected in the door detection frame of room A in the first frame and disappears in a position outside the door detection frame, it is determined that the person left from door A and left from other positions, and the number of people in the corresponding room is added or subtracted accordingly. c. If a target is detected in the non-door detection frame area in the first frame and disappears in the door detection frame of room A, it is determined that the person entered room A from another location, and the number of people in the corresponding room is added or subtracted accordingly.

8. The system for improving the management efficiency of a specific hotel according to claim 6, characterized in that, The personnel identification module further includes: further determining whether the personnel target is the target personnel, including: when the result comparison result is in the range between the minimum identification threshold and the maximum identification threshold, converting the human body image of the personnel target to the HSV color space; calculating whether the proportion of white pixels is greater than the threshold; if it is greater than the threshold, it indicates that the personnel target is the target personnel.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method for improving the management efficiency of a specific hotel as described in any one of claims 1 to 5.

Images