An intelligent operation monitoring management system and method for kitchen

Abstract

The present application relates to the field of intelligent monitoring technology, and in particular to an intelligent operation monitoring management system and method for a back kitchen. The method includes: obtaining back kitchen partition camera information through at least one camera, analyzing the back kitchen partition camera information, determining the back kitchen partition type, and determining the analysis strategy according to the back kitchen partition type; analyzing the back kitchen partition camera information according to the analysis strategy, judging whether a cross-triggering condition occurs, if a cross-triggering condition occurs, adjusting the analysis strategy in real time according to the cross-triggering condition, and analyzing the back kitchen partition camera information according to the adjusted analysis strategy to determine the analysis result; determining prompt information according to the analysis result, and controlling the voice prompt device according to the prompt information to prompt the back kitchen staff to make corresponding adjustments. The present application accurately analyzes and judges the back kitchen staff's behavior involving a single partition and the behavior involving at least two partitions, significantly reducing the risk of problems caused by human factors in the back kitchen.

Description

Technical Field

[0001] The present application relates to the field of intelligent monitoring technology, and in particular to an intelligent operation monitoring management system and method for a kitchen. Background Art

[0002] In recent years, the intelligent development of the catering industry has made significant progress. Among them, intelligent operation and monitoring of catering kitchens can improve the level of catering services while ensuring food safety and quality, thereby improving customer satisfaction.

[0003] However, the existing intelligent operation monitoring and management technology for the back kitchen only monitors the processing process of the food in the back kitchen through intelligent kitchen equipment, and is unable to accurately analyze and deal with the behaviors of various personnel in the back kitchen that are likely to cause problems, resulting in frequent problems caused by human factors in the back kitchen. Summary of the invention

[0004] The present application provides an intelligent operation monitoring management system and method for a back kitchen to solve the above technical problems.

[0005] In a first aspect, the present application provides an intelligent operation monitoring management method for a back kitchen, which is applied to an intelligent operation monitoring management device for a back kitchen, wherein the intelligent operation monitoring management device for a back kitchen includes at least one camera, a voice prompt device, and a server, and the method is applied to the server, and the method includes:

[0006] Acquire kitchen partition camera information through the at least one camera, analyze the kitchen partition camera information, determine the kitchen partition type, and determine the analysis strategy according to the kitchen partition type;

[0007] According to the analysis strategy, analyze the kitchen partition camera information to determine whether a cross-trigger condition occurs; if the cross-trigger condition occurs, adjust the analysis strategy in real time according to the cross-trigger condition, and analyze the kitchen partition camera information according to the adjusted analysis strategy to determine the analysis result;

[0008] According to the analysis result, prompt information is determined, and according to the prompt information, the voice prompt device is controlled to prompt the kitchen staff to make corresponding adjustments.

[0009] Through the method provided by the present application, it is possible to formulate analysis strategies for corresponding surveillance video information for different partitions in the back kitchen, analyze and judge whether the behavior of the monitored object will cause problems within a single partition or between at least two partitions, and adjust the analysis strategy in real time based on the results of the analysis and judgment, so that the analysis strategy can accurately analyze and judge the behavior of personnel in the back kitchen partition camera information involving a single partition and the behavior involving at least two partitions at the same time, and according to the analysis results, the corresponding back kitchen personnel can be prompted to regulate their behavior through a voice prompt device, so that the operations of personnel in each partition in the back kitchen are more standardized, which significantly reduces the risk of problems caused by human factors during the work in the back kitchen.

[0010] Optionally, determining the analysis strategy according to the kitchen partition type includes:

[0011] Analyze the camera information of the kitchen partition according to the kitchen partition type to determine the range information and monitoring objects;

[0012] Determining an analysis behavior set according to the range information and the monitoring object;

[0013] The analysis strategy is determined according to the range information, the monitoring object and the analysis behavior set.

[0014] By adopting the above technical solution, the information of the corresponding back kitchen partition is extracted according to the back kitchen partition type, and an analysis strategy is formulated for the surveillance camera information of the partition. Compared with the existing method of analyzing the entire back kitchen area monitoring, the analysis of the monitoring objects and their behaviors in different partitions is more targeted, thereby making the analysis results more accurate.

[0015] Optionally, analyzing the kitchen partition camera information according to the range information and the monitored object to determine an analysis behavior set includes:

[0016] Analyze the kitchen partition camera information according to the range information and the monitoring object, and extract the monitoring key point image;

[0017] Assume the monitoring key point image is , the gradient image of the monitoring key point is ,in, For the monitoring key point image of each size, is the number of dimensions, and the preset detection function is used to Perform detection and obtain the current size target box list , the target box lists of all sizes are fused through the following formula to obtain the target detection result:

[0018] ;

[0019] in, is the target detection result, is the number of dimensions, It is the current size target box list;

[0020] Determine monitoring key points according to the target detection results;

[0021] Matching the monitoring key point with a preset key point behavior set, and determining a single triggering behavior and a cross triggering behavior under the monitoring key point according to the matching result;

[0022] The single triggering behavior and the cross triggering behavior are taken as the analysis behavior set.

[0023] By adopting the above technical solution, in the process of identifying key monitoring points, the target detection results are calculated by using the key point gradient image, which solves the problem of low recognition accuracy of a large number of objects with high shape similarity but different sizes in the kitchen by the existing video analysis technology, and improves the speed and accuracy of identifying key monitoring points, so as to improve the accuracy of subsequent judgment on the behavior of the monitored object at the key monitoring points.

[0024] Analyzing the kitchen partition camera information according to the analysis strategy includes:

[0025] According to the range information, the kitchen camera information is divided into zones to determine the analysis area;

[0026] Tracking the monitored object in the analysis area, when the monitored object reaches the key monitoring point, analyzing the behavior of the monitored object at the key monitoring point, determining the action characteristics and action amplitude of the behavior; matching the behavior with the analysis behavior set according to the action characteristics and the action amplitude, and determining the behavior matching result;

[0027] Determining prompt information according to the analysis result includes:

[0028] The behavior matching result is analyzed, and if the behavior matches the single triggering behavior, a single matching result is obtained, and the first prompt information is determined according to the single matching result.

[0029] By adopting the above technical solution, by determining the action characteristics and action amplitude of the monitored object's behavior, it is judged whether the action of the monitored object matches the action in the preset action analysis set. Compared with the prior art method of simply judging the action characteristics of the monitored object's behavior, this is more rigorous. By introducing the method of action amplitude judgment, the accuracy of judging the action of the monitored object is improved, and the false alarm rate of the subsequent prompt process is reduced.

[0030] Optionally, if the cross-triggering condition occurs, adjusting the analysis strategy in real time according to the cross-triggering condition includes:

[0031] If the behavior matches the cross-triggering behavior, it is determined that a cross-triggering condition occurs, and a cross-matching result is obtained;

[0032] Determining a triggering time of the cross-triggering behavior according to the cross-matching result;

[0033] Determine a subsequent monitoring period according to the trigger time and the preset subsequent behavior monitoring duration;

[0034] Based on the behavior and the subsequent monitoring period, the analysis strategy is adjusted.

[0035] By adopting the above technical solution, after the monitored object makes a cross-triggering behavior, the subsequent detection period is determined by the trigger time and the preset subsequent behavior monitoring duration, and the analysis strategy is adjusted according to the subsequent monitoring period to accurately analyze and monitor the subsequent actions of the monitored object in the subsequent monitoring period corresponding to the back kitchen partition camera information.

[0036] Optionally, the method further comprises:

[0037] Analyze the behavior, and determine whether there is a related area based on the analysis result;

[0038] If the associated area exists, adjusting the analysis strategy according to the associated area;

[0039] If the associated region does not exist, a non-associated mark is generated;

[0040] The analysis strategy is adjusted according to the non-association marker.

[0041] By adopting the above technical scheme, different situations that may exist after the cross-triggering behavior of the monitored object is taken into consideration. From the perspective of whether to contact other partitions, the subsequent behavior of the cross-triggering behavior is divided into two situations: must contact other partitions and cannot contact other partitions. This enables the analysis strategy adjusted for the cross-triggering behavior to analyze and judge the subsequent behavior of the monitored object more comprehensively and accurately.

[0042] Optionally, determining prompt information according to the analysis result includes:

[0043] According to the analysis strategy, analyzing the kitchen partition camera information in the subsequent monitoring period;

[0044] If the associated area exists in the analysis strategy, acquiring and analyzing the associated area camera information of the subsequent monitoring period according to the analysis strategy to determine the associated analysis result;

[0045] Analyze the association analysis result to determine whether the monitored object appears in the associated area during the monitoring period. If not, determine the prompt information based on the association analysis result.

[0046] By adopting the above technical solution, when there is an associated area in the analysis strategy, it is analyzed and determined whether the monitored object appears in the associated area. If not, the prompt information is determined based on the monitored object and its behavior to remind the monitored object to go to the associated area to solve the problem caused previously, so that the working operation of the monitored object is more standardized.

[0047] Optionally, the method further includes:

[0048] If the non-correlation mark exists in the analysis strategy, acquiring and analyzing the camera information of other kitchen partitions during the monitoring period according to the analysis strategy to determine the non-correlation analysis result;

[0049] Analyze the non-correlation analysis result to determine whether the monitored object is close to other partitions of the back kitchen during the monitoring period. If so, determine the prompt information based on the non-correlation analysis result.

[0050] By adopting the above technical solution, when there are non-associated marks in the analysis strategy, it is analyzed and determined whether the monitored object is close to other partitions in the back kitchen. If close, the prompt information is determined based on the monitoring object and its behavior to remind the monitored object not to go to other partitions while performing the current behavior, so as to prevent the work operations of the monitored object from causing problems between at least two partitions in the back kitchen.

[0051] Optionally, the method further includes:

[0052] After analyzing the association analysis result, if the monitored object appears in the associated area during the monitoring period, analyzing the camera information of the associated area to determine whether the monitored object has stopped the behavior, and if so, determining the prompt information according to the monitored object and the behavior;

[0053] After analyzing the non-correlated analysis results, if the monitored object appears in other partitions of the back kitchen during the monitoring period, analyze the camera information of other partitions of the back kitchen to determine whether the monitored object has stopped the behavior. If not, determine the prompt information based on the monitored object and the behavior.

[0054] By adopting the above-mentioned technical scheme, on the basis of the classification discussion of cross-triggering behaviors in the above-mentioned technical scheme, a method of behavior termination judgment is further introduced to make the analysis and judgment of subsequent behaviors of the monitored objects more accurate, while reducing the false alarm rate and making the prompt information more in line with the actual status of the current monitored object.

[0055] In a second aspect, the present application provides an intelligent operation monitoring and management system for a back kitchen, the system comprising:

[0056] An analysis strategy determination module, for obtaining kitchen partition camera information through at least one camera, analyzing the kitchen partition camera information, determining the kitchen partition type, and determining the analysis strategy according to the kitchen partition type;

[0057] A behavior analysis module, for analyzing the kitchen partition camera information according to the analysis strategy, determining whether a cross-triggering condition occurs, and if the cross-triggering condition occurs, adjusting the analysis strategy in real time according to the cross-triggering condition, and analyzing the kitchen partition camera information according to the adjusted analysis strategy to determine the analysis result;

[0058] The prompt module determines prompt information according to the analysis result, and controls the voice prompt device to prompt the kitchen staff to make corresponding adjustments according to the prompt information.

[0059] Optionally, the analysis strategy determination module is specifically used to:

[0060] Analyze the camera information of the kitchen partition according to the kitchen partition type to determine the range information and monitoring objects;

[0061] Determining an analysis behavior set according to the range information and the monitoring object;

[0062] The analysis strategy is determined according to the range information, the monitoring object and the analysis behavior set.

[0063] Optionally, the analysis strategy determination module is specifically used to:

[0064] Analyze the kitchen partition camera information according to the range information and the monitoring object, and extract the monitoring key point image;

[0065] Assume the monitoring key point image is , the gradient image of the monitoring key point is ,in, For the monitoring key point image of each size, is the number of dimensions, and the preset detection function is used to Perform detection and obtain the current size target box list , the target box lists of all sizes are fused through the following formula to obtain the target detection result:

[0066] ;

[0067] in, is the target detection result, is the number of dimensions, It is the current size target box list;

[0068] Determine monitoring key points according to the target detection results;

[0069] Matching the monitoring key point with a preset key point behavior set, and determining a single triggering behavior and a cross triggering behavior under the monitoring key point according to the matching result;

[0070] The single triggering behavior and the cross triggering behavior are taken as the analysis behavior set.

[0071] Optionally, the behavior analysis module is specifically used to:

[0072] According to the range information, the kitchen camera information is divided into zones to determine the analysis area;

[0073] Tracking the monitored object in the analysis area, and when the monitored object reaches the key monitoring point, analyzing the behavior of the monitored object at the key monitoring point to determine the action characteristics and action amplitude of the behavior;

[0074] According to the action feature and the action amplitude, the action is matched with the analysis action set to determine a action matching result;

[0075] Determining prompt information according to the analysis result includes:

[0076] The behavior matching result is analyzed, and if the behavior matches the single triggering behavior, a single matching result is obtained, and the first prompt information is determined according to the single matching result.

[0077] Optionally, the behavior analysis module is specifically used to:

[0078] If the behavior matches the cross-triggering behavior, it is determined that a cross-triggering condition occurs, and a cross-matching result is obtained;

[0079] Determining a triggering time of the cross-triggering behavior according to the cross-matching result;

[0080] Determine a subsequent monitoring period according to the trigger time and the preset subsequent behavior monitoring duration;

[0081] Based on the behavior and the subsequent monitoring period, the analysis strategy is adjusted.

[0082] Optionally, the system further comprises an association analysis module, which is used to:

[0083] Analyze the behavior, and determine whether there is a related area based on the analysis result;

[0084] If the associated area exists, adjusting the analysis strategy according to the associated area;

[0085] If the associated region does not exist, a non-associated mark is generated;

[0086] The analysis strategy is adjusted according to the non-association marker.

[0087] Optionally, the association analysis module is specifically used to:

[0088] According to the analysis strategy, analyzing the kitchen partition camera information in the subsequent monitoring period;

[0089] If the associated area exists in the analysis strategy, acquiring and analyzing the associated area camera information of the subsequent monitoring period according to the analysis strategy to determine the associated analysis result;

[0090] Analyze the association analysis result to determine whether the monitored object appears in the associated area during the monitoring period. If not, determine the prompt information based on the association analysis result.

[0091] Optionally, the association analysis module is specifically used to:

[0092] If the non-correlation mark exists in the analysis strategy, acquiring and analyzing the camera information of other kitchen partitions during the monitoring period according to the analysis strategy to determine the non-correlation analysis result;

[0093] Analyze the non-correlation analysis result to determine whether the monitored object is close to other partitions of the back kitchen during the monitoring period. If so, determine the prompt information based on the non-correlation analysis result.

[0094] Optionally, the device further comprises a suspension analysis module, configured to:

[0095] After analyzing the association analysis result, if the monitored object appears in the associated area during the monitoring period, analyzing the camera information of the associated area to determine whether the monitored object has stopped the behavior, and if so, determining the prompt information according to the monitored object and the behavior;

[0096] After analyzing the non-correlated analysis results, if the monitored object appears in other partitions of the back kitchen during the monitoring period, analyze the camera information of the back kitchen partition to determine whether the monitored object has stopped the behavior. If not, determine the prompt information based on the monitored object and the behavior.

[0097] The above technical solution adopted in this application has the following advantages:

[0098] It is able to formulate analysis strategies for corresponding surveillance video information for different partitions in the back kitchen, analyze and judge whether the behavior of the monitored object will cause problems within a single partition or between at least two partitions, and adjust the analysis strategy in real time according to the results of the analysis and judgment, so that the analysis strategy can accurately analyze and judge the behavior of personnel in a single partition and the behavior involving at least two partitions in the back kitchen partition camera information at the same time, and can prompt the corresponding back kitchen personnel to regulate their behavior through voice prompt equipment according to the analysis results, so that the operations of personnel in each partition in the back kitchen are more standardized, which significantly reduces the risk of problems caused by human factors during the work in the back kitchen.

[0099] According to the type of back kitchen partition, the information of the corresponding back kitchen partition is extracted, and an analysis strategy for the surveillance camera information of the partition is formulated. Compared with the existing method of analyzing the monitoring of the entire back kitchen area, the analysis of the monitoring objects and their behaviors in different partitions is more targeted, making the analysis results more accurate.

[0100] In the process of identifying key monitoring points, the method of using key point gradient images to calculate target detection results solves the problem of low recognition accuracy of existing video analysis technology for a large number of objects with high similarity in shape but different sizes in the kitchen, improves the speed and accuracy of identifying key monitoring points, and thus improves the accuracy of subsequent judgment on the behavior of the monitored objects at the key monitoring points.

[0101] By determining the action characteristics and action amplitude of the monitored object's behavior, it is judged whether the action of the monitored object matches the action in the preset action analysis set. Compared with the existing technology of simply judging the action characteristics of the monitored object's behavior, this method is more rigorous. By introducing the method of action amplitude judgment, the accuracy of judging the action of the monitored object is improved, and the false alarm rate of the subsequent prompt process is reduced.

[0102] After the monitored object makes a cross-triggering behavior, the subsequent detection period is determined by the trigger time and the preset subsequent behavior monitoring duration. The analysis strategy is adjusted according to the subsequent monitoring period to accurately analyze and monitor the subsequent actions of the monitored object in the corresponding subsequent monitoring period in the back kitchen partition camera information.

[0103] Taking into account the different situations that may exist after the cross-triggering behavior of the monitored object, from the perspective of whether to contact with other partitions, the subsequent behavior of the cross-triggering behavior is divided into two situations: must contact with other partitions and cannot contact with other partitions. This makes the analysis strategy adjusted for the cross-triggering behavior able to analyze and judge the subsequent behavior of the monitored object more comprehensively and accurately.

[0104] When there are associated areas in the analysis strategy, analyze and determine whether the monitored object appears in the associated area. If not, determine the prompt information based on the monitored object and its behavior to remind the monitored object to go to the associated area to solve the problem caused previously, so that the work operation of the monitored object is more standardized.

[0105] When there are non-associated tags in the analysis strategy, analyze and determine whether the monitored object is close to other partitions in the back kitchen. If close, determine the prompt information based on the monitoring object and its behavior to remind the monitored object not to go to other partitions while performing the current behavior, so as to prevent the monitored object's work operations from causing problems between at least two partitions in the back kitchen. BRIEF DESCRIPTION OF THE DRAWINGS

[0106] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0107] Figure 1 A schematic diagram of an application scenario provided for an embodiment of the present application;

[0108] Figure 2 A flowchart of an intelligent operation monitoring and management method for a back kitchen provided in one embodiment of the present application;

[0109] Figure 3 A schematic diagram of the structure of an intelligent operation monitoring and management system for a back kitchen provided in one embodiment of the present application. DETAILED DESCRIPTION

[0110] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0111] In addition, the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article, unless otherwise specified, generally means that the associated objects before and after are in an "or" relationship.

[0112] The embodiments of the present application are further described in detail below in conjunction with the drawings in the specification.

[0113] Intelligent operation monitoring of restaurant kitchens can improve the level of catering services while ensuring food safety and quality, thereby increasing customer satisfaction.

[0114] However, the existing intelligent operation monitoring and management technology for the back kitchen only monitors the processing process of the food in the back kitchen through intelligent kitchen equipment, and is unable to accurately analyze and deal with the behaviors of various personnel in the back kitchen that are likely to cause problems, resulting in frequent problems caused by human factors in the back kitchen.

[0115] Based on this, the present application provides an intelligent operation monitoring management system and method for the back kitchen. According to different partitions in the back kitchen, an analysis strategy for the corresponding monitoring video information is formulated to analyze and judge whether the behavior of the monitored object will cause problems in a single partition or between at least two partitions, and according to the results of the analysis and judgment, the analysis strategy is adjusted in real time, so that the analysis strategy can accurately analyze and judge the behavior of personnel in the back kitchen partition camera information involving a single partition and the behavior involving at least two partitions at the same time, and can prompt the corresponding back kitchen personnel to standardize their behavior through a voice prompt device according to the analysis results, so that the operation of personnel in each partition in the back kitchen is more standardized, which significantly reduces the risk of problems caused by human factors during the back kitchen work process.

[0116] Figure 1 A schematic diagram of an application scenario provided by the present application. The server obtains the camera information of the back kitchen partitions through at least one camera, and applies the method of the present application to analyze the camera information of the back kitchen partitions, and determines the prompt information, so that the voice prompt device prompts the back kitchen staff to perform standardized operations according to the prompt information, so that the operations of the staff in each partition in the back kitchen are more standardized, which significantly reduces the risk of problems caused by human factors during the back kitchen work process. The specific implementation method can refer to the following embodiments.

[0117] Figure 2 This is a flow chart of an intelligent operation monitoring and management method for a back kitchen provided in an embodiment of the present application. The method of this embodiment can be applied to the server in the above scenario. Figure 2 As shown, the method includes:

[0118] S201. Obtain kitchen partition camera information through at least one camera, analyze the kitchen partition camera information, determine the kitchen partition type, and determine the analysis strategy according to the kitchen partition type.

[0119] The back kitchen partition camera information can be the video information of different areas in the back kitchen during the working hours.

[0120] The types of kitchen partitions include but are not limited to cooking area, food preparation area, cleaning area, storage area and food delivery area.

[0121] The analysis strategy may be a set of information used to analyze the camera information of the kitchen partitions.

[0122] The analysis strategy includes but is not limited to scope information, monitoring object points, and analysis behavior sets.

[0123] Specifically, in the process of using video analysis technology to analyze the back kitchen partition camera information and determine the back kitchen partition type, it is necessary to extract typical features in the back kitchen camera information. For example, there is usually a stove in the cooking area, so the stove can be used as a typical feature of the cooking area. However, in the process of using video analysis technology to extract and identify typical features, due to the large number of people walking around in the back kitchen during working hours, the video background information in the back kitchen partition camera information is cluttered, thereby reducing the accuracy of identifying typical features.

[0124] To solve the above problems, we first use video analysis to extract frames from the kitchen camera information, and calculate the difference image using formula (1):

[0125] (1);

[0126] in, is the difference image, is the image of the current frame in the kitchen camera information, It is the image of the previous frame in the kitchen camera information;

[0127] Then, the differential image is processed by formula (2) to obtain the processed image:

[0128] (2);

[0129] in, For the processed image, is the pixel value of the difference image, To distinguish the threshold;

[0130] Pixel value of the difference image It can be obtained by processing the difference image with the Sobel operator;

[0131] Calculate the discrimination threshold The difference image can be evenly divided into several small areas, and the local threshold of each small area can be calculated using formula (3):

[0132] (3);

[0133] in, is the local threshold, The difference image is represented by points The set of pixel values ​​within a small area centered on

[0134] Finally, the discrimination threshold is obtained by formula (4):

[0135] (4);

[0136] in, To distinguish the threshold, is a set of local thresholds;

[0137] According to the processed image It can accurately distinguish dynamic objects from static objects in the back kitchen partition video information. By identifying static objects, it can extract and identify typical features in the back kitchen partition, and then quickly determine the back kitchen partition type based on the typical features.

[0138] In the process of deriving the analysis strategy, it is necessary to take into account that the responsibilities of different kitchen partitions are different, which leads to differences in the personnel and actions that need to be monitored. This makes the scope information, monitoring key points and analysis behavior sets in the analysis strategies of different types of partitions different. Therefore, it is necessary to determine the analysis strategy according to the different types of kitchen partitions.

[0139] S202. Analyze the camera information of the kitchen partitions according to the analysis strategy to determine whether a cross-trigger condition occurs. If a cross-trigger condition occurs, adjust the analysis strategy in real time according to the cross-trigger condition, and analyze the camera information of the kitchen partitions according to the adjusted analysis strategy to determine the analysis result.

[0140] The cross-trigger condition may be that personnel in the current back kitchen partition take actions that affect the work in other partitions.

[0141] Specifically, in the process of monitoring the operations of personnel in the back kitchen partition, personnel in different partitions may behave in a way that affects the current partition, or behave in a way that affects the work of other partitions. In the above case, it is necessary to analyze and determine whether personnel in a single partition behave in a way that affects the work of the current partition, and at the same time determine whether personnel behave in a way that affects the work of other partitions during the monitoring process, so as to adjust the aforementioned analysis strategy based on the behavior of personnel, so that the analysis strategy can accurately analyze the behavior of personnel in a single partition and involving at least two partitions in the back kitchen partition camera information.

[0142] S203. Determine prompt information according to the analysis result, and control the voice prompt device to prompt the kitchen staff to make corresponding adjustments according to the prompt information.

[0143] The analysis results include but are not limited to the monitored object and the behaviors performed by the monitored object.

[0144] The prompt information may be suggestion information corresponding to the behavior of the monitored object.

[0145] The voice prompt device can be a portable voice receiving device that can be carried by kitchen staff, or it can be a voice output device installed in a fixed position in the kitchen. This application does not make specific limitations.

[0146] Specifically, when personnel in the back kitchen partition perform behaviors that affect the back kitchen work, the corresponding prompt information is determined by analyzing the monitoring objects and the behaviors performed by the monitored objects in the results, so as to promptly remind the corresponding monitored objects to pay attention to the current behaviors, so as to prevent the current behaviors of the monitored objects from continuing and causing negative impacts on the back kitchen work.

[0147] Through the method of the present application, it is possible to formulate analysis strategies for corresponding monitoring video information for different partitions in the back kitchen, analyze and judge whether the behavior of the monitored object will cause problems within a single partition or between at least two partitions, and adjust the analysis strategy in real time according to the results of the analysis and judgment, so that the analysis strategy can accurately analyze and judge the behavior of personnel in a single partition and the behavior involving at least two partitions in the back kitchen partition camera information at the same time, and according to the analysis results, the corresponding back kitchen personnel can be prompted to regulate their behavior through a voice prompt device, so that the operations of personnel in each partition in the back kitchen are more standardized, which significantly reduces the risk of problems caused by human factors during the work in the back kitchen.

[0148] In some embodiments, the range information and monitoring objects are determined according to the kitchen partition type; the analysis behavior set is determined according to the range information and monitoring objects; and the analysis strategy is determined according to the range information, monitoring objects and analysis behavior set.

[0149] The range information may be the size information of the range that needs to be monitored in the current back kitchen partition.

[0150] The monitoring object can be the personnel who need to be monitored in the current back kitchen partition.

[0151] The key monitoring points may be key areas in the current kitchen partition that need to be monitored.

[0152] The analysis behavior set may be a set of behaviors that need to be monitored intensively under the current monitoring key point.

[0153] Specifically, since the range information required to be monitored under different kitchen partition types is different, for example, the range that needs to be monitored in the cooking area must at least include all stoves, so the range information needs to be determined according to the kitchen partition type. After the kitchen partition type is obtained, the range information is confirmed by video analysis technology. For example, each stove in the cooking area is identified by video analysis, and the area where each stove is located is connected into a continuous area, and then the range information of the cooking area is determined according to the area;

[0154] Similarly, the monitoring objects under different types of kitchen partitions are also different. For example, the monitoring objects in the cooking area are mainly chefs, while the monitoring objects in the preparation area are mainly food processing personnel. Therefore, the monitoring objects also need to be determined according to the kitchen partition type. Video analysis technology is used to identify the appearance characteristics of personnel in different kitchen partitions. For example, chefs in the cooking area usually wear chef uniforms and chef hats, while food processing personnel in the preparation area usually wear uniform work clothes and headscarves. At the same time, the action characteristics of different personnel are identified, such as the cooking actions of chefs in the cooking area and the cutting actions of food processing personnel, so as to determine the monitoring objects under different kitchen partitions.

[0155] The method of this embodiment extracts the information corresponding to the back kitchen partition according to the back kitchen partition type, and formulates an analysis strategy for the surveillance camera information of the partition. Compared with the existing method of analyzing the entire back kitchen area monitoring, the analysis of the monitoring objects and their behaviors in different partitions is more targeted, thereby making the analysis results more accurate.

[0156] In some embodiments, the kitchen partition camera information is analyzed according to the range information and the monitored object, and the key monitoring point images are extracted;

[0157] Assume the monitoring key point image is , the gradient image of the monitoring key point is ,in, For each size of monitoring key point image, is the number of dimensions, and the preset detection function is used to Perform detection and obtain the current size target box list , the target box lists of all sizes are fused through formula (5) to obtain the target detection result:

[0158] (5);

[0159] in, is the target detection result, is the number of dimensions, It is the current size target box list;

[0160] According to the target detection results, the monitoring key points are determined; the monitoring key points are matched with the preset key point behavior set, and the single trigger behavior and cross trigger behavior under the monitoring key points are determined according to the matching results; the single trigger behavior and cross trigger behavior are used as the analysis behavior set.

[0161] The preset key point behavior set may be a set of a series of behaviors that need to be monitored intensively and will occur at different monitoring key points, obtained by statistically analyzing a large amount of historical monitoring data.

[0162] A single trigger behavior may be a behavior of the monitored object that affects the current single kitchen partition at the current monitoring key point.

[0163] The cross-triggering behavior may be a behavior in which the controlled object has an impact on at least two kitchen partitions at the current monitoring key point.

[0164] The preset detection function can adopt existing computer vision technology or machine learning algorithm.

[0165] Specifically, after confirming the range information of the back kitchen partition, since the range information is relatively broad area information, if the personnel operations in the area corresponding to the range information are directly monitored and analyzed, it is easy to cause inaccurate analysis results. Therefore, it is necessary to extract the monitoring key points in the area corresponding to the range information. For example, the chef's operation area next to the stove in the cooking area can be used as the monitoring key point in the cooking area, and the food processing personnel's operation area next to the cutting table in the preparation area can be used as the monitoring key point in the preparation area. However, when using video analysis technology to identify the monitoring key points in the back kitchen partition, due to the presence of a large number of objects with high shape similarity but different sizes in the back kitchen, such as pots and knives of different sizes, the existing video analysis technology is slow and has a high error rate in identifying the key points in the back kitchen partition.

[0166] Through the above detection method, the recognition accuracy of a large number of objects with high similarity in shape but different sizes in the kitchen can be improved during video analysis, thereby improving the speed and accuracy of identifying key monitoring points.

[0167] Furthermore, after determining the key monitoring points, since the actions that need to be monitored at different key monitoring points are different, it is necessary to match the type of key monitoring points with the preset key point behavior set to obtain the behaviors that need to be monitored at the current key monitoring points. The behaviors that need to be monitored include single trigger behaviors that will affect the current back kitchen partition and cross-trigger behaviors that will affect other back kitchen partitions.

[0168] The method of this embodiment, in the process of identifying key monitoring points, solves the problem of low recognition accuracy of existing video analysis technology for a large number of objects with high shape similarity but different sizes in the kitchen by using key point gradient images to calculate target detection results, improves the speed and accuracy of identifying key monitoring points, and thus improves the accuracy of subsequent judgment on the behavior of the monitored object at the key monitoring points.

[0169] In some embodiments, the camera information of the back kitchen is divided into zones according to the range information to determine the analysis area; the monitored object in the analysis area is tracked, and when the monitored object reaches the key monitoring point, the behavior of the monitored object at the key monitoring point is analyzed to determine the action characteristics and action amplitude of the behavior; according to the action characteristics and action amplitude, the behavior is matched with the analysis behavior set to determine the behavior matching result; the behavior matching result is analyzed, and if the behavior matches a single triggering behavior, a single matching result is obtained, and based on the single matching result, the first prompt information is determined.

[0170] The analysis area may be an area divided in the back kitchen partition camera information according to range information for analyzing the behavior of the monitored object.

[0171] The analysis behavior set includes but is not limited to the characteristic value and amplitude value of a single trigger action and the characteristic value and amplitude value of a cross-trigger behavior.

[0172] First, the coordinates of the key parts of the monitored person's body are recorded through video analysis technology. The coordinates of the key parts of the human body include but are not limited to the coordinates of the head, hands, and legs;

[0173] Formula (6) is used to describe the action feature vector of the monitored object:

[0174] (6);

[0175] in, is the action feature vector, is the coordinate of the first key part, and so on. is the number of key parts;

[0176] Then, the horizontal moving speed of the monitored object is calculated by formula (7):

[0177] (7);

[0178] in, The key parts of the monitored object are at the time point The speed in the horizontal direction, For key parts at time points The horizontal coordinate at For key parts at time points The horizontal coordinate at is the duration of the action;

[0179] At the same time, the vertical movement speed of the monitored object is calculated by formula (8):

[0180] (8);

[0181] in, The key parts of the monitored object are at the time point The vertical velocity, For key parts at time points The vertical coordinate of For key parts at time points The vertical coordinate of is the duration of the action;

[0182] Furthermore, the acceleration of the monitored object in the horizontal direction is calculated by formula (9):

[0183] (9);

[0184] in, For key parts at time points The horizontal acceleration of The key parts of the monitored object are at the time point The horizontal speed, To monitor the key parts of the object at the time point The horizontal speed, is the duration of the action;

[0185] At the same time, the acceleration of the monitored object in the vertical direction is calculated by formula (10):

[0186] (10);

[0187] in, For key parts at time points The vertical acceleration of To monitor the key parts of the object at the time point The vertical velocity, To monitor the key parts of the object at the time point The vertical velocity, is the duration of the action;

[0188] Specifically, by calculating the speed of the key parts of the monitored object in the horizontal and vertical directions per unit time, the motion characteristics of the monitored object can be obtained. At the same time, by calculating the acceleration of the key parts of the monitored object in the horizontal and vertical directions per unit time, the motion amplitude of the monitored object can be obtained. The motion characteristics and motion amplitude are respectively compared with the characteristic values ​​and amplitude values ​​of the single trigger action and cross trigger action in the preset action analysis set to determine whether the action of the monitored object matches the action in the preset action analysis set.

[0189] The method of this embodiment determines the action characteristics and action amplitude of the monitored object's behavior to determine whether the action of the monitored object matches the action in the preset action analysis set. Compared with the prior art method of simply judging the action characteristics of the monitored object's behavior, this method is more rigorous. By introducing the method of action amplitude judgment, the accuracy of judging the action of the monitored object is improved, and the false alarm rate of the subsequent prompt process is reduced.

[0190] In some embodiments, if the behavior matches the cross-triggering behavior, it is determined that a cross-triggering condition occurs and a cross-matching result is obtained; based on the cross-matching result, the triggering time of the cross-triggering behavior is determined; based on the triggering time and the preset subsequent behavior monitoring duration, the subsequent monitoring period is determined; based on the behavior and the subsequent monitoring period, the analysis strategy is adjusted.

[0191] The cross-matching results include but are not limited to the cross-triggering behavior and triggering time that successfully matches the behavior of the monitored object;

[0192] The preset duration of subsequent behavior monitoring can be a fixed period of time, or it can be adjusted according to the back-kitchen workflow. This application does not make any specific limitations.

[0193] Specifically, after the monitored object performs a cross-triggering behavior, it is necessary to focus on monitoring the actions of the monitored object in the subsequent period of time to prevent the cross-triggering behavior of the monitored object from having a negative impact on the back kitchen work.

[0194] Taking the food handlers in the food preparation area as an example, if the food handlers are holding knives to handle food, they need to temporarily take food from the storage area or send the processed food to the cooking area, but at this time the food handlers forget to put down the knives, and instead hold the knives back and forth between at least two back kitchen partitions. During this process, the knives in the hands of the food handlers may cause harm to the food handlers or other people in the back kitchen partitions. At this time, the action of the food handlers "walking with knives in hand" is a kind of cross-triggering behavior. After the food handlers make the above actions, determine the subsequent monitoring period and adjust the analysis strategy to monitor and analyze the continuous actions of the food handlers in the subsequent monitoring period.

[0195] In the method of this embodiment, after the monitored object makes a cross-triggering behavior, the subsequent detection period is determined by the trigger time and the preset subsequent behavior monitoring duration, and the analysis strategy is adjusted according to the subsequent monitoring period to accurately analyze and monitor the subsequent actions of the monitored object in the subsequent monitoring period corresponding to the back kitchen partition camera information.

[0196] In some embodiments, the analysis behavior determines whether there is a correlation area based on the analysis results; if there is a correlation area, the analysis strategy is adjusted based on the correlation area; if there is no correlation area, a non-correlation mark is generated; and the analysis strategy is adjusted based on the non-correlation mark.

[0197] The associated area may be other partitioned areas of the kitchen that may be associated with the current behavior of the monitored object.

[0198] The non-association flag may be information used to express that the current cross-triggering behavior cannot be associated with other partitions.

[0199] Specifically, after determining that the monitored object has made a cross-triggering behavior, determine whether the current behavior must be linked to other areas of the back kitchen based on the behavior of the monitored object. Taking the food processing staff in the preparation area as an example, if the food processing staff accidentally drops the food to the ground during the processing of the food, at this time, in order to ensure food safety, the food processing staff should take the dropped food to the clean area to wash the food. At this time, the "food dropping" action made by the food processing staff has an associated area, and the associated area is the clean area. According to the associated area, adjust the analysis strategy to monitor whether the subsequent behavior of the food processing staff complies with the back kitchen work specifications.

[0200] However, the cross-triggering behavior of the monitored object may not be able to be linked to other areas of the back kitchen. Taking the cross-triggering behavior of the food handler "walking with a knife in hand" in the aforementioned embodiment as an example, when the food handler "walks with a knife in hand", the food handler will not be able to go to any other area of ​​the back kitchen, otherwise the risk of personal injury caused by the food handler's movement will increase. At this time, according to the cross-triggering behavior of the food handler "walking with a knife in hand", a non-associated mark is generated, and the analysis strategy is adjusted according to the non-associated mark to monitor whether the subsequent behavior of the monitored object complies with the back kitchen work specifications.

[0201] The method of this embodiment takes into account the different situations that may exist after the cross-triggering behavior is performed by the monitored object. From the perspective of whether to contact other partitions, the subsequent behavior of the cross-triggering behavior is divided into two situations: must contact other partitions and cannot contact other partitions. This enables the analysis strategy adjusted for the cross-triggering behavior to analyze and judge the subsequent behavior of the monitored object more comprehensively and accurately.

[0202] In some embodiments, according to the analysis strategy, the back kitchen partition camera information of the subsequent monitoring period is analyzed; if there is an associated area in the analysis strategy, the associated area camera information of the subsequent monitoring period is obtained and analyzed according to the analysis strategy to determine the associated analysis result; the associated analysis result is analyzed to determine whether the monitored object appears in the associated area during the monitoring period. If not, the prompt information is determined according to the associated analysis result.

[0203] The correlation analysis result may be analysis information obtained after analyzing the video information of the correlation area.

[0204] Specifically, if there is an associated area in the analysis strategy, it means that the cross-triggering behavior made by the current monitored object must be associated with the back kitchen partition corresponding to the associated area in the analysis strategy to comply with the back kitchen work specifications. At this time, it is necessary to obtain the camera information of the associated area in the current analysis strategy, and track the current monitored object through the target tracking technology in the video analysis process. If the current monitored object does not appear in the associated area during the subsequent detection period, it means that the current monitored object did not go to the associated area to solve the problem caused by the cross-triggering behavior after making the cross-triggering behavior. At this time, according to the association analysis results, it is determined that the prompt information is to prompt the monitored object to go to the associated area as soon as possible to solve the problem caused by the cross-triggering behavior.

[0205] In the method of this embodiment, when there is an associated area in the analysis strategy, it is analyzed and determined whether the monitored object appears in the associated area. If not, based on the monitored object and its behavior, a prompt message is determined to remind the monitored object to go to the associated area to solve the problem caused previously, so that the working operation of the monitored object is more standardized.

[0206] In some embodiments, if there is a non-correlation mark in the analysis strategy, the camera information of other partitions of the back kitchen during the monitoring period is obtained and analyzed according to the analysis strategy to determine the non-correlation analysis results; the non-correlation analysis results are analyzed to determine whether the monitored object is close to other partitions of the back kitchen during the monitoring period. If close, the prompt information is determined based on the non-correlation analysis results.

[0207] Specifically, if there is a non-association mark in the analysis strategy, it means that the cross-triggering behavior performed by the current monitored object cannot be associated with other partitions in the back kitchen, otherwise it will cause problems. At this time, it is necessary to obtain the camera information of all other partitions in the back kitchen, and track the monitored object through the target tracking technology in the video analysis process. If the monitored object approaches other partitions in the back kitchen during the subsequent detection period, it means that the risk caused by the cross-triggering behavior at this time has increased significantly. According to the current non-association analysis results, it is determined that the prompt information is to prompt the monitored object not to go to other partitions while performing the current behavior.

[0208] The method of this embodiment, when there is a non-associated mark in the analysis strategy, analyzes and determines whether the monitored object is close to other partitions in the back kitchen. If close, then based on the monitoring object and its behavior, determine the prompt information to remind the monitored object not to go to other partitions while performing the current behavior, so as to prevent the working operation of the monitored object from causing problems between at least two partitions in the back kitchen.

[0209] In some embodiments, after analyzing the correlation analysis results, if the monitored object appears in the associated area during the monitoring period, the camera information of the associated area is analyzed to determine whether the monitored object has stopped the behavior. If so, the prompt information is determined based on the monitored object and the behavior. After analyzing the non-correlation analysis results, if the monitored object approaches other partitions of the back kitchen during the monitoring period, the camera information of the back kitchen partition is analyzed to determine whether the monitored object has stopped the behavior. If not, the prompt information is determined based on the monitored object and the behavior.

[0210] Specifically, when there are associated areas in the analysis strategy, by judging whether the monitored object appears in the associated area during the monitoring period in the aforementioned embodiment, although it can be determined that when the monitored object does not appear in the associated area, the problem caused by the cross-triggering behavior must not be solved, it is impossible to determine that the problem caused by the cross-triggering behavior has been solved when the monitored object appears in the associated area. For example, when the food handler in the food preparation area drops food on the ground due to misoperation, the ideal handling method is that the food handler carries the food dropped on the ground to the cleaning area to complete the cleaning of the food. However, there is a situation where the food handler goes to the cleaning area to perform other work after performing the aforementioned cross-triggering behavior, and does not carry the food dropped on the ground. At this time, it is necessary to analyze whether the food handler's behavior in the cleaning area is to clean the aforementioned food based on the camera information of the cleaning area. If not, it means that even if the food handler appears in the associated area, the problem caused by the aforementioned cross-triggering behavior has not been solved. At this time, it is judged that the food handler has terminated the behavior. According to the monitored object and behavior, it is determined that the prompt information is to remind the food handler that the food dropped before was not cleaned in time, and the food handler should take the food to the cleaning area as soon as possible to clean the food.

[0211] Similarly, when there are non-associated marks in the analysis strategy, the method of judging whether the monitored object is close to other partitions of the back kitchen during the monitoring period in the aforementioned embodiment can prompt the monitored object in time to avoid the monitored object causing problems between at least two partitions when performing cross-triggering behavior. However, if the monitored object has terminated the cross-triggering behavior in advance when approaching other partitions of the back kitchen, it is easy to cause false alarms. Therefore, while judging whether the monitored object is close to other partitions of the back kitchen during the monitoring period, it is further judged whether the monitored object has terminated the cross-triggering behavior in advance when approaching other partitions of the back kitchen. If not, it can be determined that the current behavior of the monitored object still has the risk of causing problems in at least two partitions. At this time, according to the monitored object and behavior, the monitored object is prompted to terminate the cross-triggering behavior before going to other partitions to avoid causing problems between at least two partitions in the back kitchen.

[0212] The method of this embodiment, based on the classification discussion of cross-triggering behaviors in the previous embodiment, further introduces a method of behavior termination judgment, so that the analysis and judgment of subsequent behaviors of the monitored object are more accurate, while reducing the false alarm rate, so that the prompt information is more in line with the actual status of the current monitored object.

[0213] Figure 3 An intelligent operation monitoring and management system for a kitchen provided in one embodiment of the present application is provided. Figure 3 As shown, the intelligent operation monitoring and management system 300 for the back kitchen of this embodiment includes: an analysis strategy determination module 301, a behavior analysis module 302 and a prompt module 303.

[0214] Analysis strategy determination module 301, for obtaining kitchen partition camera information through at least one camera, analyzing the kitchen partition camera information, determining the kitchen partition type, and determining the analysis strategy according to the kitchen partition type;

[0215] The behavior analysis module 302 is used to analyze the kitchen partition camera information according to the analysis strategy, determine whether a cross-triggering condition occurs, and if the cross-triggering condition occurs, adjust the analysis strategy in real time according to the cross-triggering condition, and analyze the kitchen partition camera information according to the adjusted analysis strategy to determine the analysis result;

[0216] The prompt module 303 determines prompt information according to the analysis result, and controls the voice prompt device to prompt the kitchen staff to make corresponding adjustments according to the prompt information.

[0217] Optionally, the analysis strategy determination module 301 is specifically used to:

[0218] Analyze the camera information of the kitchen partition according to the kitchen partition type to determine the range information and monitoring objects;

[0219] Determining an analysis behavior set according to the range information and the monitoring object;

[0220] The analysis strategy is determined according to the range information, the monitoring object and the analysis behavior set.

[0221] Optionally, the analysis strategy determination module 301 is specifically used to:

[0222] Analyze the kitchen partition camera information according to the range information and the monitoring object, and extract the monitoring key point image;

[0223] Assume the monitoring key point image is , the gradient image of the monitoring key point is ,in, For the monitoring key point image of each size, is the number of dimensions, and the preset detection function is used to Perform detection and obtain the current size target box list , the target box lists of all sizes are fused through the following formula to obtain the target detection result:

[0224] ;

[0225] in, is the target detection result, is the number of dimensions, It is the current size target box list;

[0226] Determine monitoring key points according to the target detection results;

[0227] Matching the monitoring key point with a preset key point behavior set, and determining a single triggering behavior and a cross triggering behavior under the monitoring key point according to the matching result;

[0228] The single triggering behavior and the cross triggering behavior are taken as the analysis behavior set.

[0229] Optionally, the behavior analysis module 302 is specifically used to:

[0230] According to the range information, the kitchen camera information is divided into zones to determine the analysis area;

[0231] Tracking the monitored object in the analysis area, and when the monitored object reaches the key monitoring point, analyzing the behavior of the monitored object at the key monitoring point to determine the action characteristics and action amplitude of the behavior;

[0232] According to the action feature and the action amplitude, the action is matched with the analysis action set to determine a action matching result;

[0233] Determining prompt information according to the analysis result includes:

[0234] The behavior matching result is analyzed, and if the behavior matches the single triggering behavior, a single matching result is obtained, and the first prompt information is determined according to the single matching result.

[0235] Optionally, the behavior analysis module 302 is specifically used to:

[0236] If the behavior matches the cross-triggering behavior, it is determined that a cross-triggering condition occurs, and a cross-matching result is obtained;

[0237] Determining a triggering time of the cross-triggering behavior according to the cross-matching result;

[0238] Determine a subsequent monitoring period according to the trigger time and the preset subsequent behavior monitoring duration;

[0239] Based on the behavior and the subsequent monitoring period, the analysis strategy is adjusted.

[0240] Optionally, the system further includes a correlation analysis module 304, which is used to:

[0241] Analyze the behavior, and determine whether there is a related area based on the analysis result;

[0242] If the associated area exists, adjusting the analysis strategy according to the associated area;

[0243] If the associated region does not exist, a non-associated mark is generated;

[0244] The analysis strategy is adjusted according to the non-association marker.

[0245] Optionally, the association analysis module 304 is specifically used to:

[0246] According to the analysis strategy, analyzing the kitchen partition camera information in the subsequent monitoring period;

[0247] If the associated area exists in the analysis strategy, acquiring and analyzing the associated area camera information of the subsequent monitoring period according to the analysis strategy to determine the associated analysis result;

[0248] Analyze the association analysis result to determine whether the monitored object appears in the associated area during the monitoring period. If not, determine the prompt information based on the association analysis result.

[0249] Optionally, the association analysis module 304 is specifically used to:

[0250] If the non-correlation mark exists in the analysis strategy, acquiring and analyzing the camera information of other kitchen partitions during the monitoring period according to the analysis strategy to determine the non-correlation analysis result;

[0251] Analyze the non-correlation analysis result to determine whether the monitored object is close to other partitions of the back kitchen during the monitoring period. If so, determine the prompt information based on the non-correlation analysis result.

[0252] Optionally, the device further comprises a suspension analysis module 305, configured to:

[0253] After analyzing the association analysis result, if the monitored object appears in the associated area during the monitoring period, analyzing the camera information of the associated area to determine whether the monitored object has stopped the behavior, and if so, determining the prompt information according to the monitored object and the behavior;

[0254] After analyzing the non-correlated analysis results, if the monitored object appears in other partitions of the back kitchen during the monitoring period, analyze the camera information of the back kitchen partition to determine whether the monitored object has stopped the behavior. If not, determine the prompt information based on the monitored object and the behavior.

[0255] The system of this embodiment can be used to execute the method of any of the above embodiments. The implementation principles and technical effects are similar and will not be described in detail here.

Claims

1. An intelligent operation monitoring and management method for a back kitchen, characterized in that: The invention is applied to an intelligent operation monitoring and management device for a back kitchen, the intelligent operation monitoring and management device for a back kitchen comprises at least one camera, a voice prompt device and a server, the intelligent operation monitoring and management method for a back kitchen is applied to the server, and the method comprises: Acquire the kitchen partition camera information through the at least one camera, analyze the kitchen partition camera information, determine the kitchen partition type, and determine the analysis strategy according to the kitchen partition type; According to the analysis strategy, the camera information of the kitchen partition is analyzed to determine whether a cross-triggering condition occurs. If the cross-triggering condition occurs, the analysis strategy is adjusted in real time according to the cross-triggering condition, and the camera information of the kitchen partition is analyzed according to the adjusted analysis strategy to determine the analysis result; the cross-triggering condition is that the personnel in the current kitchen partition take actions that affect the work of other partitions; Determine prompt information according to the analysis result, and control the voice prompt device to prompt the kitchen staff to make corresponding adjustments according to the prompt information; Determining the analysis strategy according to the kitchen partition type includes: Analyze the camera information of the kitchen partition according to the kitchen partition type to determine the range information and monitoring objects; Determining an analysis behavior set according to the range information and the monitoring object; Determining the analysis strategy according to the range information, the monitoring object and the analysis behavior set; The step of analyzing the kitchen partition camera information according to the range information and the monitored object to determine an analysis behavior set includes: Analyze the kitchen partition camera information according to the range information and the monitoring object, and extract the monitoring key point image; Assume the monitoring key point image is , the gradient image of the monitoring key point is ,in, For the monitoring key point image of each size, is the number of dimensions, and the preset detection function is used to Perform detection and obtain the current size target box list , the target box lists of all sizes are fused through the following formula to obtain the target detection result: in, is the target detection result, is the number of dimensions, It is the target box list of the current size; Determine monitoring key points according to the target detection results; Matching the monitoring key point with a preset key point behavior set, and determining a single triggering behavior and a cross triggering behavior under the monitoring key point according to the matching result; taking the single triggering behavior and the cross triggering behavior as the analysis behavior set; Analyzing the kitchen partition camera information according to the analysis strategy includes: Divide the kitchen camera information into different zones according to the range information to determine the analysis area; Tracking the monitored object in the analysis area, and when the monitored object reaches the key monitoring point, analyzing the behavior of the monitored object at the key monitoring point to determine the action characteristics and action amplitude of the behavior; According to the action feature and the action amplitude, the action is matched with the analysis action set to determine a action matching result; Determining prompt information according to the analysis result includes: Analyze the behavior matching result, if the behavior matches the single triggering behavior, obtain a single matching result, and determine the first prompt information according to the single matching result; If the cross-triggering condition occurs, adjusting the analysis strategy in real time according to the cross-triggering condition includes: If the behavior matches the cross-triggering behavior, it is determined that a cross-triggering condition occurs, and a cross-matching result is obtained; Determining a triggering time of the cross-triggering behavior according to the cross-matching result; Determine a subsequent monitoring period according to the trigger time and the preset subsequent behavior monitoring duration; Based on the behavior and the subsequent monitoring period, the analysis strategy is adjusted.

2. The method according to claim 1, characterized in that The method further comprises: Analyze the behavior, and determine whether there is a related area based on the analysis result; If the associated area exists, adjusting the analysis strategy according to the associated area; If the associated region does not exist, a non-associated mark is generated; The analysis strategy is adjusted according to the non-association marker.

3. The method according to claim 2, characterized in that Determining prompt information according to the analysis result includes: According to the analysis strategy, analyzing the kitchen partition camera information in the subsequent monitoring period; If the associated area exists in the analysis strategy, acquiring and analyzing the associated area camera information of the subsequent monitoring period according to the analysis strategy to determine the associated analysis result; Analyze the association analysis result to determine whether the monitored object appears in the associated area during the monitoring period. If not, determine the prompt information based on the association analysis result.

4. The method according to claim 3, characterized in that Determining prompt information according to the analysis result includes: If the non-correlation mark exists in the analysis strategy, acquiring and analyzing the camera information of other kitchen partitions during the monitoring period according to the analysis strategy to determine the non-correlation analysis result; Analyze the non-correlation analysis result to determine whether the monitored object is close to other partitions of the back kitchen during the monitoring period. If so, determine the prompt information based on the non-correlation analysis result.

5. The method according to claim 4, characterized in that The method further comprises: After analyzing the association analysis result, if the monitored object appears in the associated area during the monitoring period, analyzing the camera information of the associated area to determine whether the monitored object has stopped the behavior, and if so, determining the prompt information according to the monitored object and the behavior; After analyzing the non-correlated analysis results, if the monitored object appears in other partitions of the back kitchen during the monitoring period, analyze the camera information of the back kitchen partition to determine whether the monitored object has stopped the behavior. If not, determine the prompt information based on the monitored object and the behavior.

6. An intelligent operation monitoring and management system for kitchen, characterized in that: The method as claimed in any one of claims 1 to 5 comprises: An analysis strategy determination module, for obtaining kitchen partition camera information through at least one camera, analyzing the kitchen partition camera information, determining the kitchen partition type, and determining the analysis strategy according to the kitchen partition type; The behavior analysis module is used to analyze the camera information of the kitchen partition according to the analysis strategy, determine whether a cross-triggering condition occurs, and if the cross-triggering condition occurs, adjust the analysis strategy in real time according to the cross-triggering condition, and analyze the camera information of the kitchen partition according to the adjusted analysis strategy to determine the analysis result; the cross-triggering condition is that the personnel in the current kitchen partition take a behavior that affects the work of other partitions; A prompt module determines prompt information according to the analysis result, and controls a voice prompt device to prompt kitchen staff to make corresponding adjustments according to the prompt information; The analysis strategy determination module is specifically used for: Analyze the camera information of the kitchen partition according to the kitchen partition type to determine the range information and monitoring objects; Determining an analysis behavior set according to the range information and the monitoring object; Determining the analysis strategy according to the range information, the monitoring object and the analysis behavior set; The analysis strategy determination module is specifically used for: Analyze the kitchen partition camera information according to the range information and the monitoring object, and extract the monitoring key point image; Assume the monitoring key point image is , the gradient image of the monitoring key point is ,in, For the monitoring key point image of each size, is the number of dimensions, and the preset detection function is used to Perform detection and obtain the current size target box list , the target box lists of all sizes are fused through the following formula to obtain the target detection result: in, is the target detection result, is the number of dimensions, It is the target box list of the current size; Determine monitoring key points according to the target detection results; Matching the monitoring key point with a preset key point behavior set, and determining a single triggering behavior and a cross triggering behavior under the monitoring key point according to the matching result; The single triggering behavior and the cross triggering behavior are taken as the analysis behavior set.

Images