An intelligent analysis method and system for employee behavior in a catering scenario

Abstract

The present invention relates to the field of image processing, and in particular to an intelligent analysis method and system for employee behavior in a catering scene. It includes: denoising each video frame, performing edge detection based on the denoised video frame, calculating the matching value through a template matching algorithm based on the video frame after edge detection, and calculating the motion trajectory score through a nonlinear spatiotemporal smoothing motion trajectory analysis algorithm; obtaining a comprehensive score through weighted calculation based on the matching value and the motion trajectory score; and judging the employee behavior by setting a comprehensive score threshold. It solves the problem that in a catering scene, the video frames collected by the camera are often affected by environmental noise, resulting in noise in the video frames; in a catering scene with complex lighting conditions, it is difficult to accurately extract edge information in the video frame; and it is difficult to accurately match specific actions or behaviors in the video frame due to the operator's motion offset, lighting changes, and partial occlusion.

Description

Technical Field

[0001] The present invention relates to the field of image processing, and in particular to an intelligent analysis method and system for employee behavior in a catering scenario. Background Art

[0002] As an important branch of the service industry, the catering industry involves aspects such as personnel management, food material management, food processing, equipment management, hygiene management and safety management. Each link has an important impact on the final service quality and food safety. However, due to the high mobility and high staff density in the catering industry, the traditional management method relies on manual inspection and supervision, which is not only inefficient but also prone to omissions, which will bring potential risks to food safety and catering service quality. The rise of intelligent analysis technology enables catering companies to control food safety risks, improve management efficiency, and enhance the overall service quality. With the continuous advancement of technology and the gradual promotion of application scenarios, intelligent analysis systems will play an increasingly important role in the catering industry, reduce the occurrence of food safety incidents, and enhance consumers' trust in catering companies. It has broad application prospects and market value.

[0003] In catering scenarios, the existing intelligent analysis systems have the following technical problems: the video frames captured by the camera are often affected by environmental noise, including light changes, camera shake, and background clutter, resulting in noise in the video frames; in catering scenarios with complex lighting conditions, it is difficult to accurately extract edge information in the video frames, affecting the behavior recognition effect; due to the operator's motion offset, lighting changes, and partial occlusion, it is difficult to accurately match specific actions or behaviors in the video frames, resulting in the accuracy and reliability of behavior recognition being affected. Summary of the invention

[0004] The present invention provides an intelligent analysis method and system for employee behaviors in a catering scene, so as to solve the problems that in a catering scene, video frames captured by a camera are often affected by environmental noise, including light changes, camera shaking, and background clutter, resulting in noise in the video frames; in a catering scene with complex lighting conditions, it is difficult to accurately extract edge information in the video frames, affecting the effect of behavior recognition; and due to the operator's motion offset, lighting changes, and partial occlusion, it is difficult to accurately match specific actions or behaviors in the video frames, resulting in the accuracy and reliability of behavior recognition being affected.

[0005] The present invention provides an intelligent analysis method and system for employee behavior in a catering scenario, which specifically includes the following technical solutions:

[0006] An intelligent analysis method for employee behavior in a catering scenario includes the following steps:

[0007] S1, collecting video frames to generate a continuous video frame sequence; performing denoising on each video frame in the video frame sequence to obtain a denoised video frame;

[0008] S2, based on the denoised video frame, edge detection is performed by using an adaptive pixel-to-pixel interactive edge detection algorithm to obtain a video frame after edge detection;

[0009] S3, based on the video frame after edge detection, the matching value is calculated by the template matching algorithm, and the motion trajectory score is calculated by the nonlinear spatiotemporal smoothing motion trajectory analysis algorithm;

[0010] S4. Based on the matching value and the motion trajectory score, a comprehensive score is obtained through weighted calculation; a comprehensive score threshold is set, and the employee behavior is judged by comparing the comprehensive score with the comprehensive score threshold.

[0011] Preferably, the S1 specifically includes:

[0012] The implementation process of the denoising process is as follows:

[0013] In the first step, a perturbation value is generated for each pixel of each video frame in the video frame sequence;

[0014] The second step is to calculate the disturbance response value under each disturbance condition based on the disturbance value;

[0015] The third step is to calculate the stability index while generating the disturbance response value;

[0016] The fourth step is to determine whether to perform denoising on the video frame pixels based on the stability index.

[0017] Preferably, the S2 specifically includes:

[0018] In the process of implementing the adaptive pixel interaction edge detection algorithm, adaptive spatial pixel interaction function and adaptive spatial difference analysis are introduced.

[0019] Preferably, the S2 specifically includes:

[0020] The adaptive spatial pixel interaction function obtains the interaction strength between the current pixel and its neighboring pixels by analyzing each pixel in the denoised video frame.

[0021] Preferably, the S2 specifically includes:

[0022] The adaptive spatial difference analysis performs weighted calculation on the interaction strength between the current pixel and all neighboring pixels to obtain an adaptive spatial difference value; based on the adaptive spatial difference value, an edge judgment is performed on each pixel in the denoised video frame.

[0023] Preferably, the S3 specifically includes:

[0024] The template matching algorithm performs adaptive weight adjustment on the video frame after edge detection, aligns local features of the template obtained from the industry standard action library, and calculates the matching value between the video frame after edge detection and the template.

[0025] Preferably, the S3 specifically includes:

[0026] Based on the video frames after edge detection, the grayscale changes between the current frame and the next frame are compared through the nonlinear spatiotemporal smoothing motion trajectory analysis algorithm, and the movement direction and speed of the pixels in the matching area of ​​the video frames after edge detection are estimated. At the same time, the nonlinear function is combined for smoothing to obtain the motion vector; the motion vectors of the pixel points at all pixel positions in the matching area of ​​the video frames after edge detection are weighted summed to obtain the motion trajectory score.

[0027] An intelligent analysis system for employee behavior in a catering scenario, including the following parts:

[0028] Video frame acquisition module, video frame denoising processing module, edge detection module, template matching module, motion trajectory analysis module, comprehensive score calculation module, behavior determination module;

[0029] Video frame acquisition module: performs video frame acquisition, generates a continuous video frame sequence, and outputs the video frame sequence to the video frame denoising processing module;

[0030] Video frame denoising processing module: performs denoising processing on each video frame in the video frame sequence to obtain a denoised video frame, and outputs the denoised video frame to the edge detection module;

[0031] Edge detection module: Based on the denoised video frame, edge detection is performed through an adaptive pixel-to-pixel interactive edge detection algorithm to obtain the video frame after edge detection, and the video frame after edge detection is output to the template matching module and the motion trajectory analysis module;

[0032] Template matching module: Based on the video frame after edge detection, the template matching algorithm is used to calculate the matching value, and the matching value is output to the comprehensive score calculation module;

[0033] Motion trajectory analysis module: Based on the video frames after edge detection, the motion trajectory score is calculated by nonlinear spatiotemporal smoothing motion trajectory analysis algorithm, and the motion trajectory score is output to the comprehensive score calculation module;

[0034] Comprehensive score calculation module: calculates the comprehensive score based on the matching value and motion trajectory score, and outputs the comprehensive score to the behavior judgment module;

[0035] Behavior judgment module: Set a comprehensive score threshold and judge employee behavior by comparing the comprehensive score with the comprehensive score threshold.

[0036] The beneficial effects of the technical solution of the present invention are:

[0037] 1. By performing disturbance generation, disturbance response calculation and stability analysis on each video frame pixel, the noise pixels and real information pixels are effectively distinguished. By setting the stability threshold, it is decided whether to perform denoising on the video frame pixels, thereby removing the noise in the video frame, retaining useful detail information, improving the quality of the video frame, reducing noise interference, and ensuring the accuracy and reliability of subsequent processing steps.

[0038] 2. After denoising, an adaptive pixel-to-pixel interaction edge detection algorithm is used. By introducing adaptive spatial pixel interaction functions and adaptive spatial difference analysis, the edge detection capability is significantly improved. By analyzing the color difference and spatial distance between the current pixel and the neighboring pixels, the edge features in the video frame are accurately identified. In the catering scene, precise edge detection helps to accurately identify and separate different objects.

[0039] 3. Based on the video frames after edge detection, the template matching algorithm with adaptive weight adjustment and local feature alignment is adopted to effectively deal with the offset, lighting change and occlusion problems in the video frames. By increasing the weight value influence on the key area, the accuracy and robustness of the matching are improved, which provides strong support for behavior analysis and violation detection in catering scenarios.

[0040] 4. By using a nonlinear spatiotemporal smoothing motion trajectory analysis algorithm, the time and space information is integrated, and the motion vectors of the pixel points in the matching area are nonlinearly smoothed, effectively filtering out abnormal movements caused by lighting changes or subtle noises; the use of the hyperbolic tangent function (tanh) smoothes the mutations of the motion vector, ensuring the continuity and stability of the motion trajectory, thereby improving the accuracy and reliability of the motion trajectory analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 This is a structural diagram of an intelligent analysis system for employee behavior in a catering scenario according to the present invention;

[0042] Figure 2 This is a flow chart of the intelligent analysis method of employee behavior in a catering scenario described in the present invention. DETAILED DESCRIPTION

[0043] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0044] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0045] The specific scheme of the intelligent analysis method and system of employee behavior in a catering scenario provided by the present invention is described in detail below with reference to the accompanying drawings.

[0046] Refer to the attached Figure 1 , which shows a structure diagram of an intelligent analysis system for employee behavior in a catering scenario provided by an embodiment of the present invention, the system includes the following parts:

[0047] Video frame acquisition module, video frame denoising processing module, edge detection module, template matching module, motion trajectory analysis module, comprehensive score calculation module, behavior determination module;

[0048] Video frame acquisition module: It acquires video frames through high-definition cameras deployed in the dining scene, generates continuous video frame sequences, and outputs the video frame sequences to the video frame denoising processing module;

[0049] Video frame denoising processing module: receives the video frame sequence from the video frame acquisition module, performs denoising on each video frame in the video frame sequence to obtain a denoised video frame, and outputs the denoised video frame to the edge detection module;

[0050] Edge detection module: receives the denoised video frame from the video frame denoising processing module, performs edge detection through an adaptive pixel-to-pixel interactive edge detection algorithm to obtain the video frame after edge detection, and outputs the video frame after edge detection to the template matching module and the motion trajectory analysis module;

[0051] Template matching module: receives the video frame after edge detection from the edge detection module, calculates the matching value through the template matching algorithm, and outputs the matching value to the comprehensive score calculation module;

[0052] Motion trajectory analysis module: receives the video frames after edge detection from the edge detection module, calculates the motion trajectory score through the nonlinear spatiotemporal smoothing motion trajectory analysis algorithm, and outputs the motion trajectory score to the comprehensive score calculation module;

[0053] Comprehensive score calculation module: receives the matching value from the template matching module and the motion trajectory score from the motion trajectory analysis module, obtains a comprehensive score through weighted calculation, and outputs the comprehensive score to the behavior determination module;

[0054] Behavior determination module: receives the comprehensive score from the comprehensive score calculation module, sets the comprehensive score threshold, and determines the employee behavior based on the comparison between the comprehensive score and the comprehensive score threshold.

[0055] Refer to the attached Figure 2 , which shows a flow chart of an intelligent analysis method for employee behavior in a catering scenario provided by an embodiment of the present invention, the method comprising the following steps:

[0056] S1, collecting video frames to generate a continuous video frame sequence; performing denoising on each video frame in the video frame sequence to obtain a denoised video frame;

[0057] The high-definition cameras deployed in the dining scene are used to collect video frames and generate a continuous video frame sequence {I1, I2, …, I n}, n represents the number of video frames, I i represents the i-th video frame, i∈[1,n]; denoising is performed on each video frame in the video frame sequence, and the specific steps are as follows:

[0058] The first step is disturbance generation;

[0059] A perturbation value is generated for each pixel of the video frame, and the perturbation value is generated by a pseudo-random generator PRNG:

[0060]

[0061] in, Represents the disturbance value, that is, for the i-th video frame pixel I i (x, y) is the offset value of the jth perturbation; PRNG(x, y, j) represents the perturbation value generated by the pseudo-random generator PRNG according to the video frame pixel position (x, y) and the number of perturbations j;

[0062] The second step is disturbance response calculation;

[0063] Based on the disturbance value, calculate the disturbance response value under each disturbance condition:

[0064]

[0065] in, Represents the disturbance response value, i.e., the pixel I of the i-th video frame i (x, y) is the pixel value under the jth perturbation; I i(x,y) represents the pixel of the i-th video frame; Represents the disturbance value, that is, for the i-th video frame pixel I i (x,y) is the offset value under the jth disturbance;

[0066] The third step is stability analysis;

[0067] While generating the disturbance response value, the average response value is calculated separately and standard deviation σ i (x,y), to assess stability:

[0068]

[0069] in, represents the average response value of the pixels in the i-th video frame; m represents the number of disturbances; Represents the disturbance response value, i.e., the pixel I of the i-th video frame i (x, y) is the pixel value under the jth perturbation;

[0070]

[0071] Among them, σ i (x,y) is the standard deviation, which represents the stability index, that is, the degree of dispersion of the disturbance response value under different disturbances;

[0072] The fourth step is denoising;

[0073] According to the stability index σ i (x,y), decide whether to denoise the video frame pixels, the formula is as follows:

[0074]

[0075] Among them, I' i (x,y) represents the pixel value of the i-th denoised video frame; Indicates the stability threshold, which is used to determine whether the video frame pixels need to be denoised. It can be set according to the specific implementation scenario and is not limited here;

[0076] By perturbing each pixel of the video frame and analyzing the stability, it is possible to effectively distinguish between noise pixels and real information pixels, and to accurately remove noise without losing the details of the video frame, thus obtaining the denoised video frame I' i .

[0077] S2, based on the denoised video frame, edge detection is performed by using an adaptive pixel-to-pixel interactive edge detection algorithm to obtain a video frame after edge detection;

[0078] Based on the denoised video frame, an adaptive pixel-to-pixel interaction edge detection algorithm is used to perform edge detection to obtain a video frame after edge detection; the adaptive pixel-to-pixel interaction edge detection algorithm improves the edge detection capability by introducing an adaptive spatial pixel interaction function and an adaptive spatial difference analysis, and performs edge judgment on each pixel in the denoised video frame by setting an edge detection threshold. For each pixel in the denoised video frame, when the adaptive spatial difference value exceeds the edge detection threshold, the current pixel is judged as an edge pixel, otherwise it is a non-edge pixel, and a video frame after edge detection is obtained; wherein, the edge detection threshold can be set according to the specific implementation scenario and is not limited here.

[0079] For each pixel in the denoised video frame, the adaptive spatial pixel interaction function represents the interaction strength by analyzing the color difference and spatial distance between the current pixel and its neighboring pixels. Two adjustment parameters are introduced to control the sensitivity of the interaction strength to the color difference and spatial distance respectively.

[0080] Adaptive spatial difference analysis weights the interaction strength between the current pixel and all neighboring pixels to calculate the adaptive spatial difference value;

[0081] The specific calculation formula of the adaptive spatial difference value is:

[0082]

[0083] Among them, ΔB' i (x, y) represents the adaptive spatial difference value of the pixel at position (x, y) in the i-th denoised video frame; ∑ (u,v)∈N(x,y) Indicates the accumulation of the interaction results calculated for all neighboring pixels; (u, v) represents the coordinates of the neighboring pixels; N (x, y) represents the neighborhood set of the current pixel. The size of the neighborhood set can be set according to the specific implementation scenario and is not limited here; C' i (x,y), C' i (u,v) respectively represent the color values ​​of the pixels at the current (x,y) position and the neighboring (u,v) position in the i-th denoised video frame; |C' i (x,y)-C' i (u,v)| represents the absolute difference in color value between the current pixel and the neighboring pixels in the denoised video frame, which is used to directly measure the color change between pixels; F' i (x, y, u, v) represents the adaptive spatial pixel interaction function, which is used to represent the interaction strength between the current pixel and the neighboring pixels in the denoised video frame. The specific calculation formula is:

[0084]

[0085] Wherein, exp represents an exponential function, which is used to define the nonlinear relationship of the interaction strength between pixels; γ represents an adjustment parameter for controlling the sensitivity of the interaction strength between pixels to color difference; λ represents an adjustment parameter for controlling the sensitivity of the interaction strength between pixels to the pixel distance; d(x, y, u, v) represents the spatial distance between the current pixel and the neighboring pixel, which is calculated using the Euclidean distance. The specific method is a technical means well known to those skilled in the art and will not be described in detail here;

[0086] Set the edge detection threshold according to the calculated adaptive spatial difference value ΔB' i (x, y), edge detection is performed; if the adaptive spatial difference value is greater than or equal to the set edge detection threshold, the current pixel is determined to be an edge pixel, otherwise it is a non-edge pixel, and the video frame I after edge detection is obtained. i .

[0087] S3, based on the video frame after edge detection, the matching value is calculated by the template matching algorithm, and the motion trajectory score is calculated by the nonlinear spatiotemporal smoothing motion trajectory analysis algorithm;

[0088] Based on the video frames after edge detection, the template matching algorithm is used to calculate the matching value; the template matching algorithm adaptively adjusts the weights of the video frames after edge detection and aligns the local features of the template to cope with the offset, lighting changes and occlusion problems in the video frames after edge detection; the template is derived from the industry standard action library, which is a standard behavior template manually designed based on the operating specifications and standards of the catering industry.

[0089] Adaptive weight adjustment is to assign a weight value to each pixel in the edge-detected video frame according to the local features of the edge-detected video frame. The size of the weight value depends on the feature strength of the pixel location, such as edge strength. The weight value is used to increase the influence of the key area on the overall matching value.

[0090] The local feature alignment of the template is to translate each position of the template, calculate the difference between the adjusted template and the video frame area after edge detection, and find the optimal template position by comparing the differences, thereby improving the matching accuracy;

[0091] The matching value is calculated as:

[0092]

[0093] Among them, A(I" i ,M k ) represents the video frame I after edge detection i With template M k The matching value of (x,y)∈RIt means to sum the weighted differences of all pixel points at the pixel position (x, y) in the matching area R. The matching area can be set according to the specific implementation scenario and is not limited here. W(x, y) represents the weight value at the pixel position (x, y). I” i (x, y) represents the pixel value at the pixel position (x, y) of the video frame after edge detection of the i-th edge; M k (x+Δx, y+Δy) represents the pixel value of the k-th template at the pixel position (x+Δx, y+Δy) after local feature alignment. Δx and Δy represent the alignment adjustment in the x and y directions, respectively. |·| represents the absolute value sign, which is used to avoid the offset of positive and negative deviations. M k It represents the kth type of template, which comes from the industry standard action library and is a standard behavior template designed manually based on the operating specifications and standards of the catering industry;

[0094] Based on the edge-detected video frames, the motion trajectory score is calculated using a nonlinear spatiotemporal smoothing motion trajectory analysis algorithm;

[0095] The nonlinear spatiotemporal smoothing motion trajectory analysis algorithm integrates time and space information and combines nonlinear functions to smooth motion vectors. It performs weighted summation of the motion vectors of all pixel positions in the matching area of ​​the video frame after edge detection to obtain the motion trajectory score.

[0096] Specifically, it is done by comparing the position changes of the same pixel in the current frame and the next frame, using the grayscale changes between frames to estimate the movement direction and speed of the pixel, and at the same time using a nonlinear function for smoothing to effectively filter out abnormal movements caused by illumination changes or subtle noises, thereby ensuring the smoothness and continuity of the trajectory. The motion trajectory score is obtained by weighted summing the motion vectors of all pixel positions in the matching area of ​​the video frame after edge detection.

[0097] The specific calculation formula of the motion vector is:

[0098]

[0099] Wherein, V" (x, y) represents the motion vector of the pixel point at the pixel position (x, y) in the matching area of ​​the video frame after edge detection; Δt represents the time interval between adjacent video frames after edge detection; tanh represents the hyperbolic tangent function, which is used for nonlinear smoothing processing; I" (x, y, t) represents the pixel value after edge detection at the pixel position (x, y) at time t; I" (x, y, t+Δt) represents the pixel value after edge detection at the pixel position (x, y) at time t+Δt; Represents the gradient of the video frame in the x direction after edge detection, which is used to measure the degree of change of pixels in the horizontal direction; Represents the gradient of the video frame in the y direction after edge detection, which is used to measure the degree of change of the pixel in the vertical direction; ∈ represents a very small positive number, which is used to prevent the denominator from being zero; Represents the square root symbol, which is used to calculate the Euclidean norm of the gradient. It combines the horizontal and vertical gradients to calculate the overall gradient strength, reflecting the magnitude of pixel change. The motion trajectory score is obtained by weighted summing of the motion vectors of all pixel positions in the matching area of ​​the video frame after edge detection. The specific formula is:

[0100]

[0101] Where Q represents the motion trajectory score; ∑ (x,y)∈R It represents the weighted summation of the motion vectors of the pixel points at all pixel positions (x, y) in the matching area R. The matching area can be set according to the specific implementation scenario and is not limited here. W(x, y) represents the weight value at the pixel position (x, y). ∥V”(x, y)∥ represents the modulus of the motion vector of the pixel point at the pixel position (x, y) in the matching area R of the video frame after edge detection.

[0102] S4. Based on the matching value and the motion trajectory score, a comprehensive score is obtained through weighted calculation; a comprehensive score threshold is set, and the employee behavior is judged by comparing the comprehensive score with the comprehensive score threshold.

[0103] Based on the matching value of the video frame after edge detection and the template and the motion trajectory score, a comprehensive score S is obtained through weighted calculation. Among them, the matching value of the video frame after edge detection and the template reflects the similarity between the current video frame and the standard template, and the motion trajectory score evaluates the motion continuity of the object. The combination of the matching value of the video frame after edge detection and the template and the motion trajectory score effectively improves the accuracy of behavior recognition;

[0104] The specific calculation formula for the comprehensive score is:

[0105]

[0106] Among them, S represents the comprehensive score; α and β represent weight parameters, which respectively represent the contribution of the matching value between the video frame and the template after edge detection and the motion trajectory score to the comprehensive score; A(I” i ,M k ) represents the video frame I after edge detection i With template M k The matching value of; Q represents the motion trajectory score;

[0107] Set the comprehensive scoring threshold, which can be set according to the specific implementation scenario and is not limited here;

[0108] By comparing the comprehensive score with the comprehensive score threshold, the employee's behavior is judged. If the comprehensive score is less than or equal to the comprehensive score threshold, it can be considered that the employee's behavior does not meet the catering operation specifications; if the comprehensive score is greater than the comprehensive score threshold, it is considered that the employee's behavior meets the catering operation specifications;

[0109] By setting a comprehensive scoring threshold to judge employee behavior, it is possible to automatically detect whether employee behavior complies with catering operation specifications, reducing the need for manual intervention and improving detection efficiency.

[0110] In summary, an intelligent analysis method and system for employee behavior in catering scenarios have been completed.

[0111] The order of the embodiments of the invention is for description only and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0112] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0113] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. An intelligent analysis method for employee behavior in a catering scenario, characterized in that: The following steps are involved: S1, collecting video frames to generate a continuous video frame sequence; performing denoising on each video frame in the video frame sequence to obtain a denoised video frame; S2, based on the denoised video frame, edge detection is performed by using an adaptive pixel-to-pixel interactive edge detection algorithm to obtain a video frame after edge detection; S3, based on the video frame after edge detection, the matching value is calculated by the template matching algorithm, and the grayscale change between the current frame and the next frame is compared by the nonlinear spatiotemporal smoothing motion trajectory analysis algorithm, and the motion direction and speed of the pixels in the matching area of ​​the video frame after edge detection are estimated. At the same time, the nonlinear function is combined for smoothing to obtain the motion vector; the specific calculation formula of the motion vector is: Wherein, V" (x, y) represents the motion vector of the pixel point at the pixel position (x, y) in the matching area of ​​the video frame after edge detection; Δt represents the time interval between adjacent video frames after edge detection; tanh represents the hyperbolic tangent function; I" (x, y, t) represents the pixel value after edge detection at the pixel position (x, y) at time t; I" (x, y, t+Δt) represents the pixel value after edge detection at the pixel position (x, y) at time t+Δt; Represents the gradient of the video frame in the x direction after edge detection; Represents the gradient of the video frame in the y direction after edge detection; ∈ represents a very small positive number, which is used to prevent the denominator from being zero; Represents the square root symbol; The motion vectors of all pixel positions in the matching area of ​​the edge-detected video frame are weighted summed to calculate the motion trajectory score. The specific formula is: Where Q represents the motion trajectory score; Σ (x,y)∈R represents the weighted sum of the motion vectors of all pixel points at the pixel position (x, y) in the matching area R; W(x, y) represents the weight value at the pixel position (x, y); ∥V”(x, y)∥ represents the modulus of the motion vector of the pixel point at the pixel position (x, y) in the matching area R of the video frame after edge detection; S4. Based on the matching value and the motion trajectory score, a comprehensive score is obtained through weighted calculation; a comprehensive score threshold is set, and the employee behavior is judged by comparing the comprehensive score with the comprehensive score threshold.

2. The intelligent analysis method of employee behavior in a catering scenario according to claim 1 is characterized in that: The S1 specifically includes: The implementation process of the denoising process is as follows: In the first step, a perturbation value is generated for each pixel of each video frame in the video frame sequence; The second step is to calculate the disturbance response value under each disturbance condition based on the disturbance value; The third step is to calculate the stability index while generating the disturbance response value; The fourth step is to determine whether to perform denoising on the video frame pixels based on the stability index.

3. The intelligent analysis method of employee behavior in a catering scenario according to claim 1 is characterized in that: The S2 specifically includes: In the process of implementing the adaptive pixel interaction edge detection algorithm, adaptive spatial pixel interaction function and adaptive spatial difference analysis are introduced.

4. The intelligent analysis method of employee behavior in a catering scenario according to claim 3 is characterized in that: The S2 specifically includes: The adaptive spatial pixel interaction function obtains the interaction strength between the current pixel and its neighboring pixels by analyzing each pixel in the denoised video frame.

5. The intelligent analysis method of employee behavior in a catering scenario according to claim 3 is characterized in that: The S2 specifically includes: The adaptive spatial difference analysis performs weighted calculation on the interaction strength between the current pixel and all neighboring pixels to obtain an adaptive spatial difference value; based on the adaptive spatial difference value, an edge judgment is performed on each pixel in the denoised video frame.

6. The intelligent analysis method of employee behavior in a catering scenario according to claim 1 is characterized in that: The S3 specifically includes: The template matching algorithm performs adaptive weight adjustment on the video frame after edge detection, aligns local features of the template obtained from the industry standard action library, and calculates the matching value between the video frame after edge detection and the template.

7. An intelligent analysis system for employee behavior in a dining scene, applied to the intelligent analysis method for employee behavior in a dining scene as claimed in claim 1, characterized in that: Includes the following sections: Video frame acquisition module, video frame denoising processing module, edge detection module, template matching module, motion trajectory analysis module, comprehensive score calculation module, behavior determination module; Video frame acquisition module: performs video frame acquisition, generates a continuous video frame sequence, and outputs the video frame sequence to the video frame denoising processing module; Video frame denoising processing module: performs denoising processing on each video frame in the video frame sequence to obtain a denoised video frame, and outputs the denoised video frame to the edge detection module; Edge detection module: Based on the denoised video frame, edge detection is performed through an adaptive pixel-to-pixel interactive edge detection algorithm to obtain the video frame after edge detection, and the video frame after edge detection is output to the template matching module and the motion trajectory analysis module; Template matching module: Based on the video frame after edge detection, the template matching algorithm is used to calculate the matching value, and the matching value is output to the comprehensive score calculation module; Motion trajectory analysis module: Based on the video frames after edge detection, the motion trajectory score is calculated by nonlinear spatiotemporal smoothing motion trajectory analysis algorithm, and the motion trajectory score is output to the comprehensive score calculation module; Comprehensive score calculation module: calculates the comprehensive score based on the matching value and motion trajectory score, and outputs the comprehensive score to the behavior judgment module; Behavior judgment module: Set a comprehensive score threshold and judge employee behavior by comparing the comprehensive score with the comprehensive score threshold.

Images