A method and system for assessing work quality based on image recognition

Abstract

The application discloses a kind of based on image recognition's working quality evaluation method and system, its method receives to failure quality evaluation request, obtains the two images to be analyzed of adjacent frames of working progress monitoring video and reference quantity value, extracts the region of interest of each image to be analyzed, the image to be analyzed is divided into target image and background image using region of interest, the coincidence degree between two background images is calculated, according to the comparison result of coincidence degree and preset coincidence degree threshold, whether two background images are matched is judged, if two background images match, corresponding target analysis strategy is selected according to target image, target image and reference quantity value are analyzed according to target analysis strategy, and working quality evaluation result is generated, solve the technical problem that the way of existing through portable communication device or artificial naked eye inspection is judged to complete progress, work quality cannot be accurately evaluated.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a work quality assessment method and system based on image recognition. Background Technology

[0002] In heavy industrial factories, it is common for workers to engage in long hours of manual labor, with a high degree of repetitiveness in their actions. Current technologies typically rely on portable communication devices or manual visual inspection to assess progress and manage it. However, both methods suffer from drawbacks such as high manpower costs, operational inconvenience, inaccurate data collection, and increased labor costs, and fail to provide accurate assessments of work quality. Summary of the Invention

[0003] This invention provides a work quality assessment method and system based on image recognition, which solves the technical problem that existing methods of judging progress and managing progress by relying on portable communication devices or manual visual inspection are time-consuming, labor-intensive, inconvenient to operate, and have inaccurate data acquisition, and cannot accurately assess work quality.

[0004] The first aspect of this invention provides a work quality assessment method based on image recognition, comprising:

[0005] When a work quality assessment request is received, two images to be analyzed and a reference quantity value are obtained from adjacent frames of the work progress monitoring video.

[0006] Extract the region of interest (ROI) of each image to be analyzed, and use the ROI to divide the image to be analyzed into a target image and a background image;

[0007] Calculate the overlap between the two background images, and determine whether the two background images match based on the comparison result between the overlap and a preset overlap threshold.

[0008] If two background images match, the target image is matched with the corresponding target analysis strategy in the preset analysis strategy key-value pair table.

[0009] The two target images and the reference quantity value are analyzed according to the target analysis strategy to generate a work quality assessment result.

[0010] Optionally, the step of extracting the region of interest (ROI) of each image to be analyzed and using the ROI to divide the image to be analyzed into a target image and a background image includes:

[0011] Select the center point of each image to be analyzed, and construct a two-dimensional coordinate system on the center point;

[0012] Based on the center point, a region of interest is constructed on the two-dimensional coordinate system using four preset coordinate points;

[0013] The image to be analyzed is cropped within the region of interest as the target image;

[0014] The image outside the region of interest in the image to be analyzed is used as the background image.

[0015] Optionally, the step of calculating the overlap between two background images and determining whether the two background images match based on a comparison of the overlap with a preset overlap threshold includes:

[0016] Calculate the first overlap value between the two background images;

[0017] Determine whether the first overlap value is greater than or equal to a preset overlap threshold;

[0018] If the first overlap value is greater than or equal to the overlap threshold, then the two background images are determined to be a match;

[0019] If the first overlap value is less than the overlap threshold, then the two background images are determined to be mismatched.

[0020] Optionally, the step of matching the target image with the corresponding target analysis strategy in a preset analysis strategy key-value pair table if two background images match includes:

[0021] If the two background images match, then the feature identifier in the target image is obtained;

[0022] Using the aforementioned feature identifier, a corresponding composite key is generated;

[0023] The target analysis strategy corresponding to the target image is matched according to the composite key in the preset analysis strategy key-value pair table.

[0024] Optionally, the step of parsing the two target images and the reference quantity value according to the target analysis strategy to generate a work quality assessment result includes:

[0025] Calculate a second overlap value between the two target images;

[0026] The first difference value is generated by subtracting the preset standard value from the second overlap value.

[0027] The first difference is compared with the reference quantity value to generate a work quality assessment result.

[0028] Optionally, the step of parsing the two target images and the reference quantity value according to the target analysis strategy to generate a work quality assessment result further includes:

[0029] The two target images are respectively input into a pre-trained target image analysis model to generate a first target image quantity value and a second target image quantity value; wherein, the target image analysis model is used to analyze the input target images and output the corresponding target image quantity value;

[0030] The difference between the second target image quantity value and the first target image quantity value is calculated to generate a second difference value;

[0031] The second difference is compared with the reference quantity value to obtain the ratio result as the work quality assessment result.

[0032] A second aspect of the present invention provides a work quality assessment system based on image recognition, comprising:

[0033] The data acquisition module is used to acquire two images to be analyzed and a reference quantity value from adjacent frames of the work progress monitoring video when a work quality assessment request is received.

[0034] The image cropping module is used to extract the region of interest (ROI) of each image to be analyzed, and to divide the image to be analyzed into a target image and a background image using the ROI.

[0035] The judgment and analysis module is used to calculate the overlap between two background images and determine whether the two background images match based on the comparison result of the overlap with a preset overlap threshold.

[0036] The target analysis strategy selection module is used to match the target image with the corresponding target analysis strategy in a preset analysis strategy key-value pair table if two background images match.

[0037] The quality assessment module is used to analyze the two target images and the reference quantity value according to the target analysis strategy, and generate a work quality assessment result.

[0038] Optional, the judgment and analysis module includes:

[0039] The first overlap value calculation submodule is used to calculate the first overlap value between the two background images;

[0040] The first overlap value determination submodule is used to determine whether the first overlap value is greater than or equal to a preset overlap threshold.

[0041] If the first overlap value is greater than or equal to the overlap threshold, then the two background images are determined to be a match;

[0042] If the first overlap value is less than the overlap threshold, then the two background images are determined to be mismatched.

[0043] Optional, the target analysis strategy selection module includes:

[0044] The feature identifier acquisition submodule is used to acquire the feature identifier in the target image if two background images match.

[0045] The composite key generation submodule is used to generate the corresponding composite key using the feature identifier;

[0046] The target analysis strategy acquisition submodule is used to match the target analysis strategy corresponding to the target image in a preset analysis strategy key-value pair table according to the composite key.

[0047] Optionally, the target analysis strategy includes a first target analysis strategy and a second target analysis strategy, and the quality assessment module includes:

[0048] The second overlap calculation submodule is used to calculate the second overlap value between the two target images;

[0049] The first work quality assessment result acquisition submodule is used to calculate the difference between the preset standard value and the second overlap value to generate a first difference value.

[0050] The first difference is compared with the reference quantity value to generate a work quality assessment result.

[0051] As can be seen from the above technical solutions, the present invention has the following advantages:

[0052] By acquiring images of the work area before and after the work is completed, a region of interest (ROI) is extracted from each image. Based on the ROI, the image before the work is completed is divided into a first target image and a first background image. Then, using the ROI, the image after the work is completed is divided into a second target image and a second background image. The first and second background images are then compared to determine if they are in the same location. If they are, a corresponding target analysis strategy is selected based on the target image. The two target images and a reference value are then analyzed according to the target analysis strategy to calculate the work quality assessment result. This method solves the technical problem of existing technologies that rely on portable communication devices or manual visual inspection to judge progress, making it impossible to accurately assess work quality. Attached Figure Description

[0053] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0054] Figure 1 This is a flowchart illustrating the steps of an image recognition-based work quality assessment method provided in Embodiment 1 of the present invention.

[0055] Figure 2 A flowchart illustrating the steps of an image recognition-based work quality assessment method provided in Embodiment 2 of the present invention.

[0056] Figure 3 This is a structural block diagram of a work quality assessment system based on image recognition provided in Embodiment 3 of the present invention. Detailed Implementation

[0057] This invention provides a work quality assessment method and system based on image recognition, which solves the technical problem that existing methods of judging progress and managing progress by relying on portable communication devices or manual visual inspection are time-consuming, labor-intensive, inconvenient to operate, and have inaccurate data acquisition, and thus cannot accurately assess work quality.

[0058] To make the objectives, features, and advantages of this invention more apparent and understandable, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described below are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0059] Please see Figure 1 , Figure 1 The flowchart illustrates the steps of a work quality assessment method based on image recognition, as provided in Embodiment 1 of the present invention.

[0060] This invention provides a work quality assessment method based on image recognition, comprising the following steps:

[0061] Step 101: When a work quality assessment request is received, obtain two images to be analyzed and a reference quantity value from adjacent frames of the work progress monitoring video.

[0062] A work quality assessment request refers to a request sent by management to generate work quality assessment results.

[0063] The images to be analyzed refer to images of the same area before and after the work was done, and the images contain information about the specific type of work.

[0064] The reference quantity value refers to the total amount of work that needs to be completed in the current stage.

[0065] In this embodiment of the invention, when a work quality assessment request is received from a manager via a terminal, two images to be analyzed from adjacent frames of the work progress monitoring video and the total amount of work to be completed in the current stage are obtained.

[0066] It should be noted that the image size and format are the same before and after working within the same area.

[0067] Step 102: Extract the region of interest (ROI) of each image to be analyzed, and use the ROI to divide the image to be analyzed into the target image and the background image.

[0068] In this embodiment of the invention, after the image to be analyzed is acquired, a two-dimensional coordinate system is constructed with the center point of the first image to be analyzed as the origin, and the first image to be analyzed is divided into a first target image and a first background image by using four preset coordinate points as the enclosed area. Similarly, a two-dimensional coordinate system is constructed with the center point of the second image to be analyzed as the origin, and the second image to be analyzed is divided into a second target image and a second background image by using four preset coordinate points as the enclosed area.

[0069] It should be noted that the first image to be analyzed is the image of the same area before processing, and the second image to be analyzed is the image of the same area after processing.

[0070] Step 103: Calculate the overlap between the two background images. Based on the comparison result of the overlap with the preset overlap threshold, determine whether the two background images match.

[0071] In this embodiment of the invention, the overlap value between the first background image and the second background image is calculated, and the first background image and the second background image are determined to match based on the comparison result of the overlap value and the preset first overlap threshold.

[0072] It should be noted that an information digest algorithm can also be used to determine whether the first background image and the second background image match. An information digest hash value is generated between the first background image and the second background image. It is then determined whether the information digest hash value of the first background image and the information digest hash value of the second background image are consistent. If the information digest hash values ​​are consistent, the first background image and the second background image are determined to match. If the information digest hash values ​​are inconsistent, the overlap value between the first background image and the second background image is calculated. Based on the comparison result of the overlap value and the preset first overlap threshold, it is determined whether the first background image and the second background image match.

[0073] Step 104: If the two background images match, then match the corresponding target analysis strategy in the preset analysis strategy key-value pair table according to the target image.

[0074] In this embodiment of the invention, when the first background image matches the second background image, it is determined that the first image to be analyzed and the second image to be analyzed are images at the same location, and the corresponding target analysis strategy is matched from the preset analysis strategy key-value pair table according to the feature identifier in the target image.

[0075] It should be noted that the target image contains labeled feature identifiers, which can be obtained through a pre-trained neural network model.

[0076] Step 105: Analyze the two target images and reference quantity values ​​according to the target analysis strategy to generate work quality assessment results.

[0077] In this embodiment of the invention, a corresponding calculation method is selected according to the target analysis strategy. If the target analysis strategy is the first target analysis strategy, the second overlap value between the first target image and the second target image is calculated, the first difference between the preset standard value and the second overlap value is calculated, and the first ratio between the first difference and the reference quantity value is calculated. The first ratio is determined as the work quality evaluation result. If the target analysis strategy is the second target analysis strategy, the first target image and the second target image are respectively input into the pre-trained target image analysis model to generate the first target image quantity value and the second target image quantity value. The second difference between the second target image quantity value and the first target image quantity value is calculated, and the second ratio between the second difference and the reference quantity value is used as the work quality evaluation result.

[0078] In this embodiment of the invention, when a request is received from a manager via a terminal, two images to be analyzed from adjacent frames of a work progress monitoring video are acquired. Based on four preset coordinate points, the image to be analyzed before work commences is divided into a first target image and a first background image. The image to be analyzed after work commences is divided into a second target image and a second background image. By calculating the first overlap between the first and second background images, it is determined whether the images to be analyzed before and after work are in the same position. If they are, a corresponding target analysis strategy is selected based on the target image. The two target images and a reference quantity value are analyzed according to the target analysis strategy to calculate the work completion degree corresponding to the image to be analyzed, and this work completion degree is used as the evaluation result of work quality. This solves the technical problem of existing technologies that rely on portable communication devices or manual visual inspection to determine progress, making it impossible to accurately evaluate work quality.

[0079] Please see Figure 2 , Figure 2 The flowchart illustrates the steps of a work quality assessment method based on image recognition, as provided in Embodiment 2 of the present invention.

[0080] Step 201: When a work quality assessment request is received, obtain two images to be analyzed and a reference quantity value from adjacent frames of the work progress monitoring video.

[0081] In this embodiment of the invention, when a work quality assessment request is received from a terminal, two images to be analyzed in adjacent frames of the work progress monitoring video and the total amount of work to be completed are obtained, wherein the two images to be analyzed are the image to be analyzed before work and the image to be analyzed after work.

[0082] Step 202: Extract the region of interest (ROI) of each image to be analyzed, and use the ROI to divide the image to be analyzed into the target image and the background image.

[0083] Further, step 202 includes the following sub-steps:

[0084] S11. Select the center point of each image to be analyzed and construct a two-dimensional coordinate system on the center point.

[0085] S12. Based on the center point, construct the region of interest using four preset coordinate points in a two-dimensional coordinate system.

[0086] The coordinate points refer to the four preset coordinate points with values ​​of (0.5L1, 0.5L2), (-0.5L1, 0.5L2), (0.5L1, -0.5L2), and (-0.5L1, -0.5L2) based on the length L1 and width L2 of the image to be analyzed.

[0087] S13. Extract the image within the region of interest from the image to be analyzed as the target image.

[0088] S14. Extract the image outside the region of interest from the image to be analyzed as the background image.

[0089] In this embodiment of the invention, a first coordinate system is constructed with the center point of the first image to be analyzed as the origin, a region of interest is constructed based on four preset coordinate points, an image of the first image to be analyzed within the region of interest is selected as the first target image, and an image of the first image to be analyzed outside the region of interest is selected as the first background image. Then, a second coordinate system is constructed with the center point of the second image to be analyzed as the origin, a region of interest is constructed based on four preset coordinate points, an image of the second image to be analyzed within the region of interest is selected as the second target image, and an image of the second image to be analyzed outside the region of interest is selected as the second background image.

[0090] It should be noted that the first coordinate system and the second coordinate system are two-dimensional coordinate systems. In this invention, the first image to be analyzed and the second image to be analyzed are divided using the same four coordinate points.

[0091] Step 203: Calculate the overlap between the two background images. Based on the comparison result of the overlap with the preset overlap threshold, determine whether the two background images match.

[0092] Optionally, step 203 includes the following steps S21-S24:

[0093] S21. Calculate the first overlap value between the two background images.

[0094] In this embodiment of the invention, a first overlap value between the first background image and the second background image is calculated using a histogram algorithm.

[0095] S22. Determine whether the first overlap value is greater than or equal to the preset overlap threshold.

[0096] The overlap threshold refers to the fact that the overall difference between the first background image and the second background image must be less than or equal to 10%.

[0097] In this embodiment of the invention, it is determined whether the first overlap value is greater than or equal to 90%.

[0098] S23. If the first overlap value is greater than or equal to the overlap threshold, then the two background images are determined to be matched.

[0099] In this embodiment of the invention, if the first overlap value is greater than or equal to 90%, the first background image is determined to match the second background image.

[0100] S24. If the first overlap value is less than the overlap threshold, then the two background images are determined to be mismatched.

[0101] In this embodiment of the invention, if the first overlap value is less than 90%, it is determined that the first background image and the second background image do not match.

[0102] Step 204: If the two background images match, obtain the feature identifiers in the target image.

[0103] Feature identification refers to the type of work that managers add to the image to be analyzed through annotation operations.

[0104] In this embodiment of the invention, when it is determined that the first background image and the second background image are at the same position, the feature identifier in the target image is obtained.

[0105] Step 205: Use feature identifiers to generate corresponding composite bonds.

[0106] In this embodiment of the invention, feature identifiers are used to generate corresponding composite bonds.

[0107] Step 206: Match the target image with the target analysis strategy according to the preset analysis strategy key-value pair table based on the composite key.

[0108] The target analysis strategy refers to the computational analysis method selected based on the job type corresponding to the image to be analyzed.

[0109] In this embodiment of the invention, a calculation and analysis method is used to match the target image with the composite key in a preset analysis strategy key-value pair table.

[0110] It should be noted that the composite key contains the work type of the target image. For example, if the work type is tree felling, the corresponding work progress can be determined by analyzing the proportion of each color in the target image. If the work type is material transportation, the degree of work completion can be determined by calculating the change in the quantity of materials before and after the target image.

[0111] Step 207: Analyze the two target images and reference quantity values ​​according to the target analysis strategy to generate work quality assessment results.

[0112] Optionally, step 207 includes the following steps S31-S36.

[0113] S31. Calculate the second overlap value between the two target images.

[0114] In this embodiment of the invention, when the target analysis strategy is the first target analysis strategy, a second overlap value between the first target image and the second target image is calculated.

[0115] It should be noted that if the target analysis strategy is the first target analysis strategy, the progress of the work on the image to be analyzed is determined by calculating the percentage of color change between the target image before and after the work.

[0116] S32. Subtract the preset standard value from the second overlap value to generate the first difference value.

[0117] The first difference refers to the difference between the first target image and the second target image.

[0118] In this embodiment of the invention, a first difference between 100% and the second degree of overlap is calculated.

[0119] S33. Ratio the first difference with the reference quantity value to generate the work quality assessment result.

[0120] In this embodiment of the invention, a first ratio between the first difference and the reference quantity value is calculated, and the first ratio is used as the evaluation result of the work quality.

[0121] Optionally, step 207 may also include the following steps S31-S33.

[0122] S34. Input the two target images into the pre-trained target image analysis model to generate the first target image quantity value and the second target image quantity value; wherein, the target image analysis model is used to analyze the input target images and output the corresponding target image quantity value.

[0123] In this embodiment of the invention, the first target image and the second target image are respectively output to a pre-trained target image analysis model to generate the quantity values ​​of the first target image and the second target image.

[0124] The specific training process of the target image analysis model is as follows:

[0125] S341. Obtain training sample data and corresponding standard quantity values, input the training sample data into the preset image analysis model for training, and generate the corresponding training quantity values.

[0126] The image analysis model refers to the YOLOv4 neural network model.

[0127] Training sample data refers to multiple training target images, such as tree images, box images, circuit breaker switch images, etc.

[0128] In this embodiment of the invention, when training sample data and corresponding standard quantity values ​​are obtained, the training sample data is input into a preset YOLOv4 neural network model for training to generate corresponding training quantity values.

[0129] It should be noted that when the training target images in the training sample data are input into the YOLOv4 neural network model, the training target images are first scaled to 416*416 before being input into the neural network model. The features of the training target images are initially extracted through the backbone feature extraction network based on MobileNetv3, resulting in three effective feature layers. Then, the feature extraction is enhanced through the feature pyramid based on SPP and PANet, achieving the fusion of the three effective feature layers and extracting three more effective feature layers. Finally, the YOLOv3 Head structure is used to detect the objects and their positions in the training target images, thus achieving the counting of training target images.

[0130] S342. Generate the corresponding accuracy value based on the comparison results between the training quantity value and the standard quantity value.

[0131] S343. Determine whether the precision value is greater than or equal to the preset precision threshold.

[0132] S344. If the accuracy value is greater than or equal to the accuracy threshold, stop training and generate the target image analysis model.

[0133] S345. If the accuracy value is less than the accuracy threshold, adjust the network parameters of the image analysis model according to the preset gradient, jump to the step of inputting the training sample data into the preset image analysis model for training and generating the corresponding training quantity value, until the accuracy value is greater than or equal to the accuracy threshold, and output its corresponding target image analysis model.

[0134] In this embodiment of the invention, a corresponding accuracy value is calculated based on the comparison result between the training quantity value and the standard quantity value. It is then determined whether the accuracy value is greater than or equal to 98%. If the accuracy value is greater than or equal to 98%, training is stopped. A target image analysis model is obtained. If the accuracy value is less than 98%, the model parameters of the image analysis model are adjusted using gradient descent, and the process jumps to the step of inputting training sample data into a preset YOLOv4 neural network model for training to generate the corresponding training quantity value, until the accuracy value is greater than or equal to 98%.

[0135] It should be noted that multiple training target images from the training sample data are input into the YOLOv4 neural network model to generate multiple training values. The number of training values ​​that are equal to the standard values ​​is counted, and then the ratio of the number of training values ​​to the number of training values ​​is calculated. This ratio is then used as the corresponding accuracy value.

[0136] S35. Subtract the quantity value of the second target image from the quantity value of the first target image to generate a second difference value.

[0137] In this embodiment of the invention, a second difference between the second target image quantity value and the first target image quantity value is calculated.

[0138] S36. Ratio the second difference with the reference quantity value to generate the work quality assessment result.

[0139] In this embodiment of the invention, a second ratio between the second difference and the reference quantity value is calculated, and the second ratio is used as the evaluation result of the work quality.

[0140] In this embodiment of the invention, when a work quality assessment request is received from a manager via a terminal, two images to be analyzed from adjacent frames of a work progress monitoring video are acquired. Based on four preset coordinate points, the image before work is completed is divided into a first target image and a first background image. The image after work is completed is divided into a second target image and a second background image. By calculating the first overlap between the first and second background images, it is determined whether the images before and after work are in the same position. If they are, a corresponding target analysis strategy is selected based on any target image. The two target images and a reference quantity value are analyzed according to the target analysis strategy to calculate the work completion degree corresponding to the image and use this degree as the work quality assessment result. This solves the technical problem of existing technologies that rely on portable communication devices or manual visual inspection to determine progress, making it impossible to accurately assess work quality.

[0141] Please see Figure 3 , Figure 3 This invention provides a structural block diagram of a work quality assessment system based on image recognition, as shown in Embodiment 3 of the present invention.

[0142] The data acquisition module 301 is used to acquire two images to be analyzed and a reference quantity value of adjacent frames of the work progress monitoring video when a work quality assessment request is received.

[0143] The image cropping module 302 is used to extract the region of interest (ROI) of each image to be analyzed, and to divide the image to be analyzed into a target image and a background image using the ROI.

[0144] The judgment and analysis module 303 is used to calculate the overlap between two background images and determine whether the two background images match based on the comparison result of the overlap with the preset overlap threshold.

[0145] The target analysis strategy selection module 304 is used to match the target image with the corresponding target analysis strategy in the preset analysis strategy key-value pair table if two background images match.

[0146] The quality assessment module 305 is used to analyze two target images and reference quantity values ​​according to the target analysis strategy and generate work quality assessment results.

[0147] Furthermore, the image cropping module 302 includes:

[0148] The two-dimensional coordinate system construction submodule is used to select the center point of each image to be analyzed and construct a two-dimensional coordinate system at the center point;

[0149] The Region of Interest (ROI) construction submodule is used to construct a region of interest in a two-dimensional coordinate system based on a center point and using four preset coordinate points.

[0150] The target image cropping submodule is used to crop the image within the region of interest of the image to be analyzed as the target image.

[0151] The background image cropping submodule is used to crop the image outside the region of interest of the image to be analyzed as the background image;

[0152] Furthermore, the judgment and analysis module 303 includes:

[0153] The first overlap value calculation submodule is used to calculate the first overlap value between two background images;

[0154] The first overlap value judgment submodule is used to determine whether the first overlap value is greater than or equal to the preset overlap threshold.

[0155] If the first overlap value is greater than or equal to the overlap threshold, then the two background images are determined to be a match;

[0156] If the first overlap value is less than the overlap threshold, the two background images are determined to be mismatched.

[0157] Furthermore, the target analysis strategy selection module 304 includes:

[0158] The feature identification acquisition submodule is used to acquire the feature identification in the target image if two background images match.

[0159] The composite key generation submodule is used to generate corresponding composite keys using feature identifiers;

[0160] The target analysis strategy acquisition submodule is used to match the target analysis strategy corresponding to the target image in a preset analysis strategy key-value pair table according to the composite key.

[0161] Furthermore, the target analysis strategy includes a first target analysis strategy and a second target analysis strategy. The quality assessment module 305 includes:

[0162] The second overlap calculation submodule is used to calculate the second overlap value between two target images;

[0163] The first work quality assessment result acquisition submodule is used to calculate the difference between the preset standard value and the second overlap value to generate the first difference value;

[0164] The first difference is compared with the reference quantity value to generate the work quality assessment result.

[0165] The target image quantity value acquisition submodule is used to input two target images into a pre-trained target image analysis model to generate a first target image quantity value and a second target image quantity value; wherein, the target image analysis model is used to analyze the input target images and output the corresponding target image quantity value;

[0166] The second work quality assessment result acquisition submodule is used to subtract the second target image quantity value from the first target image quantity value to generate a second difference value.

[0167] The second difference is compared with the reference quantity value, and the ratio result is used as the work quality assessment result.

[0168] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the system and unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0169] In the embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between systems or units may be electrical, mechanical, or other forms.

[0170] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0171] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A work quality assessment method based on image recognition, characterized in that, include: When a work quality assessment request is received, two images to be analyzed and a reference quantity value are obtained from adjacent frames of the work progress monitoring video. Extract the region of interest (ROI) of each image to be analyzed, and use the ROI to divide the image to be analyzed into a target image and a background image; Calculate the overlap between the two background images, and determine whether the two background images match based on the comparison result between the overlap and a preset overlap threshold. If two background images match, the target image is matched with the corresponding target analysis strategy in the preset analysis strategy key-value pair table. The two target images and the reference quantity value are analyzed according to the target analysis strategy to generate a work quality assessment result; The step of matching the target image with the corresponding target analysis strategy in a preset analysis strategy key-value pair table if the two background images match includes: If the two background images match, then the feature identifier in the target image is obtained; Using the aforementioned feature identifier, a corresponding composite key is generated; The target analysis strategy corresponding to the target image is matched according to the composite key in the preset analysis strategy key-value pair table.

2. The work quality assessment method based on image recognition according to claim 1, characterized in that, The step of extracting the region of interest (ROI) of each image to be analyzed and using the ROI to divide the image to be analyzed into a target image and a background image includes: Select the center point of each image to be analyzed, and construct a two-dimensional coordinate system on the center point; Based on the center point, a region of interest is constructed on the two-dimensional coordinate system using four preset coordinate points; The image to be analyzed is cropped within the region of interest as the target image; The image outside the region of interest in the image to be analyzed is used as the background image.

3. The work quality assessment method based on image recognition according to claim 1, characterized in that, The step of calculating the overlap between two background images and determining whether the two background images match based on a comparison of the overlap with a preset overlap threshold includes: Calculate the first overlap value between the two background images; Determine whether the first overlap value is greater than or equal to a preset overlap threshold; If the first overlap value is greater than or equal to the overlap threshold, then the two background images are determined to be a match; If the first overlap value is less than the overlap threshold, then the two background images are determined to be mismatched.

4. The work quality assessment method based on image recognition according to claim 2, characterized in that, The step of parsing the two target images and the reference quantity value according to the target analysis strategy to generate a work quality assessment result includes: Calculate a second overlap value between the two target images; The first difference value is generated by subtracting the preset standard value from the second overlap value. The first difference is compared with the reference quantity value to generate a work quality assessment result.

5. The work quality assessment method based on image recognition according to claim 2, characterized in that, The step of parsing the two target images and the reference quantity value according to the target analysis strategy to generate a work quality assessment result further includes: The two target images are respectively input into a pre-trained target image analysis model to generate a first target image quantity value and a second target image quantity value; wherein, the target image analysis model is used to analyze the input target images and output the corresponding target image quantity value; The difference between the second target image quantity value and the first target image quantity value is calculated to generate a second difference value; The second difference is compared with the reference quantity value to obtain the ratio result as the work quality assessment result.

6. A work quality assessment system based on image recognition, characterized in that, include: The data acquisition module is used to acquire two images to be analyzed and a reference quantity value from adjacent frames of the work progress monitoring video when a work quality assessment request is received. The image cropping module is used to extract the region of interest (ROI) of each image to be analyzed, and to divide the image to be analyzed into a target image and a background image using the ROI. The judgment and analysis module is used to calculate the overlap between two background images and determine whether the two background images match based on the comparison result of the overlap with a preset overlap threshold. The target analysis strategy selection module is used to match the target image with the corresponding target analysis strategy in a preset analysis strategy key-value pair table if two background images match. The quality assessment module is used to analyze the two target images and the reference quantity value according to the target analysis strategy, and generate a work quality assessment result; The target analysis strategy selection module includes: The feature identifier acquisition submodule is used to acquire the feature identifier in the target image if two background images match. The composite key generation submodule is used to generate the corresponding composite key using the feature identifier; The target analysis strategy acquisition submodule is used to match the target analysis strategy corresponding to the target image in a preset analysis strategy key-value pair table according to the composite key.

7. The work quality assessment system based on image recognition according to claim 6, characterized in that, The judgment and analysis module includes: The first overlap value calculation submodule is used to calculate the first overlap value between the two background images; The first overlap value determination submodule is used to determine whether the first overlap value is greater than or equal to a preset overlap threshold. If the first overlap value is greater than or equal to the overlap threshold, then the two background images are determined to be a match; If the first overlap value is less than the overlap threshold, then the two background images are determined to be mismatched.

8. The work quality assessment system based on image recognition according to claim 6, characterized in that, The target analysis strategy includes a first target analysis strategy and a second target analysis strategy. The quality assessment module includes: The second overlap calculation submodule is used to calculate the second overlap value between the two target images; The first work quality assessment result acquisition submodule is used to calculate the difference between the preset standard value and the second overlap value to generate a first difference value. The first difference is compared with the reference quantity value to generate a work quality assessment result.

Images