A jewelry quality identification method and device based on image recognition

Abstract

The present invention provides a jewelry quality identification method and device based on image recognition, which includes: S1: based on the omnidirectional image of the jewelry to be identified and the omnidirectional video when it is illuminated by a light, respectively building a three-dimensional static model and a dynamic model under illumination; S2: obtaining a surface defect identification result based on the dimensional data in the three-dimensional static model; S3: obtaining an internal defect identification result of the jewelry based on the dynamic model under illumination; S4: obtaining a jewelry quality identification result based on the surface defect identification result and the internal defect identification result; the method is used to combine model building and image recognition with existing jewelry identification standards, thereby completing the surface and internal defect identification and quality identification of the jewelry, improving the accuracy and efficiency of jewelry identification, and making the jewelry identification process streamlined and standardized.

Description

Technical Field

[0001] The present invention relates to the technical field of jewelry identification, and in particular to a jewelry quality identification method and device based on image recognition. Background Art

[0002] At present, the quality of jewelry in the jewelry market is uneven, and identification is mostly done manually.

[0003] However, this method requires a lot of manual experience and labor costs, and the accuracy of the identification results is closely related to the technical experience of the appraiser, and the identification efficiency of the manual identification method needs to be improved. Therefore, the current jewelry identification field needs an emerging industrial identification method that can be standardized and streamlined to improve the accuracy and efficiency of jewelry identification.

[0004] Therefore, the present invention proposes a jewelry quality identification method and device based on image recognition. Summary of the Invention

[0005] The present invention provides a jewelry quality appraisal method and device based on image recognition, which are used to combine model building and image recognition with existing jewelry appraisal standards, thereby completing the identification of surface and internal defects and quality appraisal of jewelry, improving the accuracy and efficiency of jewelry appraisal, and making the jewelry appraisal process streamlined and standardized.

[0006] The present invention provides a jewelry quality identification method based on image recognition, comprising:

[0007] S1: Based on the omnidirectional image of the jewelry to be identified and the omnidirectional video of the jewelry under illumination, a 3D static model and a dynamic model of the jewelry under illumination are constructed respectively;

[0008] S2: Obtain surface defect recognition results based on the dimensional data in the 3D static model;

[0009] S3: Obtaining the intrinsic flaw recognition results of jewelry based on the lighting dynamic model;

[0010] S4: Obtain jewelry quality appraisal results based on the surface defect identification results and the internal defect identification results.

[0011] Preferably, the jewelry quality identification method based on image recognition, S1: based on the omnidirectional image of the jewelry to be identified and the omnidirectional video when illuminated by light, respectively building a three-dimensional static model and a dynamic model of the jewelry under illumination, including:

[0012] Obtain a full range of images of the jewelry to be identified under preset shooting environment conditions;

[0013] Obtain a 360-degree video of the jewelry being authenticated when it is illuminated by a preset light source;

[0014] Build a three-dimensional static model based on the omnidirectional image;

[0015] Build a dynamic lighting model based on omnidirectional video.

[0016] Preferably, the jewelry quality identification method based on image recognition obtains a full-range image of the jewelry to be identified under preset shooting environment conditions, including:

[0017] Obtaining images of the jewelry to be identified from multiple angles under preset shooting environment conditions;

[0018] Based on the display parameters of each pixel in the captured image, a plurality of divided regions corresponding to the display parameters are obtained, and edge lines of the corresponding display parameters of the corresponding transformed regions are determined;

[0019] Based on the edge lines corresponding to each display parameter of each transformation area, adjacent captured images are connected and fused to obtain a full-dimensional image of the jewelry to be identified under the preset shooting environment conditions.

[0020] Preferably, the jewelry quality identification method based on image recognition connects and fuses adjacent captured images based on edge lines corresponding to each display parameter of each transformed area to obtain a full-dimensional image of the jewelry to be identified under preset shooting environment conditions, including:

[0021] Adjacent captured images are determined based on the primary and secondary edge lines in all transformed areas, and the adjacent captured images are connected and fused to obtain a full-scale image of the jewelry to be identified under preset shooting environment conditions.

[0022] Preferably, the jewelry quality identification method based on image recognition, S2: obtaining a surface defect identification result based on the dimensional data in the three-dimensional static model, comprises:

[0023] S201: Marking the edges of the 3D static model to obtain the model edges;

[0024] S202: Match and compare the size data of the model edge with the standard size data of the jewelry to be identified to obtain the surface defect recognition result of the jewelry to be identified.

[0025] Preferably, the jewelry quality identification method based on image recognition, S202: matching and comparing the size data of the model edge with the standard size data of the jewelry to be identified to obtain the surface defect identification result of the jewelry to be identified, includes:

[0026] Measure the size data of all model edges;

[0027] Matching the size data of the model edge with the standard size data of the jewelry to be identified to obtain an edge matching result;

[0028] Based on the matching results, edge size comparison is performed to obtain the surface defect identification results of the jewelry to be identified.

[0029] Preferably, the jewelry quality identification method based on image recognition, S3: obtaining the intrinsic defect identification result of the jewelry based on the lighting dynamic model, includes:

[0030] Extracting the lighting image of the jewelry to be identified from the lighting dynamic model;

[0031] Perform image analysis on the jewelry area in the light image to obtain the inherent defect identification results of the jewelry to be identified.

[0032] Preferably, the jewelry quality identification method based on image recognition, S4: obtaining the jewelry quality identification result based on the surface defect identification result and the internal defect identification result, comprises:

[0033] Determine the surface identification result based on the preset surface identification standard and the surface defect identification result;

[0034] Determine the internal identification result based on the preset internal identification standard and the internal defect identification result;

[0035] The jewelry quality appraisal result is obtained based on the surface appraisal results and the internal appraisal results.

[0036] Preferably, the jewelry quality identification method based on image recognition obtains the jewelry quality identification result based on the surface identification result and the internal identification result, including:

[0037] Determine whether there are any unqualified surface identification items in the surface identification results or unqualified internal identification items in the internal identification results. If so, summarize the unqualified identification items to obtain the jewelry quality identification results of the jewelry to be identified. Otherwise, the jewelry to be identified meets the identification standards and is used as the jewelry quality identification results of the corresponding jewelry to be identified.

[0038] Among them, unqualified identification items include unqualified surface identification items and unqualified internal identification items.

[0039] The present invention proposes a jewelry quality identification device based on image recognition, comprising:

[0040] A model building module is used to build a three-dimensional static model and a dynamic model based on the omnidirectional image of the jewelry to be identified and the omnidirectional video of the jewelry under light;

[0041] A first recognition module is used to obtain a surface defect recognition result based on the dimensional data in the three-dimensional static model;

[0042] The second recognition module is used to obtain the inherent defect recognition result of the jewelry based on the lighting dynamic model;

[0043] The comprehensive identification module is used to obtain jewelry quality identification results based on surface defect identification results and internal defect identification results.

[0044] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present invention. The purposes and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings.

[0045] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0047] Figure 1 This is a flow chart of a jewelry quality identification method based on image recognition in an embodiment of the present invention;

[0048] Figure 2 This is a flow chart of another jewelry quality identification method based on image recognition in an embodiment of the present invention;

[0049] Figure 3 Schematic diagram of a jewelry quality identification device based on image recognition in an embodiment of the present invention. DETAILED DESCRIPTION

[0050] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0051] Example 1:

[0052] The present invention provides a jewelry quality identification method based on image recognition, referring to Figure 1 ,include:

[0053] S1: Based on the omnidirectional image of the jewelry to be identified and the omnidirectional video of the jewelry under illumination, a 3D static model and a dynamic model of the jewelry under illumination are constructed respectively;

[0054] S2: Obtain surface defect recognition results based on the dimensional data in the 3D static model;

[0055] S3: Obtaining the intrinsic flaw recognition results of jewelry based on the lighting dynamic model;

[0056] S4: Obtain jewelry quality appraisal results based on the surface defect identification results and the internal defect identification results.

[0057] In this embodiment, the jewelry to be identified is the jewelry that needs to be identified in the present invention.

[0058] In this embodiment, the omnidirectional image is an image that includes all appearance details of the jewelry to be authenticated.

[0059] In this embodiment, the omnidirectional video is a video of the jewelry to be identified under preset lighting conditions, including all appearance details of the jewelry to be identified.

[0060] In this embodiment, the three-dimensional static model is a model constructed based on the omnidirectional image to represent the appearance characteristics of the jewelry to be identified under preset environmental conditions.

[0061] In this embodiment, the lighting dynamic model is a dynamic model that represents all appearance details of the jewelry to be identified under preset lighting conditions.

[0062] In this embodiment, the dimensional data is the dimensional data of the three-dimensional static model.

[0063] In this embodiment, the surface defect recognition result is the result obtained after the surface defects of the jewelry to be identified are recognized based on the size data in the three-dimensional static model.

[0064] In this embodiment, the intrinsic defect recognition result is the result obtained after the intrinsic defects of the jewelry to be identified are recognized based on the lighting dynamic model.

[0065] In this embodiment, the jewelry quality appraisal result is the appraisal result of the jewelry quality of the jewelry to be appraised obtained based on the surface defect recognition result and the internal defect recognition result.

[0066] The beneficial effects of the above technology are: combining model building and image recognition with existing jewelry identification standards, realizing the identification of surface and internal defects and quality appraisal of jewelry based on the three-dimensional static model and lighting dynamic model of jewelry, improving the accuracy and efficiency of jewelry identification, and making the jewelry identification process streamlined and standardized.

[0067] Example 2:

[0068] Based on Example 1, the jewelry quality identification method based on image recognition, S1: Based on the omnidirectional image of the jewelry to be identified and the omnidirectional video when illuminated by a light, respectively build a three-dimensional static model and a dynamic model of the light, including:

[0069] Obtain a full range of images of the jewelry to be identified under preset shooting environment conditions;

[0070] Obtain a 360-degree video of the jewelry being authenticated when it is illuminated by a preset light source;

[0071] Build a three-dimensional static model based on the omnidirectional image;

[0072] Build a dynamic lighting model based on omnidirectional video.

[0073] In this embodiment, the preset shooting environment conditions are preset environment conditions for obtaining omnidirectional images, for example, the brightness is greater than a preset brightness threshold and the clarity is greater than a preset clarity threshold.

[0074] In this embodiment, the preset lighting is a preset lighting condition for obtaining a full-range video of the jewelry to be identified when it is illuminated by light, for example: the light intensity is greater than the preset light intensity and the brightness is within a preset brightness range.

[0075] The beneficial effects of the above technology are: by constraining the shooting environment conditions and the illumination when obtaining omnidirectional images and omnidirectional videos of the jewelry to be identified, the quality of the obtained omnidirectional images and omnidirectional videos is guaranteed, which is conducive to the subsequent complete construction of three-dimensional static models and lighting dynamic models.

[0076] Example 3:

[0077] Based on Example 2, the jewelry quality identification method based on image recognition obtains a full-range image of the jewelry to be identified under preset shooting environment conditions, including:

[0078] Obtaining images of the jewelry to be identified from multiple angles under preset shooting environment conditions;

[0079] Obtaining display parameter distribution data of the captured image based on the display parameter of each pixel in the captured image, performing cluster analysis on sub-distribution data of corresponding types of display parameters in the display parameter distribution data, until the clustering result obtained by the current clustering process is consistent with the clustering result obtained by the previous clustering process, then obtaining multiple display parameter clusters of corresponding types of display parameters based on the latest clustering result;

[0080] Aggregating the pixel points corresponding to all display parameters included in the display parameter cluster to obtain a plurality of pixel point clusters corresponding to the types of display parameters, and dividing the captured image based on the pixel point clusters to obtain a plurality of divided areas corresponding to the types of display parameters;

[0081] Determining a list of enhancement coefficients for the corresponding type of display parameters based on the parameter ranges and limit values ​​of the corresponding type of display parameters in each divided area;

[0082] Performing enhancement processing on the display parameters in the corresponding divided areas based on the enhancement coefficients included in the enhancement coefficient list, to obtain a set of transformed areas of the corresponding types of display parameters;

[0083] Extracting an underlying feature distribution map of each transformation region in the transformation region set based on a feature extraction model, and marking common features of the underlying feature distribution maps of all transformation regions in the transformation region set to obtain underlying common feature distribution maps of corresponding types of display parameters of corresponding transformation regions;

[0084] Perform edge extraction on the underlying common feature distribution map to obtain edge lines of corresponding display parameters of corresponding transformation areas;

[0085] Based on the edge lines corresponding to each display parameter of each transformation area, adjacent captured images are connected and fused to obtain a full-dimensional image of the jewelry to be identified under the preset shooting environment conditions.

[0086] In this embodiment, the captured image is an image obtained by capturing the jewelry to be identified at a preset position under preset shooting environment conditions.

[0087] In this embodiment, the preset orientation is the orientation for photographing the jewelry to be identified and obtaining the photographed image.

[0088] In this embodiment, the display parameters are parameters such as the grayscale value, chromaticity value, brightness value, and contrast of the pixel points.

[0089] In this embodiment, the display parameter distribution data is data including the display parameters of each pixel in the captured image.

[0090] In this embodiment, the sub-distribution data is data including corresponding types of display parameters of all pixels in the captured image.

[0091] In this embodiment, the display parameter cluster is to perform cluster analysis on the sub-distribution data of the corresponding type of display parameters in the display parameter distribution data until the clustering result obtained by the current clustering process is consistent with the clustering result obtained by the previous clustering process, and then the clusters of multiple display parameters containing pixel points in the captured image contained in the latest clustering result are included.

[0092] In this embodiment, the pixel cluster is a cluster consisting of pixel points of corresponding display parameters obtained by aggregating pixel points corresponding to all display parameters included in the display parameter cluster.

[0093] In this embodiment, the divided area is an area obtained by dividing the captured image based on pixel clusters.

[0094] In this embodiment, the parameter range is the value range of the corresponding type display parameter in the corresponding divided area.

[0095] In this embodiment, the limit value is the limit value of the corresponding display parameter. For example, if the range of the chromaticity value is -100 to 100, the limit values ​​of the grayscale value are -100 and 100.

[0096] In this embodiment, based on the parameter range and the limit value of the corresponding type parameter in each divided area, the enhancement coefficient list of the corresponding type display parameter is determined, namely:

[0097] When the lower limit value of the parameter range of the corresponding type of display parameter is not less than 0, the ratio of the larger limit value to the upper limit value of the parameter range is used as the maximum enhancement coefficient;

[0098] When the lower limit value of the parameter range of the corresponding type of display parameter is less than 0, the minimum value of the ratio of the larger limit value and the upper limit value of the parameter range to the smaller range value and the lower limit value of the parameter range is used as the maximum enhancement coefficient;

[0099] The difference between the maximum enhancement coefficient and 1 is used as the enhancement coefficient coverage value, and the ratio of the enhancement coefficient coverage value to the preset number of enhancement coefficients is used as the enhancement coefficient interval value;

[0100] The enhancement coefficients included in the enhancement coefficient list include: the sum of 1 and the enhancement coefficient interval value, the sum of 1 and 2 times the enhancement coefficient interval value, the sum of 1 and 3 times the enhancement coefficient interval value, and up to the sum of 1 and n times the enhancement coefficient interval value, where n is the total number of enhancement coefficients.

[0101] In this embodiment, the enhancement coefficient is the multiple by which the corresponding display parameter is multiplied in each enhancement process included in the enhancement coefficient list.

[0102] In this embodiment, the enhancement process is to multiply the display parameters in the corresponding divided areas by the corresponding enhancement coefficients to obtain the image areas corresponding to the new display parameters.

[0103] In this embodiment, the transformation region set is a set consisting of multiple image regions obtained by sequentially enhancing the display parameters in the corresponding divided regions based on the enhancement coefficients included in the enhancement coefficient list.

[0104] In this embodiment, the feature extraction model is a model used to extract the underlying feature distribution map in the transformation area.

[0105] In this embodiment, the underlying feature distribution map is a distribution map representing the underlying features in the corresponding transformation area that is extracted from the transformation area based on the feature kicking model.

[0106] In this embodiment, the underlying common feature distribution map is the underlying feature distribution of the image areas corresponding to the multiple enhanced display parameters of the corresponding types of display parameters of the corresponding transformation areas obtained after commonality marking of the underlying feature distribution maps of all transformation areas in the transformation area set.

[0107] The beneficial effects of the above technology are: by performing cluster analysis on the corresponding type display parameters of the pixel points in the captured image, the area division results corresponding to each display parameter are obtained, and a personalized enhancement coefficient list is determined based on the parameter range in the divided area contained in the area division results of each display parameter, so as to realize personalized enhancement processing of the divided area of ​​the corresponding type display parameter, and based on the commonality extraction of the underlying features of the multiple transformed areas obtained after the personalized enhancement processing, the common features are extracted after the captured image is personalized divided and personalized enhanced. On the one hand, the features in the captured image are made more prominent after the enhancement processing, and on the other hand, the display error of the captured image caused by the enhancement processing is avoided, so that the accurate edge lines corresponding to the corresponding type display parameters can be extracted subsequently, thereby realizing accurate connection and fusion between the captured images, and obtaining a full-dimensional image that can reflect all the features of the jewelry to be identified.

[0108] Example 4:

[0109] Based on Example 3, the jewelry quality identification method based on image recognition connects and fuses adjacent captured images based on the edge lines corresponding to each display parameter of each transformed area to obtain a full-dimensional image of the jewelry to be identified under preset shooting environment conditions, including:

[0110] The edge line portion included in the edge line corresponding to each display parameter in the corresponding transformation area is regarded as the main edge line of the corresponding transformation area;

[0111] and determining sub-edge lines that exist in all edge lines corresponding to multiple display parameters except the main edge line, and determining the total number of display parameter types corresponding to the edge lines including the sub-edge lines;

[0112] Calculating the inclusion ratio based on the total number of display parameter types corresponding to the edge line containing the sub-edge line and the total number of display parameter types;

[0113] When the inclusion ratio is not less than the inclusion ratio threshold, the corresponding sub-edge line is used as the secondary edge line;

[0114] Marking the primary and secondary edge lines in all transformed regions at corresponding positions in the corresponding captured image to obtain the initial edge line of the corresponding captured image, performing shape matching based on the initial edge line and the standard edge line of the jewelry to be authenticated to obtain a shape matching result, restoring and supplementing the initial edge line with the shape matching result and the standard edge line to obtain the complete edge line of the corresponding captured image;

[0115] Based on the complete edge lines in the corresponding captured image and the preset feature shape list, the feature shapes in the captured image are identified, and based on the distribution position of each feature shape in the captured image, the distribution characteristics of each feature shape in the captured image are determined;

[0116] Based on the distribution characteristics of each shape in the captured images, pairwise matching is performed in all captured images to determine adjacent captured images;

[0117] Based on the complete outline of the captured image, adjacent captured images are locally matched to determine the overlapping areas in the adjacent captured images. Based on the overlapping areas, the adjacent captured images are connected and fused to obtain a full-dimensional image of the jewelry to be identified under the preset shooting environment conditions.

[0118] In this embodiment, the main edge line is the edge line portion included in the edge lines corresponding to each display parameter in the corresponding transformation area.

[0119] In this embodiment, the sub-edge lines are edge lines corresponding to all types of display parameters of the transformation region, except for the main edge lines, which are present in all edge lines corresponding to multiple types of display parameters.

[0120] In this embodiment, the inclusion ratio is calculated based on the total number of display parameter types corresponding to the edge lines including the sub-edge lines and the total number of display parameter types:

[0121]

[0122] Where α is the inclusion ratio, a is the total number of display parameter types corresponding to the edge line containing the sub-edge line, and b is the total number of display parameter types;

[0123] For example, if a is 5 and a is 10, then 0.5.

[0124] In this embodiment, the ratio threshold is the minimum inclusion ratio corresponding to when a sub-edge line is regarded as a secondary edge line.

[0125] In this embodiment, the initial edge line is an initially determined edge line in the corresponding captured image obtained by marking the primary edge lines and secondary edge lines in all transformed regions at corresponding positions in the corresponding captured image.

[0126] In this embodiment, the shape matching result is the result obtained by performing shape matching between the initial edge line and the standard edge line of the jewelry to be identified.

[0127] In this embodiment, the complete edge line is a complete edge line in the corresponding captured image obtained by supplementing the missing portion of the initial edge line based on the shape matching result and the standard edge line.

[0128] In this embodiment, the preset feature shape list is a list containing preset feature shapes, wherein the preset feature shapes include: angular shapes with preset angles (i.e., shapes composed of two edge lines forming an angle of 60 degrees to each other), squares, triangles, hexagons, etc.

[0129] In this embodiment, the characteristic shape is a characteristic shape identified in the captured image based on the complete edge line in the corresponding captured image and a preset characteristic shape list.

[0130] In this embodiment, the distribution feature is a feature based on characterizing the distribution position of the corresponding characteristic shape in the captured image.

[0131] In this embodiment, based on the distribution characteristics of each shape in the captured images, pairwise matching is performed in all captured images to determine adjacent captured images, namely:

[0132] Determine the center coordinate point of each characteristic shape in the distribution feature, determine the center coordinate point spacing between each pair of characteristic shapes of the corresponding type in the captured images, and determine the total number of groups of characteristic shapes of the corresponding types in the two captured images with the same center coordinate point spacing. Based on the total number of groups of each characteristic shape, calculate the distribution feature matching degree in the two captured images:

[0133]

[0134] Where β is the distribution feature matching degree in the two captured images, i is the i-th feature shape, n is the total number of feature shape types, q i is the total number of groups with consistent spacing between central coordinate points corresponding to the i-th characteristic shape, and Q is the total number of groups of pairwise characteristic shapes corresponding to the i-th characteristic shape;

[0135] For example, if n is 2, the total number of groups with consistent spacing between central coordinate points corresponding to the first characteristic shape is 6, the total number of groups of pairwise characteristic shapes corresponding to the first characteristic shape is 10, the total number of groups with consistent spacing between central coordinate points corresponding to the second characteristic shape is 8, and the total number of groups of pairwise characteristic shapes corresponding to the second characteristic shape is 10, then β is 0.7.

[0136] When the distribution feature matching degree is greater than the matching degree threshold, it is determined that the corresponding two captured images are adjacent captured images.

[0137] In this embodiment, adjacent captured images are captured images containing the same portion of the jewelry to be identified, which are determined by performing pairwise matching among all captured images based on the distribution characteristics of each shape in the captured images.

[0138] In this embodiment, local matching is to match local contours in adjacent captured images based on the complete contours of the captured images.

[0139] In this embodiment, the overlapping area is the overlapping area in adjacent captured images determined after local matching of adjacent captured images based on the complete outline of the captured images (that is, the image area showing the same area of ​​the jewelry to be identified).

[0140] In this embodiment, the omnidirectional image is an image containing all the appearance details of the jewelry to be identified, which is obtained by connecting and fusing adjacent captured images based on the overlapping area.

[0141] The beneficial effects of the above technology are: by overlapping the edge lines corresponding to all types of display parameters contained in the transformation area, the overlapping parts are determined and statistically compared, and then the initial edge line of the corresponding transformation area is determined, and the initial edge line is matched with the standard edge line movement of the jewelry to be identified to achieve complete restoration of the initial edge line, and then the characteristic shape in the completely restored edge line is marked and the corresponding distribution feature is obtained. Based on the matching of distribution features and local contours between the captured images, the overlapping areas between adjacent captured images and adjacent captured images are accurately determined, and then the accurate connection and fusion between the captured images are achieved, so that the obtained all-round image is sufficiently complete and clear.

[0142] Example 5:

[0143] On the basis of Example 1, the jewelry quality identification method based on image recognition, S2: obtaining surface defect recognition results based on the size data in the three-dimensional static model, referring to Figure 2 ,include:

[0144] S201: Marking the edges of the 3D static model to obtain the model edges;

[0145] S202: Match and compare the size data of the model edge with the standard size data of the jewelry to be identified to obtain the surface defect recognition result of the jewelry to be identified.

[0146] In this embodiment, the model edge is the edge of the three-dimensional static model obtained after edge recognition and marking are performed on the three-dimensional static model.

[0147] In this embodiment, the standard size data is the size data corresponding to the jewelry to be identified in a flawless state.

[0148] The beneficial effects of the above technology are: by identifying all edges in the three-dimensional static model and matching and comparing the size data of the model edges in the three-dimensional static model with the standard size data of the jewelry to be identified, the appearance size data of the jewelry to be identified can be inspected, and then the surface defects of the jewelry to be identified can be identified.

[0149] Example 6:

[0150] Based on Example 5, the jewelry quality identification method based on image recognition, S202: matching and comparing the size data of the model edge with the standard size data of the jewelry to be identified to obtain the surface defect identification result of the jewelry to be identified, includes:

[0151] Measure the size data of all model edges;

[0152] Matching the size data of the model edge with the standard size data of the jewelry to be identified to obtain an edge matching result;

[0153] Based on the matching results, edge size comparison is performed to obtain the surface defect identification results of the jewelry to be identified.

[0154] In this embodiment, the edge matching result is the result obtained by matching the size data of the model edge with the standard size data of the jewelry to be identified.

[0155] The beneficial effect of the above technology is: by matching the size data of the model edge with the standard size data of the jewelry to be identified and then comparing them accordingly, the appearance size data of the jewelry to be identified can be inspected, and then the surface defects of the jewelry to be identified can be identified.

[0156] Example 7:

[0157] Based on Example 1, the jewelry quality identification method based on image recognition, S3: obtaining the jewelry intrinsic defect identification result based on the lighting dynamic model, includes:

[0158] Extracting the lighting image of the jewelry to be identified from the lighting dynamic model;

[0159] Perform image analysis on the jewelry area in the light image to obtain the inherent defect identification results of the jewelry to be identified.

[0160] In this embodiment, the lighting image is an image of the jewelry to be identified captured in the lighting dynamic model when it is illuminated.

[0161] In this embodiment, image analysis is performed on the jewelry area in the light image to obtain the intrinsic defect identification result of the jewelry to be identified, which is:

[0162] By analyzing the image area of ​​the jewelry to be identified in the light image, the transmission characteristics of light in the material of the jewelry to be identified when the jewelry to be identified is analyzed, which can reflect the intrinsic defects of the material of the jewelry to be identified other than the surface when the jewelry to be identified is illuminated by preset light. Based on the transmission characteristics, the intrinsic defect identification result of the jewelry to be identified is obtained.

[0163] In this embodiment, the jewelry area is the image area where the jewelry is located in the light image.

[0164] The beneficial effect of the above technology is: by performing image analysis on the jewelry area in the lighting image of the jewelry to be identified that is intercepted in the lighting dynamic model, the transmission characteristics of the preset light in the jewelry to be identified are obtained, thereby realizing the identification of the inherent defects of the jewelry to be identified.

[0165] Example 8:

[0166] Based on Example 1, the jewelry quality identification method based on image recognition, S4: obtaining a jewelry quality identification result based on the surface defect identification result and the internal defect identification result, includes:

[0167] Determine the surface identification result based on the preset surface identification standard and the surface defect identification result;

[0168] Determine the internal identification result based on the preset internal identification standard and the internal defect identification result;

[0169] The jewelry quality appraisal result is obtained based on the surface appraisal results and the internal appraisal results.

[0170] In this embodiment, the preset surface identification standard is a pre-prepared surface identification standard for checking whether the jewelry to be identified has surface defects, for example: the deviation of the size data is less than the deviation threshold corresponding to the corresponding standard size.

[0171] In this embodiment, the surface identification result is the result of the surface identification items including whether the surface of the jewelry to be identified does not meet the corresponding identification standards, which is determined based on the preset surface identification standards and the surface defect identification results.

[0172] In this embodiment, the preset intrinsic identification standard is a pre-prepared identification standard for the intrinsic identification item used to check whether the jewelry to be identified has intrinsic defects, for example: the transmission angle deviation of the corresponding section is less than the transmission angle deviation threshold of the corresponding section, etc.

[0173] In this embodiment, the intrinsic identification result is the result of determining whether there are any intrinsic identification items in the interior of the jewelry to be identified (i.e., in the material other than the surface of the jewelry) that do not meet the corresponding identification standards based on the preset intrinsic identification standards and the intrinsic defect identification results.

[0174] The beneficial effects of the above technology are: based on the preset identification standards and the surface defect identification results and the internal defect identification results, the defect identification results are converted into the identification results corresponding to the preset identification standards, thereby standardizing the jewelry identification results.

[0175] Example 9:

[0176] Based on Example 8, the jewelry quality identification method based on image recognition obtains the jewelry quality identification result based on the surface identification result and the internal identification result, including:

[0177] Determine whether there are any unqualified surface identification items in the surface identification results or unqualified internal identification items in the internal identification results. If so, summarize the unqualified identification items to obtain the jewelry quality identification results of the jewelry to be identified. Otherwise, the jewelry to be identified meets the identification standards and is used as the jewelry quality identification results of the corresponding jewelry to be identified.

[0178] Among them, unqualified identification items include unqualified surface identification items and unqualified internal identification items.

[0179] In this embodiment, the surface identification item is an identification item for identifying whether the surface of the jewelry to be identified is qualified, such as an item for identifying the edge size deviation of the jewelry to be identified, or an item for identifying the flatness of the cut surface.

[0180] In this embodiment, the intrinsic identification item is an identification item for identifying whether the intrinsic quality of the jewelry to be identified is qualified, such as an item for identifying the purity of the material of the jewelry to be identified.

[0181] The beneficial effect of the above technology is that the final jewelry quality appraisal result of the jewelry to be appraised is obtained by classifying and summarizing the surface appraisal results and the internal appraisal results, so that the obtained jewelry quality appraisal results are more unified and standardized.

[0182] Example 10:

[0183] The present invention provides a jewelry quality identification device based on image recognition, referring to Figure 3 ,include:

[0184] A model building module is used to build a three-dimensional static model and a dynamic model based on the omnidirectional image of the jewelry to be identified and the omnidirectional video of the jewelry under light;

[0185] A first recognition module is used to obtain a surface defect recognition result based on the dimensional data in the three-dimensional static model;

[0186] The second recognition module is used to obtain the inherent defect recognition result of the jewelry based on the lighting dynamic model;

[0187] The comprehensive identification module is used to obtain jewelry quality identification results based on surface defect identification results and internal defect identification results.

[0188] The beneficial effects of the above technology are: combining model building and image recognition with existing jewelry identification standards, realizing the identification of surface and internal defects and quality appraisal of jewelry based on the three-dimensional static model and lighting dynamic model of jewelry, improving the accuracy and efficiency of jewelry identification, and making the jewelry identification process streamlined and standardized.

[0189] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A jewelry quality identification method based on image recognition, characterized in that: include: S1: Based on the omnidirectional image of the jewelry to be identified and the omnidirectional video of the jewelry under illumination, a three-dimensional static model and a dynamic model of the jewelry under illumination are constructed. The process of obtaining the omnidirectional image of the jewelry to be identified under preset shooting environment conditions includes: Obtaining images of the jewelry to be identified from multiple angles under preset shooting environment conditions; Obtaining display parameter distribution data of the captured image based on the display parameter of each pixel in the captured image, performing cluster analysis on sub-distribution data of corresponding types of display parameters in the display parameter distribution data, until the clustering result obtained by the current clustering process is consistent with the clustering result obtained by the previous clustering process, then obtaining multiple display parameter clusters of corresponding types of display parameters based on the latest clustering result; Aggregating the pixel points corresponding to all display parameters included in the display parameter cluster to obtain a plurality of pixel point clusters corresponding to the types of display parameters, and dividing the captured image based on the pixel point clusters to obtain a plurality of divided areas corresponding to the types of display parameters; Determining a list of enhancement coefficients for the corresponding type of display parameters based on the parameter ranges and limit values ​​of the corresponding type of display parameters in each divided area; Performing enhancement processing on the display parameters in the corresponding divided areas based on the enhancement coefficients included in the enhancement coefficient list, to obtain a set of transformed areas of the corresponding types of display parameters; Extracting an underlying feature distribution map of each transformation region in the transformation region set based on a feature extraction model, and marking common features of the underlying feature distribution maps of all transformation regions in the transformation region set to obtain underlying common feature distribution maps of corresponding types of display parameters of corresponding transformation regions; Perform edge extraction on the underlying common feature distribution map to obtain edge lines of corresponding display parameters of corresponding transformation areas; Based on the edge lines corresponding to each display parameter of each transformed area, adjacent captured images are connected and fused to obtain a full-dimensional image of the jewelry to be identified under the preset shooting environment conditions; S2: Obtain surface defect recognition results based on the dimensional data in the 3D static model; S3: Obtaining the intrinsic flaw recognition results of jewelry based on the lighting dynamic model; S4: Obtain jewelry quality appraisal results based on the surface defect identification results and the internal defect identification results.

2. The jewelry quality identification method based on image recognition according to claim 1, characterized in that: S1: Based on the 3D image of the jewelry to be identified and the 3D video of the jewelry under illumination, a 3D static model and a dynamic model of the jewelry under illumination are constructed, including: Obtain a full range of images of the jewelry to be identified under preset shooting environment conditions; Obtain a 360-degree video of the jewelry being authenticated when it is illuminated by a preset light source; Build a three-dimensional static model based on the omnidirectional image; Build a dynamic lighting model based on omnidirectional video.

3. The jewelry quality identification method based on image recognition according to claim 1, characterized in that: Based on the edge lines corresponding to each display parameter in each transformed area, adjacent captured images are connected and fused to obtain a full-scale image of the jewelry to be identified under the preset shooting environment conditions, including: The edge line portion included in the edge line corresponding to each display parameter in the corresponding transformation area is regarded as the main edge line of the corresponding transformation area; and determining sub-edge lines that exist in all edge lines corresponding to multiple display parameters except the main edge line, and determining the total number of display parameter types corresponding to the edge lines including the sub-edge lines; Calculating the inclusion ratio based on the total number of display parameter types corresponding to the edge line containing the sub-edge line and the total number of display parameter types; When the inclusion ratio is not less than the inclusion ratio threshold, the corresponding sub-edge line is used as the secondary edge line; Adjacent captured images are determined based on the primary and secondary edge lines in all transformed areas, and the adjacent captured images are connected and fused to obtain a full-scale image of the jewelry to be identified under preset shooting environment conditions.

4. The jewelry quality identification method based on image recognition according to claim 1, characterized in that: S2: Obtain surface defect recognition results based on the dimensional data in the 3D static model, including: S201: Marking the edges of the 3D static model to obtain the model edges; S202: Match and compare the size data of the model edge with the standard size data of the jewelry to be identified to obtain the surface defect recognition result of the jewelry to be identified.

5. The jewelry quality identification method based on image recognition according to claim 4, characterized in that: S202: Matching and comparing the size data of the model edge with the standard size data of the jewelry to be identified to obtain the surface defect recognition result of the jewelry to be identified, including: Measure the size data of all model edges; Matching the size data of the model edge with the standard size data of the jewelry to be identified to obtain an edge matching result; Based on the matching results, edge size comparison is performed to obtain the surface defect identification results of the jewelry to be identified.

6. The jewelry quality identification method based on image recognition according to claim 1, characterized in that: S3: Obtaining intrinsic flaw recognition results for jewelry based on the dynamic lighting model, including: Extracting the lighting image of the jewelry to be identified from the lighting dynamic model; Perform image analysis on the jewelry area in the light image to obtain the inherent defect identification results of the jewelry to be identified.

7. The jewelry quality identification method based on image recognition according to claim 1, characterized in that: S4: Obtain jewelry quality appraisal results based on surface defect identification results and internal defect identification results, including: Determine the surface identification result based on the preset surface identification standard and the surface defect identification result; Determine the internal identification result based on the preset internal identification standard and the internal defect identification result; The jewelry quality appraisal result is obtained based on the surface appraisal results and the internal appraisal results.

8. The jewelry quality identification method based on image recognition according to claim 7, characterized in that: Obtain jewelry quality appraisal results based on surface and internal appraisal results, including: Determine whether there are any unqualified surface identification items in the surface identification results or unqualified internal identification items in the internal identification results. If so, summarize the unqualified identification items to obtain the jewelry quality identification results of the jewelry to be identified. Otherwise, the jewelry to be identified meets the identification standards and is used as the jewelry quality identification results of the corresponding jewelry to be identified. Among them, unqualified identification items include unqualified surface identification items and unqualified internal identification items.

9. A jewelry quality identification device based on image recognition, characterized in that: include: The model building module is used to build a three-dimensional static model and a dynamic model of the jewelry under illumination based on the omnidirectional image of the jewelry to be identified and the omnidirectional video when illuminated by the light. The process of obtaining the omnidirectional image of the jewelry to be identified under preset shooting environment conditions includes: Obtaining images of the jewelry to be identified from multiple angles under preset shooting environment conditions; Obtaining display parameter distribution data of the captured image based on the display parameter of each pixel in the captured image, performing cluster analysis on sub-distribution data of corresponding types of display parameters in the display parameter distribution data, until the clustering result obtained by the current clustering process is consistent with the clustering result obtained by the previous clustering process, then obtaining multiple display parameter clusters of corresponding types of display parameters based on the latest clustering result; Aggregating the pixel points corresponding to all display parameters included in the display parameter cluster to obtain a plurality of pixel point clusters corresponding to the types of display parameters, and dividing the captured image based on the pixel point clusters to obtain a plurality of divided areas corresponding to the types of display parameters; Determining a list of enhancement coefficients for the corresponding type of display parameters based on the parameter ranges and limit values ​​of the corresponding type of display parameters in each divided area; Performing enhancement processing on the display parameters in the corresponding divided areas based on the enhancement coefficients included in the enhancement coefficient list, to obtain a set of transformed areas of the corresponding types of display parameters; Extracting an underlying feature distribution map of each transformation region in the transformation region set based on a feature extraction model, and marking common features of the underlying feature distribution maps of all transformation regions in the transformation region set to obtain underlying common feature distribution maps of corresponding types of display parameters of corresponding transformation regions; Perform edge extraction on the underlying common feature distribution map to obtain edge lines of corresponding display parameters of corresponding transformation areas; Based on the edge lines corresponding to each display parameter of each transformed area, adjacent captured images are connected and fused to obtain a full-dimensional image of the jewelry to be identified under the preset shooting environment conditions; A first recognition module is used to obtain a surface defect recognition result based on the dimensional data in the three-dimensional static model; The second recognition module is used to obtain the inherent defect recognition result of the jewelry based on the lighting dynamic model; The comprehensive identification module is used to obtain jewelry quality identification results based on surface defect identification results and internal defect identification results.

Images