Mirror surface object automatic focusing shooting and depth-of-field continuation method under common light source

Abstract

The invention discloses an automatic focusing shooting and depth-of-field continuation method for a mirror surface object under a common light source, and relates to the technical fields of automatic focusing, depth-of-field continuation and the like. The method comprises the following steps: image acquisition: acquiring serialized image information of a mirror surface object under different shooting conditions; automatic focusing is carried out, a focusing interval is determined from the image sequence, and sub-sequence images used for fusion are selected; image fusion: fusing the effective picture subsequences which are selected by automatic focusing and carry the local detail information of the target object to generate a full-focus image with the detail features of the object; image enhancement: increasing detail information and definition of the full-focus image; and image classification: identifying personalized information of different objects, including color, pattern, process and category. According to the invention, the defects that the mirror surface object is fine in structure and easy to reflect light to influence the imaging quality and macro shooting is limited by depth of field are overcome, the full-focus image is automatically and efficiently output, and all details of the mirror surface object are presented on one image.

Description

Technical Field

[0001] The present invention belongs to the technical field of optical imaging and image processing, and relates to technical fields such as automatic focusing and depth of field extension, and in particular to a method for automatic focusing shooting and depth of field extension of mirror objects under ordinary light sources. Background Art

[0002] Jewelry is widely favored for its unique beauty and artistic value. With the rise of online jewelry sales, high-quality jewelry images have become a necessary prerequisite for promotion and transactions. High-precision imaging technology is the focus of the industry to show the detailed features of jewelry. Traditional jewelry imaging methods mainly rely on manual focus, which requires a high level of skill from the photographer. When faced with large-scale, high-frequency imaging needs, it is not only time-consuming and labor-intensive, but the stability of image quality is even more difficult to guarantee.

[0003] In recent years, image processing technology has been continuously developing, and obtaining high-definition images of jewelry products has become a reality. It is the basis for two-dimensional and three-dimensional display of products and their derivative applications. Jewelry is small in size and fine in structure. Macro photography is generally used when shooting. The depth of field is limited, and it is impossible to obtain a fully focused image at one focal length, resulting in loss of detail information. At present, most of the display pictures of jewelry products focus on the design feature parts, and the rest of the parts are out of focus and blurred, which cannot present all the wonderful details of the jewelry. On the other hand, jewelry raw materials are generally mirror materials such as metals and gemstone crystals. When photographed under ordinary light, they are easily affected by mirror reflection, refraction and overexposure. The image is prone to bright spots, reflections, blurs and other defects. It is difficult to obtain a clear image only through photography techniques. In addition, the post-processing method of jewelry images usually uses some image processing software to manually trim and beautify the image, which has high time and labor costs, and cannot achieve real-time image display. Summary of the invention

[0004] In view of the above problems, the present invention proposes a method for automatic focus shooting and depth of field extension of mirror objects under ordinary light sources. When faced with the needs of imaging large quantities and multiple types of mirror objects, the imaging defects of mirror objects such as easy reflection and refraction under ordinary light sources can be overcome, the acquisition efficiency and quality stability of full-focal-length images can be significantly improved, the high-definition details of the objects can be fully displayed, and the user's visual experience can be improved.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] The present invention is a method for automatically focusing and photographing a mirror object under a common light source and extending the depth of field, including the following sub-methods:

[0007] Image acquisition is used to collect serialized image information of objects under different shooting conditions (angle, focal length, lighting, etc.); autofocus is used to automatically and accurately determine the focus interval from the image sequence and select subsequence images for fusion; image fusion is used to fuse the effective pictures with local detail information of the target object selected by autofocus to generate a fully focused image with all the detail features of the object; image enhancement is used to enhance the clarity of the fully focused image, further increase the high-frequency detail information of the image, display details, and optimize the visual effect; image classification is used to identify information of different objects, including color, pattern, craftsmanship, category, etc.

[0008] The present invention is a method for automatically focusing and photographing a mirror object under a common light source and extending the depth of field, comprising the following steps:

[0009] Step S1, the macro surround shooting device is initialized, and parameters such as light source brightness and shooting angle are adjusted according to different surface features of the mirror object;

[0010] Step S2, intermittently adjusting the camera shooting focal length through image acquisition, and shooting a sequence of images of the mirror object at different focal lengths from top to bottom;

[0011] Step S3, screening the image sequence, and obtaining an image subsequence containing focus information through automatic focusing;

[0012] Step S4, fusing the filtered image subsequences in sequence by an image fusion method until all images are fused to generate a fully focused image of the mirror object;

[0013] Step S5, further improving the display clarity and detail presentation of the fully focused image by image enhancement;

[0014] Step S6, collecting the fully focused images of different mirror objects generated in the above steps, and classifying the input new mirror object images through image classification.

[0015] Preferably, in the above step S1, the macro surround shooting device includes an imaging system, and the imaging system is centered on the optical axis of the lens, and from top to bottom in order of spatial hierarchy: a camera; a lens, a fixed-focus lens; a light source, an ordinary white light source, including a top light source coaxially installed with the lens, and a front light source and a rear light source with adjustable illumination angles; a placement table, a flat surface, about 100 mm away from the lens, and the light sources are surrounded and distributed above, in front of, and behind the placement table.

[0016] Preferably, in the above step S1, the macro surround shooting device includes a driving system to realize zoom focusing from the top to the bottom of the object when the camera is shooting; to realize up and down and pitch movement of the camera; and to realize the translation of the plane stage.

[0017] Preferably, in the above step S2, when collecting the image of the object, under the illumination of ordinary white light source, a fixed ratio parameter calculation is performed on the macro surround shooting device through the clarity curve, and then the device is driven to scan and shoot from top to bottom to obtain the image sequence.

[0018] The above fixed ratio parameter calculation includes the following steps:

[0019] Step S21, acquiring an image sequence and calculating a clarity evaluation curve;

[0020] Step S22, recording the peak position of the clarity evaluation curve, and recording the image sequence positions where the top and bottom clear positions of the photographed object are located by observing the image sequence;

[0021] Step S23, respectively calculating the ratio parameters of the top relative to the peak position and the bottom relative to the peak position.

[0022] Preferably, in the above step S3, interfering pixels in the image sequence are removed by image segmentation, and the image subsequence to be fused is selected by calculating the clarity of the segmented image through automatic focusing.

[0023] The above-mentioned auto-focus is composed of, but not limited to, an image clarity evaluation function based on a Laplace operator and an image segmentation algorithm based on a morphological operator.

[0024] The above-mentioned selection of the image subsequence to be fused includes the following steps:

[0025] Step S31, performing image segmentation on the entire image sequence;

[0026] Step S32: locating the focus image interval according to the clarity of each area after the image segmentation.

[0027] The above-mentioned image segmentation includes but is not limited to grayscale value, binarization, and morphological operations, which can separate the foreground and background of the image so that the focus window only includes the part of the photographed object.

[0028] Preferably, in the above step S4, the subsequence images are fused in sequence through an image fusion model to generate a fully focused image.

[0029] The above-mentioned image fusion model includes but is not limited to a Laplacian pyramid model for multi-focus images.

[0030] The fusion of subsequence images includes the following steps:

[0031] Step S41, extracting the focused image sequence to be fused;

[0032] Step S42, calculating and comparing the high-frequency and low-frequency information of the image, and fusing the high-frequency information and the low-frequency information respectively;

[0033] Step S43: reconstructing a fully focused image of the object.

[0034] Fusion of subsequence images, including but not limited to fusion of two images.

[0035] Preferably, in the above step S5, the fully focused image is processed by an image enhancement algorithm based on human visual perception to achieve overall and edge enhancement of the image region of interest.

[0036] The above-mentioned image enhancement algorithm based on human visual perception adopts a model of visual center-surround antagonistic receptive field.

[0037] The above-mentioned overall and edge enhancement of the image region of interest includes the following steps:

[0038] Step S51, extracting a region of interest in the image;

[0039] Step S52, performing overall pixel enhancement on the image region of interest;

[0040] Step S53: perform edge pixel enhancement on the image region of interest.

[0041] Preferably, in the above step S6, firstly, image sequences of objects of different types and styles at different shooting angles and illuminations are collected according to steps S1 to S4, and fully focused images of each object are synthesized according to step S5. These fully focused images are used as data sets of the image classification network to train the image classification model to determine the object category.

[0042] The above-mentioned image classification models include but are not limited to convolutional neural network image classification models.

[0043] After adopting the above scheme, the present invention can automatically achieve depth of field extension when shooting objects through automatic focus and image fusion processing, make full use of the effective information of each image in the image sequence, reduce the loss of high-frequency information of the image, and achieve high-definition full-focus image. The present invention effectively solves the image blur caused by the reflection of objects under ordinary light sources through image enhancement, and further realizes image beautification on the basis of retaining the realism of the image. The present invention adopts an automated processing flow, which significantly reduces the processing time and labor costs compared with traditional manual shooting and subsequent manual photo editing, improves the efficiency of image acquisition and processing, ensures the stability of the image, and provides convenience for merchants and customers. The present invention is suitable for shooting various mirror objects with subtle reflections and fine structures. The shooting environment is easy to build, the operation is simple, and the robustness is high. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 is a flow chart of one embodiment of the present invention.

[0045] Figure 2is a schematic diagram of a macro surround shooting device in one embodiment of the present invention.

[0046] Figure 3 FIG. 4 is a flowchart of automatic focusing according to an embodiment of the present invention.

[0047] Figure 4 The figure is a flowchart of image fusion according to an embodiment of the present invention. DETAILED DESCRIPTION

[0048] The technical solution of the present invention will be described in detail below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] It should be noted that the drawings are all in a very simplified form and use non-precise proportions, and are only used to conveniently and clearly assist in explaining the purpose of the embodiments of the present invention. The structures shown in the drawings are part of the actual structures.

[0050] The mirrored objects mentioned in the present invention have the characteristics of complex surface, fine structure and surface reflection, including but not limited to ring-type jewelry products.

[0051] The various algorithms mentioned in the present invention are not limited to a certain algorithm.

[0052] The embodiment of the present invention takes an ordinary annular ring (hereinafter referred to as ring) as an example, and assumes that the height of each type of ring is basically the same.

[0053] See also Figure 1 As shown, the present invention provides a method for automatically focusing and photographing a mirror object under a common light source and extending the depth of field, comprising the following steps:

[0054] Step S1, the macro surround shooting device is initialized, and the light source brightness and other parameters are adjusted according to the different surface features of the ring (gemstone color, metal material, inlay process and shape, etc.).

[0055] This embodiment can be applied to macro surround shooting equipment, see Figure 2As shown, the macro surround shooting device includes an imaging system 100 and a driving system 200, the imaging system includes a lens 110, a camera 120, a light source 130, and a storage table 140; the resolution of the camera 120 is 4024×3036; the light source 130 is divided into a top light source 131, a front light source 132 and a rear light source 133, wherein the top light source 131 is a ring light source, the front light source 132 is a surface light source with an illumination angle of 30°, and the rear light source 133 is a surface light source with an illumination angle of 30°, and the brightness of the three light sources can be adjusted; the driving system 200 includes a fixing frame 210 for mounting the imaging system 100, and driving motors 211, 212, 213, and 214 for driving the lens 110 and the camera 120, the storage table 140, and the fixing frame 210 to move. The driving motor 211 controls the lens to adjust the aperture so as to achieve focusing shooting of different areas of the ring from the top to the bottom. The driving motor 212 drives the placement table 140 to rotate in a plane. The rotation angle is called longitude. The horizontal direction is 0°, and the longitude adjustment range is 0° to 360°. The driving motor 213 drives the lens 110 and the camera 120 in the imaging system 100 to move in the pitch direction. The pitch angle is called latitude. The latitude adjustment range is 0° to 90°. When the lens 110 and the camera 120 in the imaging system 100 are parallel to the horizontal plane, it is recorded as 0°. When the lens 110 and the camera 120 are perpendicular to the horizontal plane, it is recorded as 90°. The placement table 140 is a flat disk with a claw in the center for fixing the ring 300. The placement table 4 and the top light source 131 are both parallel to the horizontal plane.

[0056] In step S1, image acquisition includes the following steps:

[0057] Step S11, place the ring 300 on the storage table 140, and the macro surround shooting device is automatically initialized after being started. The brightness of the light source 130 is the default brightness, and the driving mechanism 200 drives the lens 110, the camera 120, and the top light source 131 in the imaging system 100 to a latitude of 90°.

[0058] Step S12, adjusting the shooting position and the brightness of the light source 130, the annular light source 131, the front light source 132, and the rear light source 133 can be adjusted independently.

[0059] Step S2, intermittently adjusting the shooting focal length of the camera 120 through image acquisition, and shooting a sequence of images of the ring 300 at different focal lengths from top to bottom.

[0060] In step S2, when collecting object images, under the illumination of ordinary white light source, for each category of mirror objects, firstly, a fixed ratio parameter calculation is performed on the macro surround shooting device through the clarity curve, and then the device is driven to scan and shoot from top to bottom to obtain an image sequence.

[0061] The mirrored object selected in this embodiment is a ring. If it is replaced with other categories such as pendants, bracelets, etc., it is sufficient to calculate a fixed ratio parameter for each category.

[0062] The above-mentioned macro surround shooting device performs a fixed ratio parameter calculation, including the following steps:

[0063] Step S21, placing and fixing the product, adjusting the light source, using a surround macro camera to shoot the product layer by layer from top to bottom, and saving the shot image sequence.

[0064] In step S22, please refer to the image clarity function of formula (1) and the clarity function of a certain point in the image of formula (2), input the image into the image Laplace clarity evaluation function written in MATLAB, calculate the clarity evaluation curve, and record the peak position F of the clarity evaluation curve by observing the image sequence positions where the top and bottom clear positions of the photographed object are located in the image sequence. By observing the clarity changes of each image in the image sequence, record the top clear position F1 and the bottom clear position F2 of the photographed product.

[0065]

[0066] z(x,y)=g(x-1,y)+g(x+1,y)+g(x,y-1)+g(x,y+1)-4g(x,y) (2)

[0067] Among them, f is the image clarity value, M×N is the number of image pixels, g(x,y) is the pixel value of a certain point in the image, and z(x,y) is the clarity value of the point.

[0068] Step S23, respectively calculate the ratio parameters UB and LB of the top of the product relative to the peak position and the bottom of the product relative to the peak position. Please refer to formula (3) for the function definition:

[0069]

[0070] Step S24, replace the product with a different type and repeat the above steps.

[0071] Steps S21 to S24 are preparations for each type of mirror object. If the type of ring does not change, there is no need to perform this operation every time you take a photo.

[0072] Step S3, screening the image sequence, and obtaining an image subsequence containing focus information through automatic focusing.

[0073] The shooting modes are divided into single shooting, surround shooting and custom angle range shooting. You can freely choose according to the characteristics of the ring and the placement posture to obtain the full sequence D of the ring's image.

[0074] See also Figure 3, the autofocus process includes the following steps:

[0075] Step S31, performing image segmentation on the entire sequence D of images obtained by photographing the ring, comprises the following steps:

[0076] Step S311, selecting an image located at the center of the image sequence, and processing the brightness and pixels of the image to obtain a mask of the image sequence. The processing includes:

[0077] Step S311-1, convert the obtained RGB image into a grayscale image. Please refer to formula (4) for the process:

[0078] Gray=R*0.299+G*0.578+B*0.114 (4)

[0079] Step S311-12 performs a binarization operation on the grayscale image so that it contains only black and white pixels. Please refer to formula (5) for the process:

[0080]

[0081] Wherein, x is the pixel value of a certain point in the image, T is the set threshold, and in this embodiment, the T value is set to 90;

[0082] Step S311-3, perform morphological operator operation on the binary image, alternating dilation operation and erosion operation for 3 times. Please refer to formula (6) for the process:

[0083]

[0084] Wherein, A is a binary image, B is a morphological operator, and in this embodiment, the size of the dilation and erosion operators is assumed to be the same, both of which are 3×3 matrices. represents the dilation operation, $ represents the erosion operation, and z is the pixel at the current operation position.

[0085] Step S311-4, performing a small area removal operation on the binary image after the morphological operation, by detecting the eight-neighborhood pixel points of each 0-value pixel in the image to determine whether the pixel point is a pixel point in a small area. If the pixel point in a small area exceeds the set threshold, the area is changed to a 255-value pixel point to obtain the final image mask.

[0086] Step S312, weighted fusion is performed on the image mask and each image in the entire image sequence. The process can be described as separating the main object area and the background area in the image, retaining the main ring area, removing the background area, forming a new image sequence D1 and saving it. The process can be referred to in formula (7):

[0087] M f =M*I src +(1-M)*Mb (7)

[0088] Among them, M is the image mask, M f is the image in the image sequence D1, i.e. the image of the main area of ​​the product after separation, M b For white background, I src is an image in the image sequence D.

[0089] Step S313, perform secondary segmentation on the image mask in step S311 to generate a top mask and a bottom mask. The image sequence D1 is weightedly fused with the top mask and the bottom mask respectively, thereby segmenting the top and bottom of the ring to form a top image sequence D2 and a bottom image sequence D3.

[0090] Step S32, respectively calculate and compare the image clarity values ​​of the image sequence D2 and the image sequence D3, locate and save their peak positions, record the top sequence peak position as P1, the bottom sequence peak position as P2, and save the image subsequence within [P1, P2] as the image to be fused.

[0091] The mirror object selected in this embodiment is a ring. When a flat mirror object such as a pendant is selected, the proportion parameter method can be used to obtain the interval to be fused.

[0092] All the image sequences obtained are stored in the order in which they were taken.

[0093] Step S4: The filtered image subsequences are sequentially fused by an image fusion module until all images are fused to generate a fully focused image of the object.

[0094] In this embodiment, a Laplacian pyramid image fusion model with a total of 3 layers is constructed, and the images within the focus interval to be fused are input and the fused image is output to achieve depth of field extension.

[0095] See also Figure 4 , taking the fusion of two images and the three-layer image pyramid as an example, A and B are the images to be fused, LP1 is the high-frequency information image of the first layer of the image pyramid, LP2 and LP3 are the low-frequency information images of the second and third layers of the pyramid respectively. After the corresponding images of each layer of the pyramid are fused, new LP1, LP2, and LP3 are generated, and G0 is the fully focused image after fusion.

[0096] Image fusion includes the following steps:

[0097] Step S41, constructing a Gaussian image pyramid of the image in the focus interval, with a total of 3 layers. Taking one of the images as an example, the generation process of the n+1th layer image is shown in equation (8):

[0098]

[0099] Among them, W(m,n)=W(m)*W(n) is a Gaussian convolution kernel of size 5×5.

[0100] Step S42, constructing a Laplacian image pyramid through the Gaussian image pyramid, with a total of 3 layers, and the nth Laplacian pyramid is obtained by subtracting the n+1th Gaussian pyramid from the nth Gaussian pyramid after upsampling. Taking one of the images as an example, the process is as follows:

[0101] L n =G n -expand(G n+1 ) (9)

[0102] The first layer of the Laplace image pyramid is the high-frequency information layer of the image, and the second and third layers are the low-frequency information layers.

[0103] Step S43 , according to different fusion strategies, the images of the same layers of each image pyramid are fused respectively to obtain the overall Laplacian image pyramid of the focus interval image.

[0104] Fusion strategies are divided into pixel-level, feature-level and region-level. This embodiment adopts the highest-level region-level fusion strategy.

[0105] For the focus determination of the lowest frequency information layer, variance and entropy are used to perform region focus determination, which can be described as:

[0106]

[0107] Among them, formula (10) is the variance calculation formula; formula (11) is the image entropy calculation formula.

[0108] In this embodiment, entropy uses the algorithm of image unary entropy. First, the frequency of each gray level 0-255 in the entire image is counted, and then the entropy of the center point is calculated according to formula (11). In the Laplace image, the edge area accounts for a small proportion and the smooth area accounts for a large proportion. Since the -log function is a negative correlation curve, the entropy of a pixel in an edge area (with a small frequency value) will be greater than the entropy of a pixel in a smooth area (with a large frequency value); formula (12) is a focus discrimination method. When the variance and entropy of the A image in the local block where a certain pixel is located are both greater than those of the B image, the value of the current pixel of the top layer of the fused Laplace is the value of the corresponding pixel of the top layer of the Laplace of the A image; otherwise, the value of the current pixel of the top layer of the fused Laplace is the value of the corresponding pixel of the top layer of the Laplace of the B image. When it does not belong to the above two situations, the value of the current pixel of the top layer of the fused Laplace is the average value of the corresponding pixels of the top layers of the Laplace of the A image and the B image.

[0109] The fusion strategy for the remaining layers of the image pyramid is as follows: use a convolution kernel similar to a Gaussian kernel to convolve each pixel of the image, use the convolution result as the entropy of the current pixel, and use the convolution operation to aggregate the pixel values ​​of the neighborhood. Finally, the value of a pixel in the fused Laplacian image is equal to the value of the corresponding pixel in the layer with the largest entropy value.

[0110] Step S44, fuse the whole Laplacian image pyramid, reconstruct the full focus image of the ring area, and realize the depth of field extension. The process is shown in equation (13):

[0111] G0=LP0+Expand(LP1+Expand(LP2+...+Expand(LP N ))) (13)

[0112] The Laplacian image pyramid construction is not limited to 3 layers, but must be greater than or equal to 3 layers, that is, N ≥ 3.

[0113] Step S5, further improving the display clarity and detail presentation of the fully focused image through image enhancement, removing blur caused by mirror reflection, etc., and improving the visual experience.

[0114] Step S51, extracting the region of interest in the image.

[0115] Step S52, performing image enhancement processing on the extracted area based on the principle of the visual center-surround antagonistic receptive field model, and the process thereof is shown in equation (14);

[0116]

[0117] G1=G0*K R×R (15)

[0118] Among them, K is the visual center-surround antagonistic receptive field operator matrix, with a size of R×R, R is the pixel outside the receptive field radius, and r is the receptive field range radius. R1 and R2 are the enhancement coefficients corresponding to R and r respectively. G1 is the convolution of the Laplacian pyramid reconstructed image and matrix K.

[0119] Step S53, change The edge pixels of the image G1 are subjected to secondary image enhancement using the method of step S52.

[0120] Step S6, replace the rings of different categories, generate a large number of fully focused images of the rings with different features through steps S1 to S5, collect these fully focused images of the rings to form a ring dataset, and classify unknown new objects through the image classification module.

[0121] After constructing the data set, you can set classifications according to various standards, such as color (fine goods, defects), pattern (flowers, figures, fonts, etc.), craftsmanship (plain rings, clusters, halos, etc.), and categories (rings, necklaces, pendants, etc.). Then, you can classify the data set according to these categories, and then input each category of data set into the neural network, set the loss function, and train the model multiple times using the gradient descent method. After the training is completed, you can input a new ring image into the model, and the model will get the classification result of the ring.

[0122] At this point, the specific implementation steps of this method have been described.

[0123] In summary, the above are only preferred embodiments provided by the present invention and are not intended to limit the protection scope of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the invention shall be included in the protection scope of the present invention.

[0124] Those skilled in the art can understand that the modules in the device in the embodiment can be distributed in the device in the embodiment according to the description of the embodiment, or can be changed accordingly and located in one or more devices different from the embodiment. The modules in the above embodiment can be combined into one module, or can be further divided into multiple sub-modules.

[0125] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for automatically focusing and extending the depth of field of a mirror object under a common light source, characterized in that: The method comprises: Image acquisition: collecting serialized image information of mirror objects under different shooting conditions (including angle, focal length, and lighting); Autofocus, determine the focus interval from the image sequence and select the subsequence image for fusion; Image fusion: fuse the valid image subsequences with local detail information of the target object selected by autofocus to generate a fully focused image with object detail features; Image enhancement, which increases high-frequency detail information and clarity of fully focused images; Image classification, identifying personalized information of different objects, including color, pattern, craftsmanship, and category.

2. The common light source according to claim 1, characterized in that: It is a white light source.

3. The mirror object according to claim 1, characterized in that: Mirrored objects, including jewelry objects, with different characteristics, including complex surfaces, fine structures, and surface reflections.

4. The image acquisition according to claim 1, characterized in that: The shooting focal length is adjusted intermittently, and the image sequence of the mirror object at different focal lengths is shot from top to bottom by zooming, and obtained in the following manner: When collecting an image of an object, under the illumination of the ordinary light source, a fixed-ratio parameter calculation is performed on the macro surround shooting device through a clarity curve, and then the macro surround shooting device is driven to scan and shoot from top to bottom to obtain an image sequence; The fixed ratio parameter calculation is performed once and obtained by the following method: Acquire the image sequence, and calculate the clarity evaluation curve, where the clarity evaluation curve is an image clarity evaluation function based on a Laplace operator; Recording the peak position of the clarity evaluation curve, and recording the image sequence positions where the top and bottom clear positions of the mirror object are located by observing the image sequence; The ratio parameters of the top relative to the peak position and the ratio parameters of the bottom relative to the peak position are calculated respectively.

5. The automatic focusing according to claim 1, characterized in that: Filtering the image sequence, and obtaining an image subsequence containing focus information by autofocusing; The automatic focusing is composed of an image clarity evaluation function based on a Laplace operator and an image segmentation algorithm based on a morphological operator; The image segmentation, including gray value, binarization, and morphological operation processing, is obtained in the following way: Performing image segmentation on the image sequence; The focus interval of the image sequence is located according to the clarity of each area after image segmentation.

6. The image fusion according to claim 1, characterized in that: fusing the filtered image subsequences in sequence until all images are fused to generate a fully focused image of the mirror object; The image fusion includes a Laplacian pyramid model for multi-focus images; The Laplace pyramid model for multi-focus images has 3 pyramid layers, the low-frequency image fusion strategy is a comprehensive judgment of image variance and image entropy, and the high-frequency image only uses image entropy for focus judgment; The fused subsequence image is obtained in the following manner: Extracting the obtained focused image subsequence to be fused; Calculating and comparing the high-frequency and low-frequency information of the image, and fusing the high-frequency information with the low-frequency information respectively; reconstructing the fully focused image of the specular object; The fused subsequence images include the fusion of two or more images.

7. The image enhancement according to claim 1, characterized in that: Performing image enhancement algorithm processing based on human visual perception on the fully focused image; The image enhancement algorithm based on human visual perception adopts a model of visual center-surround antagonistic receptive field.

8. The image classification according to claim 1, characterized in that: Collecting fully focused images of different types of mirror objects, and classifying input new mirror object images by image classification; The image classification includes a convolutional neural network image classification model; The image classification, the classification label includes color, pattern, process, and category information.

9. The macro surround shooting device according to claim 4, characterized in that: Including imaging system and driving system: The imaging system comprises a lens, a camera, a light source, and a storage platform; The camera has a resolution of 4024×3036; The light source is a white light source, which is divided into a top light source, a front light source and a rear light source, and the brightness can be adjusted; The storage table is in the form of a flat disk with a retractable card in the center for fixing different mirror objects; The driving system comprises a fixing frame and a driving motor; The fixing frame is used to load the imaging system and the driving motor; The drive motors are divided into three groups, including: A group of devices for driving the zoom and fixed-focus shooting of the lens and camera in the imaging system; A group of components for driving the lens and the camera in the imaging system to move in the vertical and pitch directions, wherein the pitch angle is called latitude, and the latitude adjustment range is 0° to 90°. When the lens and the camera in the imaging system are parallel to the horizontal plane, it is recorded as 0°, and when the lens and the camera in the imaging system are perpendicular to the horizontal plane, it is recorded as 90°; One group is used to drive the storage table to rotate in a plane. The rotation angle is called longitude, with the horizontal direction being 0° and the longitude adjustment range being 0° to 360°.

Images