An image pre-processing method and system for consumable management

Abstract

The application provides an image preprocessing method and system for consumable management, the method comprising: performing double-mode definition detection on an acquired single-channel image of the consumable based on local permutation entropy and frequency domain energy; acquiring a multispectral image of the single-channel image of the consumable satisfying the definition, and determining a material type of the consumable based on the multispectral image; obtaining a depth map of the material type by adopting three-dimensional vision processing on the material type, and realizing automatic segmentation of stacked consumables based on the depth map; performing stability verification on the segmented consumable image, and storing the image after sub-pixel level edge positioning and passing the stability verification. Based on the method, an image preprocessing system for consumable management is also provided. The application fuses multi-dimensional data such as spectral reflectivity, depth information and spatial-frequency domain features, and breaks through the limitation of a single sensor; parameters are dynamically adjusted through meta-learning, and different consumable types and environmental light are adapted.

Description

Technical Field

[0001] This invention belongs to the field of image processing technology, and specifically relates to an image preprocessing method and system for consumables management. Background Technology

[0002] Warehouse consumables management is a key aspect of business operations. For manufacturing companies, the management and use of consumables directly affect production efficiency and cost control. For general companies, reasonable consumables management can ensure smooth production and reduce capital occupation and inventory backlog.

[0003] In the field of consumables data acquisition, traditional data acquisition methods mainly rely on manual operation. Manual shooting is inefficient, and the captured images are easily affected by factors such as shooting angle and lighting conditions, resulting in image occlusion and inconsistent data quality, failing to meet the requirements of high-quality data acquisition. Furthermore, the general image processing methods used in the image preprocessing of the acquired data are unsuitable for the complex environment of warehouse consumables management, cannot effectively denoise different types of noise in the images, and cannot preserve the edge details of the images. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention proposes an image preprocessing method and system for consumables management. It fuses multi-dimensional data, including spectral reflectance, depth information, and spatial-frequency domain features, overcoming the limitations of a single sensor.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] An image preprocessing method for consumables management includes the following steps:

[0007] Dual-modal sharpness detection of acquired single-channel images of consumables is performed based on local permutation entropy and frequency domain energy.

[0008] Acquire multispectral images of consumables that meet the required resolution, and determine the material type of the consumables based on the multispectral images; use 3D vision processing to obtain a depth map of the material type, and achieve automatic segmentation of stacked consumables based on the depth map;

[0009] The segmented consumable images are subjected to stability verification, and images that pass the stability verification are stored after sub-pixel level edge localization.

[0010] This invention also proposes an image preprocessing system for consumable management, including a sharpness detection module, a segmentation module, and a localization module;

[0011] The sharpness detection module is used to perform dual-modal sharpness detection on the acquired single-channel image of the consumable based on local permutation entropy and frequency domain energy;

[0012] The segmentation module is used to acquire multispectral images of consumables that meet the required clarity, and to determine the material type of the consumables based on the multispectral images; it then uses 3D vision processing to obtain a depth map of the material type, and uses the depth map to achieve automatic segmentation of stacked consumables.

[0013] The positioning module is used to verify the stability of the segmented consumable image, and to store the image that has passed the stability verification after performing sub-pixel level edge positioning.

[0014] The effects described in the invention are merely those of the embodiments, and not all the effects of the invention. One of the above technical solutions has the following advantages or beneficial effects:

[0015] This invention proposes an image preprocessing method and system for consumables management. The method includes the following steps: performing dual-modal sharpness detection on the acquired single-channel image of the consumables based on local permutation entropy and frequency domain energy; acquiring a multispectral image of the consumables that meets the sharpness requirements, and determining the material type of the consumables based on the multispectral image; using 3D vision processing to obtain a depth map of the material type, and automatically segmenting stacked consumables based on the depth map; verifying the stability of the segmented consumable images, and storing the images that pass the stability verification after sub-pixel level edge localization. Based on the image preprocessing method for consumables management, an image preprocessing system for consumables management is also proposed. This invention fuses multi-dimensional data such as spectral reflectance, depth information, and spatial-frequency domain features, breaking through the limitations of a single sensor; and dynamically adjusts parameters through meta-learning to adapt to different consumable types and ambient lighting. Attached Figure Description

[0016] Figure 1 This is a flowchart of an image preprocessing method for consumables management proposed in Embodiment 1 of the present invention;

[0017] Figure 2 This is a schematic diagram of the dual-modal sharpness detection process proposed in Embodiment 1 of the present invention;

[0018] Figure 3 This is a schematic diagram of the segmentation and positioning process proposed in Embodiment 1 of the present invention;

[0019] Figure 4 This is a schematic diagram of an image preprocessing system for consumable management proposed in Embodiment 2 of the present invention. Detailed Implementation

[0020] To clearly illustrate the technical features of this solution, the invention will be described in detail below through specific embodiments and in conjunction with the accompanying drawings. The following disclosure provides many different embodiments or examples for implementing different structures of the invention. To simplify the disclosure of the invention, components and arrangements of specific examples are described below. Furthermore, reference numerals and / or letters may be repeated in different examples. This repetition is for simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or arrangements discussed. It should be noted that the components illustrated in the drawings are not necessarily drawn to scale. Descriptions of well-known components, processing techniques, and processes are omitted in this invention to avoid unnecessarily limiting the invention.

[0021] Example 1

[0022] Embodiment 1 of the present invention proposes an image preprocessing method for consumable management, which is used to solve the technical problems existing in the image preprocessing process of consumable management in the prior art.

[0023] Figure 1 This is a flowchart of an image preprocessing method for consumables management proposed in Embodiment 1 of the present invention;

[0024] In step S100, dual-modal sharpness detection is performed on the acquired single-channel image of the consumables based on local permutation entropy and frequency domain energy. Specifically, this process includes: performing blur detection on the acquired single-channel image of the consumables based on local permutation entropy; and performing frequency domain filtering on the acquired single-channel image of the consumables based on frequency domain energy.

[0025] Figure 2 This is a schematic diagram of the dual-modal sharpness detection process proposed in Embodiment 1 of the present invention;

[0026] This algorithm uses the unfold method to divide the image into multiple local windows, flattens and rearranges the window data, extracts all possible subsequences of three consecutive pixel values ​​in each window, sorts these subsequences, calculates the hash value of each subsequence based on the sorted index, calculates the probability distribution of the corresponding pattern by statistically analyzing the frequency of different hash values, and uses the probability to calculate the entropy value. The entropy value reflects the complexity of the pixel arrangement within the local window and judges the degree of image blur. The higher the entropy value, the more complex and disordered the pixel arrangement within the local window, the richer the image details, and the higher the clarity; conversely, if the entropy value is low, it indicates that the pixel arrangement is simple and highly regular, and the image is more blurry.

[0027] For consumable single-channel images Perform sliding window division, window size is Step size is , obtain window set ;in, Image height; Image width; The x-axis of the single-channel image; The vertical coordinate of a single-channel image;

[0028] For each window Triple extraction yielded: ;

[0029] Calculate the permutation hash value of triples: ,in ;in The sort index for the triplet elements;

[0030] Statistical hash value probability distribution: ; For indicator functions;

[0031] Calculate window entropy ; This represents the total number of triples within the window.

[0032] Frequency domain filtering algorithms based on Fast Fourier Transform (FFT) recognize that in the frequency domain, the high-frequency components of an image primarily reflect details such as edges and textures, while the low-frequency components contain the overall structure and grayscale variation trends. Clear images typically possess abundant high-frequency components, corresponding to sharp edges and details. Conversely, when an image is blurry, its high-frequency components significantly weaken or even disappear, resulting in unclear details and blurred edges.

[0033] A Fast Fourier Transform (FFT) is applied to the grayscale image to convert it from the spatial domain to the frequency domain. A circular mask is generated to filter out high-frequency components containing edge and detail information. The normalized high-frequency energy value is obtained by dividing the energy value of the filtered high-frequency components by the total number of pixels in the image.

[0034] If a grayscale image has a low spatial entropy value and a small normalized high-frequency energy value, it indicates that the image exhibits blurred characteristics in both the spatial and frequency domains. Conversely, if the spatial entropy value is high and the normalized high-frequency energy value is large, it indicates that the image retains clear structural and detailed features in both the spatial and frequency domains. Ultimately, it can be determined whether the image is in a blurred state in the spatial and frequency domains.

[0035] Therefore, the process of performing frequency domain filtering on the acquired single-channel image of consumables based on frequency domain energy includes:

[0036] For consumable single-channel images Performing a two-dimensional discrete Fourier transform to obtain the frequency ,

[0037] ;

[0038] in, , ; These are complex coefficients in the frequency domain; Indicates the amplitude of the corresponding frequency component; For imaginary units;

[0039] Calculate the high-frequency energy ratio: ;

[0040] ; This is a high-frequency region;

[0041] Calculate fuzzy scores: ;in To pre-define clear sample benchmark values.

[0042] Local permutation entropy reflects spatial domain sharpness by analyzing the complexity of pixel arrangement patterns, while frequency domain energy analysis reflects image sharpness by the proportion of high-frequency components. An image is considered acceptable if and only if both metrics meet the criteria. Therefore, the sharpness determination process includes:

[0043] Average window entropy ;

[0044] in, Total number of windows; Entropy threshold ;

[0045] Fuzzy fractions ,in ,

[0046] in, For fuzzy threshold; .

[0047] In this application, image sharpness is determined by combining the local arrangement entropy in the spatial domain and the high-frequency energy ratio in the frequency domain.

[0048] In step S200, a multispectral image of the consumables with sufficient clarity is acquired, and the material type of the consumables is determined based on the multispectral image. The material type is then processed using three-dimensional vision to obtain a depth map of the material type, and the stacked consumables are automatically segmented based on the depth map.

[0049] Figure 3 This is a schematic diagram of the segmentation and positioning process proposed in Embodiment 1 of the present invention;

[0050] The process of acquiring multispectral images of consumables includes: acquiring visible light images, near-infrared images, and ultraviolet light images of consumables through a multispectral image acquisition array fixed to the consumable storage area.

[0051] By coordinating three bands—visible light (400-700nm), near-infrared (900-1700nm), and ultraviolet (300-400nm)—the limitations of single-spectrum recognition can be overcome.

[0052] The material type of consumables is identified by matching the spectral characteristics of the visible, near-infrared, and ultraviolet bands. Weighted Euclidean distance is used as a similarity measure, with the visible light channel having the highest weight (60%), followed by near-infrared (30%), and ultraviolet having the lowest weight (10%).

[0053] The process of determining the material type of consumables based on multispectral images includes:

[0054] Establish a standard spectral library ;

[0055] in, ={VIS,NIR,UV} is a set of spectral bands; VIS = visible light, NIR = near-infrared, UV = ultraviolet light; Indicates the first Consumables in the band Standard reflectance at the specified value; Indicates the type of consumables; Indicates the total number of consumables;

[0056] The method for determining the material type is as follows: ;

[0057] ;

[0058] in, To acquire images in real time; Material type.

[0059] The process of automatically segmenting stacked consumables based on 3D vision processing to obtain a depth map of the material type includes:

[0060] The method for generating depth maps is as follows:

[0061] ;

[0062] in, The focal length of the camera; This is the binocular reference distance; This represents the disparity value at position (x, y) in the disparity map.

[0063] Stack quantity calculation: ; The nominal height of a single consumable item.

[0064] Automatic segmentation and counting of stacked consumables is achieved based on depth maps. First, the depth value is compensated and corrected according to the material type. Then, the number of stacked layers is calculated based on the height difference. When the remaining height exceeds 30% of the height of a single piece, the count is compensated.

[0065] In step S300, the segmented consumable image is subjected to stability verification, and the image that passes the stability verification is stored after sub-pixel level edge localization.

[0066] Subpixel-level edge localization is achieved using Zernike moment phase analysis. By calculating the real and imaginary parts of the 31st-order Zernike moment, the normal direction of the edge is derived, and finally, the subpixel-level edge coordinates are obtained.

[0067] The process of performing stability verification on the segmented consumable images and subpixel-level edge localization on the images that pass the stability verification includes:

[0068] Edge point correction formula: ;

[0069] in, For Zernike moments;

[0070] ;

[0071] in, ; ; ρ is the order; ρ is the polar radius; Polar angle; The grayscale function is used to represent the image's grayscale value.

[0072] Decomposition of real and imaginary parts:

[0073]

[0074]

[0075] During subpixel edge detection:

[0076] Edge normal angle ;

[0077] Subpixel offset ;

[0078] Image feature hash values ​​are written to a distributed ledger for storage. The image feature hash is then bound to a geographic location and a timestamp before being written to the blockchain. Only after a smart contract verifies that the image quality, stability, and clarity meet the standards is storage allowed on the blockchain.

[0079] The entire process of this invention is executed sequentially: multispectral imaging → 3D reconstruction → sharpness detection → material identification → stacking counting → edge positioning → blockchain-based evidence storage.

[0080] Embodiment 1 of this invention proposes an image preprocessing method for consumable management, which fuses multi-dimensional data such as spectral reflectance, depth information, and spatial-frequency domain features to overcome the limitations of a single sensor; and dynamically adjusts parameters through meta-learning to adapt to different consumable types and ambient lighting.

[0081] Example 2

[0082] Based on Embodiment 1 of the present invention, an image preprocessing method for consumables management is proposed. Embodiment 2 of the present invention further proposes an image preprocessing system for consumables management. Figure 4 This is a schematic diagram of an image preprocessing system for consumables management proposed in Embodiment 2 of the present invention, including a sharpness detection module, a segmentation module, and a positioning module;

[0083] The sharpness detection module is used to perform dual-modal sharpness detection on the acquired single-channel images of consumables based on local permutation entropy and frequency domain energy;

[0084] The segmentation module is used to acquire multispectral images of consumables that meet the required clarity, and to determine the material type of the consumables based on the multispectral images. The material type is then processed using 3D vision to obtain a depth map of the material type, and the stacked consumables are automatically segmented based on the depth map.

[0085] The localization module is used to verify the stability of the segmented consumable images, and to store the images that pass the stability verification after performing sub-pixel level edge localization.

[0086] The process of the sharpness detection module includes: performing blur detection on the acquired single-channel image of the consumables based on local permutation entropy; and performing frequency domain filtering on the acquired single-channel image of the consumables based on frequency domain energy.

[0087] The process of performing blur detection on the acquired single-channel image of consumables based on local permutation entropy includes:

[0088] For consumable single-channel images Perform sliding window division, window size is Step size is , obtain window set ;in, Image height; Image width; The x-axis of the single-channel image; The vertical coordinate of a single-channel image;

[0089] For each window Triple extraction yielded: ;

[0090] Calculate the permutation hash value of triples: ,in ;in The sort index for the triplet elements;

[0091] Statistical hash value probability distribution: ; For indicator functions;

[0092] Calculate window entropy ; This represents the total number of triples within the window.

[0093] The process of performing frequency domain filtering on the acquired single-channel image of consumables based on frequency domain energy includes:

[0094] For consumable single-channel images Performing a two-dimensional discrete Fourier transform to obtain the frequency ,

[0095] ;

[0096] in, , ; These are complex coefficients in the frequency domain; Indicates the amplitude of the corresponding frequency component; For imaginary units;

[0097] Calculate the high-frequency energy ratio: ;

[0098] ; This is a high-frequency region;

[0099] Calculate fuzzy scores: ;in To pre-define clear sample benchmark values.

[0100] The process of determining sharpness includes:

[0101] Average window entropy ;

[0102] in, Total number of windows; Entropy threshold ;

[0103] Fuzzy fractions ,in ,

[0104] in, For fuzzy threshold; .

[0105] In the segmentation module, visible light, near-infrared, and ultraviolet light images of the consumables are acquired by a multispectral image acquisition array fixed to the consumable storage area.

[0106] The process of determining the material type of consumables based on multispectral images includes:

[0107] Establish a standard spectral library ;

[0108] in, ={VIS,NIR,UV} is a set of spectral bands; VIS = visible light, NIR = near-infrared, UV = ultraviolet light; Indicates the first Consumables in the band Standard reflectance at the specified value; Indicates the type of consumables; Indicates the total number of consumables;

[0109] The method for determining the material type is as follows: ;

[0110] in, To acquire images in real time; Material type.

[0111] The process of automatically segmenting stacked consumables based on 3D vision processing to obtain a depth map of the material type includes:

[0112] The method for generating depth maps is as follows:

[0113] ;

[0114] in, The focal length of the camera; This is the binocular reference distance; This represents the disparity value at position (x, y) in the disparity map.

[0115] Stack quantity calculation: ; The nominal height of a single consumable item.

[0116] In the localization module, the process of verifying the stability of the segmented consumable image and performing sub-pixel-level edge localization on the image that has passed the stability verification includes:

[0117] Edge point correction formula: ;in, For Zernike moments;

[0118] Image feature hash values ​​are written to a distributed ledger for storage. The image feature hash is then bound to a geographic location and a timestamp before being written to the blockchain. Only after a smart contract verifies that the image quality, stability, and clarity meet the standards is storage allowed on the blockchain.

[0119] Embodiment 2 of this invention proposes an image preprocessing system for consumable management, which fuses multi-dimensional data such as spectral reflectance, depth information, and spatial-frequency domain features to overcome the limitations of a single sensor; and dynamically adjusts parameters through meta-learning to adapt to different consumable types and ambient lighting.

[0120] The description of the relevant parts of the image preprocessing system for consumable management provided in Embodiment 2 of this application can be found in the detailed description of the corresponding parts of the image preprocessing method for consumable management provided in Embodiment 1 of this application, and will not be repeated here.

[0121] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that the elements inherent in a process, method, article, or apparatus that includes a list of elements are included. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. Additionally, portions of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of corresponding technical solutions in the prior art have not been described in detail to avoid excessive elaboration.

[0122] While specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art can make other modifications or variations based on the above description. It is neither necessary nor possible to exhaustively describe all embodiments here. Various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. An image pre-processing method for consumable management, characterized by, Includes the following steps: Dual-modal sharpness detection of acquired single-channel images of consumables based on local permutation entropy and frequency domain energy; the process of dual-modal sharpness detection of acquired single-channel images of consumables based on local permutation entropy and frequency domain energy: Blur detection is performed on the acquired single-channel images of consumables based on local permutation entropy; And frequency domain filtering of the acquired single-channel images of consumables based on frequency domain energy; The process of performing blur detection on the acquired single-channel image of consumables based on local permutation entropy includes: Single-channel image of consumable Sliding window partitioning is performed, with a window size of and a step size of , to obtain a window set ; wherein, is the image height; is the image width; is the horizontal coordinate of the single-channel image; is the vertical coordinate of the single-channel image; For each window Triple extraction yielded: ; Calculate the permutation hash value of triples: ,in ;in The sort index for the triplet elements; Statistical hash value probability distribution: ; For indicator functions; Calculate window entropy ; This represents the total number of triples within the window. The process of performing frequency domain filtering on the acquired single-channel image of consumables based on frequency domain energy includes: For consumable single-channel images Performing a two-dimensional discrete Fourier transform to obtain the frequency , ； in, , ; These are complex coefficients in the frequency domain; Indicates the amplitude of the corresponding frequency component; For imaginary units; Calculate the high-frequency energy ratio: ; ; This is a high-frequency region; Calculate fuzzy scores: ;in To pre-define clear sample baseline values; The process of determining sharpness includes: Average window entropy ; in, Total number of windows; Entropy threshold ; Fuzzy fractions , in, For fuzzy threshold; ; Acquire multispectral images of consumables that meet the required resolution, and determine the material type of the consumables based on the multispectral images; use 3D vision processing to obtain a depth map of the material type, and achieve automatic segmentation of stacked consumables based on the depth map; The process of determining the material type of consumables based on multispectral images includes: Establish a standard spectral library ; in, ={VIS,NIR,UV} is a set of spectral bands; VIS = visible light, NIR = near-infrared, UV = ultraviolet light; Indicates the first Consumables in the band Standard reflectance at the specified value; Indicates the type of consumables; Indicates the total number of consumables; The method for determining the material type is as follows: ; in, To acquire images in real time; Material type; The process of automatically segmenting stacked consumables based on 3D vision processing to obtain a depth map of the material type includes: The method for generating depth maps is as follows: ； in, The focal length of the camera; This is the binocular reference distance; This represents the disparity value at position (x, y) in the disparity map. Stack quantity calculation: ; The nominal height of a single consumable item; The segmented consumable images are subjected to stability verification, and images that pass the stability verification are stored after sub-pixel level edge localization.

2. The image preprocessing method for consumables management according to claim 1, characterized in that, The process of acquiring multispectral images of consumables includes: acquiring visible light images, near-infrared images, and ultraviolet light images of consumables through a multispectral image acquisition array fixed to the consumable storage area.

3. The image preprocessing method for consumables management according to claim 1, characterized in that, The process of performing stability verification on the segmented consumable images and subpixel-level edge localization on the images that pass the stability verification includes: Edge point correction formula: ;in, For Zernike moments; The image feature hash value is written into a distributed ledger for storage.

4. An image preprocessing system for consumables management, used to execute the image preprocessing method for consumables management as described in any one of claims 1 to 3, characterized in that, Includes a sharpness detection module, a segmentation module, and a positioning module; The sharpness detection module is used to perform dual-modal sharpness detection on the acquired single-channel image of the consumable based on local permutation entropy and frequency domain energy; The segmentation module is used to acquire multispectral images of consumables that meet the required clarity, and to determine the material type of the consumables based on the multispectral images; it then uses 3D vision processing to obtain a depth map of the material type, and uses the depth map to achieve automatic segmentation of stacked consumables. The positioning module is used to verify the stability of the segmented consumable image, and to store the image that has passed the stability verification after performing sub-pixel level edge positioning.

Images