An image post-processing optimization management method for intelligent taking of electric power materials

Abstract

The application provides an image post-processing optimization management method for intelligent taking of a cabinet of electric power materials, and belongs to the technical field of electric power material management, and comprises the following steps: S1, taking and placing registration; S2, overlay image acquisition; S3, feature extraction; S4, recognition matching; S5, inventory management whole process and the like; the application collects multi-angle images of the materials through a high-resolution camera, constructs a multi-directional convolution framework by using a spatial gradient energy kernel, and strengthens the extraction capability of geometric contour, surface texture and identification features. The core technology adopts three-dimensional differential operation to analyze the edge strength, curvature change and deformation characteristics step by step, combines a multi-scale texture analysis and an adaptive light compensation algorithm, eliminates environmental interference and generates a unique feature coupling factor, matches a pre-stored feature fingerprint library in real time through an elliptical envelope detection algorithm, accurately identifies the types and quantities of the materials, and forms a closed-loop management from permission verification, accurate identification to intelligent early warning.

Description

Technical Field

[0001] This invention relates to the field of power material management, and in particular to an image post-processing optimization management method for power materials in an intelligent requisition cabinet. Background Technology

[0002] With the rapid development of the power industry, the demand for intelligent management of power materials (such as composite insulators and cable terminations) is becoming increasingly urgent. Traditional power material requisition management relies heavily on manual registration and inventory checks, which suffers from low efficiency, error-proneness, and poor real-time performance. In recent years, intelligent requisition cabinets based on image recognition have been gradually introduced to optimize processes by automating the collection and identification of material information.

[0003] For example, patent CN119006764A discloses a smart medicine cabinet medicine management method and system. It utilizes an in-cabinet camera to collect medicine entry / exit slips and medicine images, obtains entry / exit information through image recognition, and then verifies compliance by combining it with a real-time inventory table. This automated process reduces manual intervention and improves the speed and accuracy of medicine entry / exit. However, in practical applications, electrical materials are diverse and structurally complex (such as anti-slip threads, umbrella skirt structures, etc.). Traditional image feature extraction methods (such as single gradient or texture analysis) struggle to fully capture their multi-dimensional features, leading to difficulties in distinguishing similar materials, a high misidentification rate, and insufficient image recognition accuracy.

[0004] Therefore, an image post-processing optimization management method for intelligent requisition cabinet power materials was introduced. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to overcome the shortcomings of existing image feature extraction methods, which have great difficulty in distinguishing similar materials and have a high misidentification rate when faced with a variety of power materials with complex structures. The present invention proposes an image post-processing optimization management method for power materials in intelligent requisition cabinets.

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: an image post-processing optimization management method for intelligent requisition cabinet power materials, comprising the following steps:

[0007] S1: Retrieval and Placement Registration. Users click the function button on the touch screen of the smart requisition cabinet, select the function of retrieving or storing power materials, and unlock the door after facial recognition to determine authorization, thus completing the operator registration.

[0008] S2: Coverage image acquisition. After the user takes out or puts in the power supplies, he or she closes the door lock. During this process, the camera inside the cabinet takes all-round images of the items taken out and the power supplies inside the cabinet, completes the acquisition of raw image data, and constructs a multi-directional convolutional framework containing spatial gradient energy kernel.

[0009] S3: Feature extraction. Construct a multi-feature dynamic coupling model for power materials to extract features from the original image. Use the nonlinear multi-feature dynamic coupling model to process the original image and calculate the feature coupling factor of the power materials based on the original image.

[0010] S4: Identification and matching. A feature fingerprint database of power materials is established using feature coupling factors. The calculated feature coupling factors are matched with the feature fingerprint database in real time using the elliptical envelope detection method to determine and identify the type and quantity of power materials to be picked up and placed.

[0011] S5: Full-process inventory management, updating the inventory status of materials in the smart requisition cabinet based on the matching results and triggering early warning signals.

[0012] Furthermore, in S3, the steps for calculating the feature coupling factor in the multi-feature dynamic coupling model are as follows:

[0013] Step 1: Perform three-dimensional differential operations on the original image data, and calculate the 1st to 3rd order partial derivatives along the X-axis to extract the geometric features of the power materials;

[0014] Step 2: Parallel calculation of multi-scale differential texture density of power materials, using 12 directional kernels to cover 360-degree space, with each directional kernel calculating an adaptive illumination compensation factor based on the mean and variance of local image patches;

[0015] Step 3: Nonlinearly couple the differential result from Step 1 with the output from Step 2 to generate characteristic numerator terms;

[0016] Step 4: Calculate the cubic gradient integral of the second-order mixed partial derivative of the image as the denominator of the noise suppression term, and divide the numerator by the denominator to generate the characteristic coupling factor of the power material.

[0017] Furthermore, the formula for calculating the feature coupling factor in the multi-feature dynamic coupling model is as follows:

[0018]

[0019] In the formula: I represents the input image of power materials, which is a two-dimensional matrix, where each element I(i,j) corresponds to the gray value of a pixel, i represents the row coordinate, and j represents the column coordinate; k is the loop index used to traverse n feature dimensions, and the value of k is from 1 to n; m represents the order of the derivative used for each feature dimension k, and the value of m is from 1 to 3. This represents the convolution output of the gradient energy kernel in the k-th direction; This represents multi-scale differential texture density; (W, H) represents the adaptive illumination compensation factor; (W, H) represents the image size. Represents the image gradient energy kernel Perform partial derivatives from the first to the third order along the x-direction; This represents the calculation of the second-order mixed partial derivatives of the image in the x and y directions.

[0020] Furthermore, in S4, the elliptical envelope detection method verifies the characteristic coupling factor of power materials as follows:

[0021] Let the fingerprint database of electrical materials be Λ={Γ(I1), Γ(I2), ..., Γ(I... m )};

[0022] Real-time calculation of the image Γ(Ix) of the material to be identified;

[0023] Matching is performed using the elliptic envelope detection method: Where ε is the dynamic tolerance threshold, which is a constant. When the feature coupling factor Γ(Ix) of the image of the material to be identified is different from that of Γ(Ix) in the feature fingerprint database, the dynamic tolerance threshold is determined. k )satisfy The match is considered successful upon confirmation.

[0024] Furthermore, the formula for calculating the illumination compensation factor in step two is as follows:

[0025] ,

[0026] in, This represents the mean value of a local image patch in an image of power materials. This represents the local image patch variance of an image of electrical materials.

[0027] Furthermore, the mean value of local image patches in the image of power materials is calculated as follows:

[0028] ,

[0029] The local image patch variance of power material images is calculated as follows:

[0030]

[0031] in, This represents a local image block centered at pixel k; N is the total number of pixels within the local block.

[0032] Furthermore, in step one, when performing three-dimensional differential operations on the original image: a 5×5 Sobel convolution kernel is used along the x-axis to calculate the first, second, and third order partial derivatives. The first order partial derivative is used to extract the edge intensity of the power materials, the second order partial derivative is used to extract the curvature change of the power materials, and the third order partial derivative is used to extract the higher-order deformation features of the power materials.

[0033] Furthermore, in S2, the key information of the original images of the power materials collected by the camera inside the requisition cabinet includes: geometric features including the shape, edges and curvature of the power materials, texture features including the surface material of the power materials, and semantic features including the nameplate text and model identification. When the camera inside the requisition cabinet acquires the original images of the power materials, the image resolution is at least 1920×1080 pixels or higher, and the frame rate is not less than 15fps.

[0034] Furthermore, in S1, electrical materials include, but are not limited to: composite insulators, cable terminations, wiring fittings, and surge arrester assemblies. When calculating the characteristic coupling factor, the surface features of the electrical materials include at least two of the following: anti-slip threads, awning structures, and metal flanges.

[0035] Furthermore, the specific implementation steps for the entire inventory management process in S5 are as follows:

[0036] S501, Inventory Status Update: Adjusts the material inventory data of the smart requisition cabinet in real time based on the matching results;

[0037] S502, Intelligent Early Warning: When the inventory level is lower than the lower limit set by the intelligent requisition cabinet or higher than the upper limit, an audible and visual alarm signal is triggered and pushed to the operation and maintenance management system via the MQTT protocol;

[0038] S503, Threshold Dynamic Adjustment: Automatically optimizes the smart requisition cabinet inventory threshold based on historical requisition data, with an adjustment cycle of 24 hours.

[0039] Compared with existing technologies, the beneficial effects of this invention include: by using high-resolution image acquisition and a multi-directional convolutional framework constructed with spatial gradient energy kernels, it can accurately extract the geometric contours, surface textures, and identification features of power materials; by combining three-dimensional differential operations, it can progressively analyze edge intensity, curvature changes, and deformation information; and by using adaptive illumination compensation technology, it can effectively eliminate environmental interference and enhance feature stability in complex scenarios. By using a multi-feature dynamic coupling model, it can generate unique feature factors by fusing multi-scale texture density and high-order differential features, and by combining an elliptical envelope detection algorithm, it can achieve real-time high-precision matching, ensuring the accuracy of material type and quantity identification during intelligent requisition cabinet power material management. Attached Figure Description

[0040] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts. Wherein:

[0041] Figure 1 The schematic diagram illustrates the overall flowchart of the optimized management method according to one embodiment of the present invention;

[0042] Figure 2 The schematic diagram illustrates the feature extraction implementation steps of the optimized management method according to an embodiment of the present invention;

[0043] Figure 3 The diagram illustrates the entire process of inventory management implementation steps of the optimized management method proposed according to one embodiment of the present invention. Detailed Implementation

[0044] It is readily understood that, based on the technical solution of this invention, those skilled in the art can propose various interchangeable structural methods and implementations without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention.

[0045] According to one embodiment of the present invention, Figures 1-3 The following is illustrated: A method for image post-processing optimization management of electrical materials in an intelligent requisition cabinet, comprising the following implementation steps:

[0046] Step 1: User Operations and Permission Verification

[0047] Users select the "receive" or "store" function through the smart locker's touch interface. The system then uses facial recognition to verify user permissions. Once verified, the locker door unlocks automatically and records the user's information.

[0048] Step Two: Image Acquisition and Data Preprocessing. After the user completes the retrieval and placement of materials and closes the cabinet door, the high-definition camera installed inside the cabinet monitors the environment and electrical materials inside.

[0049] Multi-angle image acquisition. The acquired raw image data is processed through a multi-directional convolutional framework that fuses spatial gradient energy kernels to enhance the recognizability of image features. The image resolution is no less than 1920×1080 pixels, and the frame rate is ≥15fps. These are raw images of electrical materials captured by the smart requisition cabinet camera. They contain the following key information:

[0050] Geometric features: the shape, edges, curvature, etc. of a material (such as the skirt structure of an insulator).

[0051] Texture characteristics: Surface material (such as the brushed texture of metal flanges, the grainy texture of rubber parts).

[0052] Semantic features: Key information such as nameplate text and model identification.

[0053] Step 3: Multi-dimensional Feature Extraction and Coupling

[0054] Geometric feature extraction: Three-dimensional differential operation is performed on the original image along the X-axis. The 1st to 3rd order partial derivatives are calculated using a 5×5 Sobel convolution kernel to extract the edge intensity (1st order), curvature change (2nd order), and higher-order deformation features (3rd order) of the material.

[0055] Texture and lighting compensation: Calculating multi-scale differential texture density based on 12 directional kernels covering 360 degrees of space. ), and by the mean of local image patches ( ) and variance ( Generate adaptive illumination compensation factor ( To eliminate lighting interference, the mean value of local image blocks in power material images is calculated as follows:

[0056] ,

[0057] The local image patch variance of power material images is calculated as follows:

[0058]

[0059] in, This represents a local image block centered at pixel k; N is the total number of pixels within the local block.

[0060] Feature fusion: The result of the differential operation is nonlinearly coupled with the texture density to generate a feature numerator term, which is then combined with the cubic gradient integral of the second-order mixed partial derivative of the image (the noise suppression denominator term) to finally output the feature coupling factor (Γ(I)). The formula for calculating the feature coupling factor in the multi-feature dynamic coupling model is as follows:

[0061]

[0062] In the formula: I represents the input image of power materials, which is a two-dimensional matrix, where each element I(i,j) corresponds to the gray value of a pixel, i represents the row coordinate, and j represents the column coordinate; k is the loop index used to traverse n feature dimensions, and the value of k is from 1 to n; m represents the order of the derivative used for each feature dimension k, and the value of m is from 1 to 3. This represents the convolution output of the gradient energy kernel in the k-th direction; This represents multi-scale differential texture density; (W, H) represents the adaptive illumination compensation factor; (W, H) represents the image size. Represents the image gradient energy kernel Perform partial derivatives from the first to the third order along the x-direction; This represents the calculation of the second-order mixed partial derivatives of the image in the x and y directions.

[0063] By mean ( ) and variance ( The dynamic ratio of ) enables the following functions:

[0064] Light compensation factor design principle

[0065] Lighting balance: Suppresses overexposure in highlight areas and enhances detail in shadow areas.

[0066] Noise suppression: in low contrast regions ( Small) Increase the effective signal weight in high-noise regions ( (Greatly) Reduce interference impact.

[0067] Mechanism of action of light compensation factor

[0068] Highlight area ( big, Small), The value increases significantly → through log(1+Ψ) k Compress the dynamic range to prevent loss of detail in reflective metallic areas.

[0069] shaded area ( Small, big), Decreasing the value preserves the texture of dark areas (such as nameplate text) while suppressing noise amplification.

[0070] Complex texture areas ( medium, big), Moderate value → Balances texture enhancement and noise suppression (such as anti-slip texture on cable terminations).

[0071] Step 4: Feature Matching and Material Identification

[0072] Feature fingerprint database construction: The feature coupling factors of historical power material images are stored in the database, denoted as Λ={Γ(I1),Γ(I2),...,Γ(Im)}.

[0073] Real-time matching: The feature coupling factor Γ(Ix) of the image of the material to be identified is matched using the elliptical envelope detection method, satisfying the condition. Where ε is the dynamic tolerance threshold, which is a constant. When the feature coupling factor Γ(Ix) of the image of the material to be identified is different from that of Γ(Ix) in the feature fingerprint database, the dynamic tolerance threshold is determined. k )satisfy Once a match is found, the type and quantity of supplies are determined.

[0074] Step 5: Inventory Dynamic Management and Early Warning

[0075] Inventory Update: Adjusts the inventory data of materials in the cabinet in real time based on the matching results.

[0076] Threshold warning: When the inventory level is lower than the preset lower limit or exceeds the upper limit, an audible and visual alarm is triggered, and a warning message is pushed to the operation and maintenance system via the MQTT protocol.

[0077] Adaptive optimization: Based on historical requisition data within 24 hours, it automatically optimizes inventory thresholds to improve management efficiency.

[0078] In this embodiment, when retrieving and placing electrical materials from the intelligent requisition cabinet based on user authorization verification, multi-angle image acquisition is performed using a camera inside the cabinet. A high-resolution camera captures the geometric contours, surface textures, and identification information of the electrical materials, and a multi-directional convolutional framework is constructed using spatial gradient energy kernels to enhance feature extraction capabilities. In the core processing stage, three-dimensional differential operations are used to progressively analyze edge intensity, curvature changes, and deformation features. Multi-scale texture analysis and adaptive illumination compensation techniques are combined to eliminate environmental interference. Finally, feature factors are generated through nonlinear coupling, forming the "feature fingerprint" of the materials. The matching process utilizes a dynamic elliptical envelope algorithm to dynamically compare real-time feature factors with a pre-stored fingerprint database. By determining the threshold of the ratio and derivative deviation, the type and quantity of materials are accurately identified. The inventory management module updates data in real time based on the identification results and dynamically adjusts the inventory threshold. Through multi-dimensional feature fusion and dynamic optimization mechanisms, the identification accuracy and robustness under complex lighting and deformation conditions are significantly improved, achieving real-time monitoring and automated management of electrical material inventory.

[0079] The technical scope of this invention is not limited to the content described above. Those skilled in the art can make various modifications and variations to the above embodiments without departing from the technical concept of this invention, and all such modifications and variations should fall within the protection scope of this invention.

Claims

1. An image post-processing optimization management method for intelligent use of a cabinet power material, characterized in that, Includes the following steps: S1: Retrieval and Placement Registration. Users click the function button on the touch screen of the smart requisition cabinet, select the function of retrieving or storing power materials, and unlock the door after facial recognition to determine authorization, thus completing the operator registration. S2: Coverage image acquisition. After the user takes out or puts in the power supplies, he or she closes the door lock. During this process, the camera inside the cabinet takes all-round images of the items taken out and the power supplies inside the cabinet, completes the acquisition of raw image data, and constructs a multi-directional convolutional framework containing spatial gradient energy kernel. S3: Feature extraction. Construct a multi-feature dynamic coupling model for power materials to extract features from the original image. Use the nonlinear multi-feature dynamic coupling model to process the original image and calculate the feature coupling factor of the power materials based on the original image. S4: Identification and matching. A feature fingerprint database of power materials is established using feature coupling factors. The calculated feature coupling factors are matched with the feature fingerprint database in real time using the elliptical envelope detection method to determine and identify the type and quantity of power materials to be picked up and placed. S5: Inventory management process, updates the inventory status of materials in the smart requisition cabinet based on matching results and triggers early warning signals; In S3, the steps for calculating the feature coupling factor in the multi-feature dynamic coupling model are as follows: Step 1: Perform three-dimensional differential operations on the original image data, and calculate the 1st to 3rd order partial derivatives along the X-axis to extract the geometric features of the power materials; Step 2: Parallel calculation of multi-scale differential texture density of power materials, using 12 directional kernels to cover 360-degree space, with each directional kernel calculating an adaptive illumination compensation factor based on the mean and variance of local image patches; Step 3: Nonlinearly couple the differential result from Step 1 with the output from Step 2 to generate characteristic numerator terms; Step 4: Calculate the cubic gradient integral of the second-order mixed partial derivative of the image as the denominator of the noise suppression term, and divide the numerator by the denominator to generate the characteristic coupling factor of the power material. The formula for calculating the feature coupling factor in the multi-feature dynamic coupling model is as follows: , In the formula: I represents the input image of power materials, which is a two-dimensional matrix, where each element I(i,j) corresponds to the gray value of a pixel, i represents the row coordinate, and j represents the column coordinate; k is the loop index used to traverse n feature dimensions, and the value of k is from 1 to n; m represents the order of the derivative used for each feature dimension k, and the value of m is from 1 to 3. This represents the convolution output of the gradient energy kernel in the k-th direction; This represents multi-scale differential texture density; (W, H) represents the adaptive illumination compensation factor; (W, H) represents the image size. Represents the image gradient energy kernel Perform partial derivatives from the first to the third order along the x-direction; This indicates the calculation of the second-order mixed partial derivatives of the image in the x and y directions; In S4, the elliptical envelope detection method verifies the characteristic coupling factor of power materials as follows: Let the power material characteristic fingerprint library Λ = {Γ(I1), Γ(I2), …, Γ(I m )}; Real-time calculation of the image Γ(Ix) of the material to be identified; Matching is performed using the elliptic envelope detection method: Where ε is the dynamic tolerance threshold, which is a constant. When the feature coupling factor Γ(Ix) of the image of the material to be identified is different from that of Γ(Ix) in the feature fingerprint database, the dynamic tolerance threshold is determined. k )satisfy The match is considered successful upon confirmation.

2. The image post-processing optimization management method of electric power materials in the intelligent taking cabinet according to claim 1, characterized in that, The formula for calculating the illumination compensation factor in step two is as follows: ， wherein, denotes the local image block mean of the image of the electric power good, denotes the local image block variance of the image of the electric power good.

3. The image post-processing optimization management method of electric power materials in the intelligent taking cabinet according to claim 2, characterized in that, The calculation method for the mean value of local image blocks in the power material image is as follows: ， The local image block variance of the power material image is calculated as follows: , wherein, represents a local image block centered at pixel point k; N is the total number of pixels within the local block.

4. The image post-processing optimization management method of smart issue cabinet electric power materials according to claim 1, characterized in that, In step one, when performing three-dimensional differential operations on the original image: 1st, 2nd, and 3rd order partial derivatives are calculated using a 5×5 Sobel convolution kernel along the x-axis. The 1st order partial derivative is used to extract the edge intensity of the power materials, the 2nd order partial derivative is used to extract the curvature change of the power materials, and the 3rd order partial derivative is used to extract the higher-order deformation features of the power materials.

5. The image post-processing optimization management method of smart issue cabinet electric power materials according to claim 1, characterized in that, In S2, the key information of the original images of the power materials collected by the camera inside the requisition cabinet includes: geometric features including the shape, edges and curvature of the power materials, texture features including the surface material of the power materials, and semantic features including the nameplate text and model identification. When the camera inside the requisition cabinet acquires the original images of the power materials, the image resolution is at least 1920×1080 pixels or higher, and the frame rate is not less than 15fps.

6. The method of claim 1, wherein the method further comprises: In S1, the power materials include, but are not limited to: composite insulators, cable terminations, wiring hardware, and surge arrester assemblies. When calculating the characteristic coupling factor, the surface features of the power materials include at least two of the following: anti-slip threads, umbrella skirt structure, and metal flange.

7. The method of claim 1, wherein the method further comprises: The specific implementation steps of the entire inventory management process in S5 are as follows: S501, Inventory Status Update: Adjusts the material inventory data of the smart requisition cabinet in real time based on the matching results; S502, Intelligent Early Warning: When the inventory level is lower than the lower limit set by the intelligent requisition cabinet or higher than the upper limit, an audible and visual alarm signal is triggered and pushed to the operation and maintenance management system via the MQTT protocol; S503, Threshold Dynamic Adjustment: Automatically optimizes the smart requisition cabinet inventory threshold based on historical requisition data, with an adjustment cycle of 24 hours.

Images