CN-121998927-A - Illite and pine needle oil load uniformity detection method based on image recognition

CN121998927ACN 121998927 ACN121998927 ACN 121998927ACN-121998927-A

Abstract

The invention relates to the field of image detection, in particular to an illite and pine needle oil load uniformity detection method based on image recognition, which comprises the steps of acquiring a composite powder microscopic image under a lateral illumination condition to obtain an effective detection gray image and a seed point set; the method comprises the steps of carrying out joint evaluation on local gradient and curvature response of an effective detection gray image to obtain an artifact suppression factor, carrying out joint evaluation on regional growth direction consistency and local flux characteristics to obtain a flux compensation factor, carrying out fusion modulation on gray difference judging cost through the artifact suppression factor and the flux compensation factor to obtain final growth judging cost and generate a load region binary mask, carrying out gridding load rate statistics and dispersion calculation on the load region binary mask to obtain a load uniformity index, and thus solving the problems of shadow false growth leakage and load uniformity quantification misalignment caused by early stop and leakage detection on oil film weak edges.

Inventors

MENG WAN
Pi Longquan
MENG JINGBI

Assignees

延边大学

Dates

Publication Date: 20260508
Application Date: 20260120

Claims (10)

1. An illite and pine needle oil load uniformity detection method based on image recognition is characterized by comprising the following steps: Step S1, acquiring a composite powder microscopic image under a lateral illumination condition to obtain an effective detection gray level image and a seed point set; S2, obtaining an artifact suppression factor by carrying out joint evaluation on local gradients and curvature responses of the effective detection gray level image; S3, obtaining flux compensation factors by carrying out joint evaluation on the consistency of the growth directions of the areas and the local flux characteristics; S4, fusion modulation is carried out on the gray level difference judgment cost through the artifact suppression factor and the flux compensation factor, the final growth judgment cost is obtained, and a binary mask of a load area is generated; and S5, carrying out gridding load rate statistics and dispersion calculation on the binary mask of the load area to obtain a load uniformity index.
2. The method for detecting the load uniformity of illite and pine needle oil based on image recognition according to claim 1, wherein the steps of acquiring the effective detection gray level image and the seed point set by acquiring the composite powder microscopic image under the lateral illumination condition comprise the following steps: an industrial microscopic camera and a lateral light source are arranged on a microscopic imaging platform, the incident angle of the lateral light source relative to the surface of a composite powder sample is fixedly set to be 45 degrees, microscopic imaging acquisition is carried out on the spray-dried illite and pine needle oil composite powder sample to obtain an original color image of the composite powder microscopic image, gray processing is carried out on the original color image to correspondingly obtain an effective detection gray image, gray minimum pixel points are searched in the effective detection gray image, a preset number of pixel points with the lowest gray are used as seed points, and a set of all the seed points is used as a seed point set.
3. The method for detecting the load uniformity of illite and pine needle oil based on image recognition according to claim 1, wherein the obtaining the artifact suppression factor by performing joint evaluation on the local gradient and the curvature response of the effective detected gray image comprises: extracting and normalizing local gradient characteristic data and Hessian curvature response characteristic data of an effective detection gray image to obtain pseudo-influence evaluation data; And obtaining the artifact suppression factor by carrying out punishment enhancement modulation and lower limit constraint processing on the artifact influence to-be-evaluated data.
4. The method for detecting the load uniformity of illite and pine needle oil based on image recognition according to claim 3, wherein the obtaining the pseudo-influence-should-be-evaluated data by extracting and normalizing local gradient characteristic data and Hessian curvature response characteristic data of an effective detection gray image comprises: For any target pixel point in the effective detection gray level image, based on the gray level difference calculation of adjacent pixels of the target pixel point in the horizontal direction and the vertical direction, obtaining a local gradient feature vector corresponding to the target pixel point, and taking the module length of the local gradient feature vector as the gradient intensity estimation corresponding to the target pixel point; Acquiring Hessian curvature response characteristic data corresponding to a target pixel point through second-order difference and cross-difference calculation of the target pixel point in the horizontal and vertical directions, sequentially carrying out matrix multiplication operation on a transpose of a local gradient characteristic data matrix, the Hessian curvature response characteristic data matrix and the local gradient characteristic data matrix to acquire curvature projection response assessment corresponding to the target pixel point; And taking the calculated result of multiplying the curvature response normalization evaluation and the incidence angle compensation coefficient of the side illumination as the pseudo influence on the target pixel point to evaluate data.
5. The method for detecting the load uniformity of illite and pine needle oil based on image recognition according to claim 3, wherein the obtaining the artifact suppression factor by performing punishment enhancement modulation and lower limit constraint processing on the artifact-to-be-evaluated data comprises the following steps: Setting a punishment enhancement modulation adjustment coefficient, extracting pseudo influence to be evaluated data corresponding to a target pixel point for effectively detecting any target pixel point in a gray level image, taking a calculation result obtained by multiplying the punishment enhancement modulation adjustment coefficient and the pseudo influence to be evaluated data as punishment enhancement modulation amount evaluation, taking a constant 1 as a lower limit constraint standard, and taking a calculation result obtained by adding the constant 1 and the punishment enhancement modulation amount evaluation as a pseudo influence suppression factor corresponding to the target pixel point.
6. The method for detecting the load uniformity of illite and pine needle oil based on image recognition according to claim 1, wherein the obtaining the flux compensation factor by jointly evaluating the consistency of the growth direction of the region and the local flux characteristics comprises the following steps: the method comprises the steps of carrying out direction consistency analysis processing on regional growth direction vector data and local gradient direction data of an effective detection gray level image to obtain growth direction consistency evaluation data; Local flux characteristic evaluation data are obtained by carrying out divergence characteristic and rotation characteristic extraction processing on local gradient field data of an effective detection image; And obtaining the flux compensation factor by carrying out gate fusion and upper limit constraint processing on the growth direction consistency evaluation data and the local flux characteristic evaluation data.
7. The method for detecting the uniformity of illite and pine needle oil loads based on image recognition according to claim 6, wherein the obtaining the growth direction uniformity evaluation data by performing a direction uniformity analysis process on the regional growth direction vector data and the local gradient direction data of the effective detection gray level image comprises: For any target pixel point in the effective detection gray level image, determining a target seed point corresponding to the target pixel point from a seed point set, and acquiring a two-dimensional pixel coordinate of the target pixel point and a two-dimensional pixel coordinate of the target seed point; Taking the calculation result of the difference between the two-dimensional pixel coordinates of the target pixel point and the two-dimensional pixel coordinates of the target seed point as region growth direction vector data corresponding to the target pixel point, taking the modular length of the region growth direction vector data as a normalization scale, and dividing the calculation result of the region growth direction vector data by the normalization scale as unit region growth direction vector data corresponding to the target pixel point; Carrying out vector inner product operation on unit area growth direction vector data corresponding to a target pixel point and a local gradient vector to obtain direction consistency inner product evaluation corresponding to the target pixel point, multiplying the modulo length of the area growth direction vector data corresponding to the target pixel point by the modulo length of the local gradient vector and adding a calculation result of preventing from being divided by 0 to be used as a target pixel point direction consistency normalization scale, dividing the calculation result of the direction consistency inner product evaluation corresponding to the target pixel point by the direction consistency normalization scale to be used as a target pixel point direction consistency normalization evaluation, mapping the direction consistency normalization evaluation corresponding to the target pixel point by a hyperbolic tangent function, and taking the mapping result obtained correspondingly as a target pixel point corresponding growth direction consistency evaluation.
8. The method for detecting the load uniformity of illite and pine needle oil based on image recognition according to claim 6, wherein the obtaining the local flux characteristic evaluation data by performing a divergence characteristic and a rotation characteristic extraction process on the local gradient field data of the effective detected image comprises: setting a local range of a pixel point, and for any target pixel point in an effective detection image, acquiring a local gradient field of the target pixel point through a gradient vector of the pixel point in the local range of the target pixel point, wherein the local gradient field of the target pixel point at least comprises a horizontal gradient component and a vertical gradient component of the pixel point; Performing first-order difference calculation on the horizontal gradient component in the horizontal direction, performing first-order difference calculation on the vertical gradient component in the vertical direction, and taking the added calculation result as a corresponding divergence characteristic evaluation of the target pixel point; Performing first-order difference calculation on the vertical gradient component in the horizontal direction, performing first-order difference calculation on the horizontal gradient component in the vertical direction, and taking the absolute value of the difference calculation result of the first-order difference calculation and the second-order difference calculation as the rotation characteristic evaluation corresponding to the target pixel point; and dividing the rotation characteristic evaluation corresponding to the target pixel point by the calculation result of the rotation characteristic evaluation to obtain local flux characteristic evaluation data corresponding to the target pixel point.
9. The method for detecting the load uniformity of illite and pine needle oil based on image recognition according to claim 6, wherein the obtaining the flux compensation factor by performing gating fusion and upper limit constraint processing on growth direction consistency evaluation data and local flux characteristic evaluation data comprises the following steps: Setting a compensation intensity coefficient, regarding any target pixel point in an effective detection image, taking a constant 1 as a gating reference, taking a calculation result of subtracting local flux characteristic evaluation data from the constant 1 as a flux gating coefficient corresponding to the target pixel point, multiplying the compensation intensity coefficient, the growth direction consistency evaluation and the flux gating coefficient, taking the opposite number of the calculation result as an exponential decay driving amount, and mapping the exponential decay driving amount by an exponential function taking a natural constant as a base to obtain a flux compensation factor corresponding to the target pixel point.
10. The method for detecting the load uniformity of illite and pine needle oil based on image recognition according to claim 1, wherein the steps of performing fusion modulation on gray level difference judgment cost through artifact suppression factors and flux compensation factors, obtaining final growth judgment cost and generating a load region binary mask comprise the following steps: For any target pixel point in the effective detection image, acquiring the gray value of the target pixel point and the gray value of a target seed point corresponding to the target pixel point, and taking the absolute value of the difference value between the gray value of the target pixel point and the gray value of the target seed point as the gray difference judgment cost corresponding to the target pixel point; Taking the gray level difference judgment cost corresponding to the target pixel point and the calculation result of multiplying the artifact suppression factor by the flux compensation factor as the final growth judgment cost corresponding to the target pixel point; When the final growth judgment cost is smaller than the threshold value of the gray level difference judgment cost, marking the target pixel point as a load area pixel point and updating the load area binary mask, when the final growth judgment cost is larger than or equal to the preset gray level difference judgment cost threshold value, marking the target pixel point as an empty area pixel point and keeping the load area binary mask unchanged, and traversing all the target pixel points in the effective detection gray level image to obtain the load area binary mask.

Description

Illite and pine needle oil load uniformity detection method based on image recognition Technical Field The invention relates to the technical field of image detection, in particular to an illite and pine needle oil load uniformity detection method based on image recognition. Background In the field of fine processing and functional composite modification of inorganic nonmetallic materials, volatile active ingredients such as pine needle oil and the like are loaded on the surface of submicron illite powder, and the functions of bacteriostasis, slow release or odor regulation and the like can be endowed to the materials while the stability of a mineral matrix is maintained, so that the preparation method is an important process route for preparing functional composite powder. The compound solidification of active ingredients and illite powder is realized in an industrial manner usually by adopting a spray drying mode, the compound powder formed after spray drying usually exists in a spherical or quasi-spherical aggregate form, and the product performance and the batch stability are highly dependent on the coverage degree, the infiltration distribution and the spatial uniformity of pine needle oil on the surface of the aggregate, so that the coverage rate and the distribution uniformity of a pine needle oil load area are required to be detected on line or quasi-on line on a production line so as to support closed-loop adjustment of process parameters and avoid the reduction of functional attenuation and batch consistency caused by uneven load. In the existing online detection means, an industrial microscopic camera is adopted to obtain a micro-distance image of the composite powder after spray drying under a fixed illumination condition, then a load area formed by wetting pine needle oil is distinguished from an empty area formed by a drying substrate through an image segmentation algorithm, further load coverage rate is counted based on a binary mask, and uniformity is evaluated through gridding statistics or space dispersion indexes. In the common segmentation algorithm, the gradient-based seed region growth algorithm is widely used for extracting the wettable coverage area because the gradient-based seed region growth algorithm can simulate the continuous permeation and diffusion behavior of oil on the solid surface in the growth process of 'from dark to bright and from inside to outside', and has the characteristics of strong constraint on region connectivity, simplicity in implementation, easiness in engineering deployment and the like. The algorithm generally takes an image gray minimum value pixel as a seed point, takes gray difference between a neighborhood pixel and the seed point or a current growing region as a judging cost, judges whether the neighborhood pixel is integrated into the growing region through a preset threshold value, considers the pixel to belong to a load region and continues to expand outwards when the gray difference is smaller than the threshold value, and stops expanding when the gray difference is larger than the threshold value, so that a segmentation result of the load region is obtained. However, for the sub-micron illite and pine needle oil composite powder sample after spray drying, the conventional region growth segmentation based on single gray level difference measurement is easy to generate identification deviation under the actual macro imaging condition. The reason is that submicron illite powder is easy to agglomerate, a large number of microcosmic stacking gaps and structural fluctuation exist on the surface of the agglomerate after spray drying, the microstructures can generate high-frequency geometrical self-shielding shadows under the condition of lateral illumination or oblique illumination, the gray value of a shadow area is low and is highly overlapped with the gray characteristic of a pine needle oil infiltration area, a dry gap shadow is easily incorporated into a load area by mistake when only the gray difference is taken as a criterion by an area growth algorithm, growth leakage occurs, a large number of pseudo-load pixels are introduced, and therefore the coverage rate and uniformity statistics generate false lifting and false fluctuation. Meanwhile, pine needle oil has volatility and certain translucency, and is often subjected to nonlinear gradient attenuation caused by gradual oil film thickness change and gradual optical path shortening at the infiltration edge, the gray level of an edge area is gradually reduced, the contrast with a substrate is reduced, and an area growth algorithm is easy to stop in advance due to the increase of gray level difference when an oil film is not completely extracted under the constraint of a fixed threshold value, so that weak edge transition zone detection and area breakage are caused, and further underestimation of a load area and uniformity evaluation distortion are caused. The two errors often coexist and are mutu