An intelligent screening method and system for traditional Chinese medicine decoction pieces integrated with image recognition technology

Abstract

The application relates to the technical field of image recognition, and discloses a traditional Chinese medicine decoction piece intelligent screening method and system integrated with image recognition technology. The method comprises the following steps: extracting surface image texture features of traditional Chinese medicine decoction pieces, distinguishing differences from standard surface texture features, and obtaining surface texture difference areas; sequentially imaging the surface texture difference areas under illumination to obtain multi-angle images, reconstructing three-dimensional morphologies of the multi-angle images, and obtaining three-dimensional morphological data; analyzing morphological parameters of the three-dimensional morphological data, and obtaining surface attached material areas of the traditional Chinese medicine decoction pieces; analyzing deformation evolution of the surface attached material areas, obtaining a time sequence deformation characteristic spectrum of the surface attached material areas; clustering deformation areas of the time sequence deformation characteristic spectrum, obtaining moldy precursor areas of the traditional Chinese medicine decoction pieces; and judging moldy risks of the moldy precursor areas, and obtaining intelligent screening results of the traditional Chinese medicine decoction pieces.

Description

Technical Field

[0001] This invention relates to the field of image recognition technology, and in particular to a method and system for intelligent screening of traditional Chinese medicine decoction pieces that incorporates image recognition technology. Background Technology

[0002] As the core form of TCM clinical medication, the integrity and cleanliness of the surface properties of prepared herbal pieces are key indicators for measuring their quality grade. In the traditional TCM prepared herbal piece production and quality control process, the detection of the surface quality of the prepared herbal pieces mainly relies on human sensory evaluation, which has problems such as strong subjectivity, low efficiency, and difficulty in standardization. With the development of machine vision and image processing technology, image-based automated detection has gradually become a research hotspot in this field. By collecting images of the surface of prepared herbal pieces and using image processing algorithms to quantify and analyze their appearance information such as texture features and color distribution, a new technical path is provided for the objective evaluation of the quality of prepared herbal pieces. Existing technologies mostly focus on obtaining the overall appearance of prepared herbal pieces through single-modal imaging and combining general image segmentation algorithms to identify obvious surface defects.

[0003] Existing detection technologies rely solely on conventional two-dimensional texture comparison, making it difficult to distinguish between the inherent texture differences of medicinal slices and the micro-texture distortions of early mold precursors. For example, some medicinal slices with naturally occurring fine wrinkles or significant variations in their texture often misjudge normal texture fluctuations as abnormal areas, leading to a large number of false positive screening results. Furthermore, the lack of a precise frequency domain texture purification and spatial coding difference mapping mechanism for multispectral fusion images makes it difficult to stably locate texture distortion areas under complex background interference. Conventional three-dimensional morphology reconstruction can only acquire static surface structures. For the minute morphological changes that appear in the early stages of mold precursors, such as the slight bulging or edge expansion of surface attachments within a few hours, effective temporal deformation vector quantization and rate-increasing clustering analysis are not possible. This hinders early prediction and graded screening of mold risk and fails to meet the high-precision intelligent sorting requirements of medicinal slices. Summary of the Invention

[0004] This invention provides a method and system for intelligent screening of traditional Chinese medicine decoction pieces that incorporates image recognition technology, in order to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides an intelligent screening method for traditional Chinese medicine decoction pieces that incorporates image recognition technology, comprising: S1, extracting texture features from the surface image of traditional Chinese medicine decoction pieces, and comparing the extraction results with the standard surface texture features of traditional Chinese medicine decoction pieces to obtain the surface texture difference region of traditional Chinese medicine decoction pieces; S2. Perform sequential illumination imaging on the surface texture difference region to obtain multi-angle images of the surface texture difference region, and construct three-dimensional morphology data of the surface texture difference region based on the multi-angle images. S3. Perform morphological parameter analysis on the three-dimensional morphology data to obtain the surface attachment area of ​​the Chinese herbal medicine slices; S4. Perform dynamic displacement field tracking on the surface attachment area, and analyze the deformation displacement vector field of the surface attachment area based on the displacement tracking results. Construct a temporal deformation feature map of the surface attachment area based on the deformation displacement vector field. S5. Deconstruct the temporal deformation feature map in space-time, identify the abnormal deformation region based on the deconstruction result, and locate the mold precursor region of the Chinese herbal medicine slices based on the persistence and expansion characteristics of the abnormal deformation region. S6. Assess the risk of mold growth in the precursor areas to obtain intelligent screening results for Chinese herbal medicine pieces.

[0006] In a preferred embodiment, the process of obtaining the surface texture difference region of the Chinese herbal medicine slices is as follows: Multispectral imaging of Chinese medicinal herbs was performed to obtain the original surface images of the herbs. Based on preset feature bands, the original surface image is separated and reconstructed in the frequency domain to obtain a texture highlighting image of Chinese herbal medicine slices; Neighborhood relation quantization is performed on the texture-highlighting image to obtain the local texture coding value of the texture-highlighting image; Obtain the standard texture coding value of Chinese herbal medicine slices, and perform a deviation space mapping between the local texture coding value and the standard texture coding value to obtain the texture difference localization map of Chinese herbal medicine slices; Connectivity analysis was performed on the texture difference localization map to obtain the surface texture difference regions of Chinese herbal medicine slices.

[0007] In a preferred embodiment, the process of obtaining multi-angle images of surface texture difference regions is as follows: Discretize the surface texture difference region to obtain the imaging block of the surface texture difference region; According to the preset illumination angle sequence, the imaging block is illuminated sequentially from multiple angles to obtain the reflected light field distribution of the imaging block; By spatially stitching together the reflected light field distribution of the imaging blocks under the same illumination angle, multi-angle images of areas with different surface textures are obtained.

[0008] In a preferred embodiment, the process of constructing three-dimensional topographic data of surface texture difference regions based on multi-angle images is as follows: Height field reconstruction is performed on multi-angle images to obtain depth information fields of surface texture difference regions; Based on the pixel mapping relationship between the depth information field and the multi-angle image, texture mapping is performed on the multi-angle image to obtain the texture attachment depth map of the surface texture difference region; The texture attachment depth map is reconstructed into a point cloud according to spatial coordinates to obtain the three-dimensional morphological data of the surface texture difference region.

[0009] In a preferred embodiment, the process of obtaining the surface attachment area of ​​the Chinese herbal medicine slices is as follows: Feature extraction is performed on the three-dimensional topography data to obtain the topography parameter set of the surface texture difference region; The topographic parameter set is compared with the elevation to generate an elevation difference map of the surface texture difference area; Perform region growing on the elevation difference map to obtain candidate regions for attachments in areas with surface texture differences; Morphological identification is performed on the candidate areas of the attachments to obtain the surface attachment areas of Chinese herbal medicine slices.

[0010] In a preferred embodiment, the process of obtaining the temporal deformation feature map of the surface attachment region is as follows: Multi-temporal sampling imaging of the surface attachment area yields a temporal image sequence of the surface attachment area; Based on the temporal image sequence, edge contour tracking is performed on the surface attachment region to obtain the edge contour information of the surface attachment region; Based on the edge contour information, deformation vector mapping is performed on the surface attachment region to obtain the deformation displacement vector field of the surface attachment region; By arranging the spatiotemporal distribution of the deformation displacement vector field, a temporal deformation characteristic map of the surface attachment region is obtained.

[0011] In a preferred embodiment, the process of obtaining the mold precursor region of the traditional Chinese medicine decoction piece is as follows: The deformation rate hierarchy of the time-series deformation feature map is obtained by dividing the map into rate intervals. Based on the hierarchical distribution of deformation rate, the increasing trend of the time-series deformation feature map is identified to obtain the distribution of the rate increasing region of the time-series deformation feature map; Spatial proximity aggregation is performed on the distribution of rate-increasing regions to obtain the rate-increasing connected domain distribution of the temporal deformation feature map; By performing continuous duration authentication on the distribution of rate-increasing connected regions, the mold precursor region of traditional Chinese medicine decoction pieces is obtained.

[0012] In a preferred embodiment, the process of obtaining the mold precursor region of the traditional Chinese medicine decoction piece is as follows: Perform duration-continuous frame tracking on the rate-increasing connected component distribution to obtain a set of continuous frame sequences of the rate-increasing connected component distribution; A time-varying trend analysis of the rate was performed on a continuous frame sequence set to obtain a rate growth feature set of the rate-increasing connected component distribution; By identifying the regional expansion pattern of the rate growth feature set, the mold precursor region of Chinese herbal medicine slices is obtained.

[0013] In a preferred embodiment, the process of obtaining the mold risk level and intelligent screening results of Chinese herbal medicine slices is as follows: Multidimensional feature parameter extraction was performed on the mold precursor region to obtain the area feature parameter, morphological feature parameter, and deformation rate feature parameter of the mold precursor region; Based on area characteristic parameters, morphological characteristic parameters, and deformation rate characteristic parameters, risk factors are weighted and fused in the mold precursor region to obtain the comprehensive mold risk index of the mold precursor region. Based on the comprehensive index of mold risk, the risk level of the mold precursor area is mapped to obtain the mold risk level of the mold precursor area. By making decisions based on the risk level of mold growth, intelligent screening results for Chinese herbal medicine slices are generated.

[0014] To address the aforementioned problems, the present invention also provides an intelligent screening system for traditional Chinese medicine decoction pieces incorporating image recognition technology, the system comprising: The surface texture difference identification module is used to extract texture features from the surface image of Chinese herbal medicine slices and to identify the differences between the extraction results and the standard surface texture features of Chinese herbal medicine slices to obtain the surface texture difference areas of Chinese herbal medicine slices. The 3D topography reconstruction module is used to perform sequential illumination imaging on areas with different surface textures, obtain multi-angle images of these areas, and construct 3D topography data of these areas based on the multi-angle images. The surface attachment analysis module is used to perform morphological parameter analysis on three-dimensional morphology data to obtain the surface attachment area of ​​Chinese herbal medicine slices; The temporal deformation feature analysis module performs dynamic displacement field tracking on the surface attachment region, and analyzes the deformation displacement vector field of the surface attachment region based on the displacement tracking results, and constructs the temporal deformation feature map of the surface attachment region based on the deformation displacement vector field. The mold precursor region clustering module performs spatiotemporal deconstruction of the temporal deformation feature map, identifies abnormal deformation regions based on the deconstruction results, and locates the mold precursor region of Chinese herbal medicine slices based on the persistence and expansion characteristics of the abnormal deformation regions. The mold risk assessment and screening module is used to assess the mold risk of the precursor area and obtain intelligent screening results for Chinese herbal medicine pieces.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. By capturing images of the surface of medicinal slices using multispectral frequency bands, the effective texture signal is purified through frequency domain separation and reconstruction, and irrelevant noise is filtered out. Then, by comparing the deviation of the neighboring texture quantization encoding with the standard texture, combined with connected component analysis, abnormal areas are located. This effectively avoids the problems of stray light interference and normal texture fluctuations being misjudged as defects in conventional detection. It can accurately locate abnormal areas of surface texture, laying a solid foundation for subsequent accurate detection.

[0016] 2. By splitting the texture abnormal area into independent imaging units, collecting the reflected light field according to the fixed angle sequence illumination and completing the stitching imaging, and then relying on height field reconstruction, texture mapping and point cloud reconstruction to generate three-dimensional morphology data, it breaks through the limitation of conventional three-dimensional imaging that cannot distinguish the structure of the medicine slices themselves from the tiny attachments. It can accurately extract the area of ​​tiny attachments and greatly improve the precision of attachment detection.

[0017] 3. Continuous temporal image acquisition of surface attachments is carried out. Temporal deformation maps are generated by edge contour tracking and deformation vector quantization. Based on the deformation rate, growth trend and duration, the early signs of mold are identified. This makes up for the shortcomings of conventional detection in capturing early, weak and short-term mold deformation. It can identify mold risk in advance and provide advance prediction for mold prevention and intelligent sorting of medicinal slices, effectively reducing mold loss. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating an intelligent screening method for traditional Chinese medicine decoction pieces incorporating image recognition technology, provided in an embodiment of the present invention. Figure 2 This is a functional module diagram of an intelligent screening system for traditional Chinese medicine decoction pieces incorporating image recognition technology, provided in an embodiment of the present invention. The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0020] This application provides an intelligent screening method for traditional Chinese medicine (TCM) decoction pieces that incorporates image recognition technology. The executing entity of this intelligent screening method for TCM decoction pieces that incorporates image recognition technology includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the intelligent screening method for TCM decoction pieces that incorporates image recognition technology can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms.

[0021] Reference Figure 1 The diagram shown is a flowchart illustrating an intelligent screening method for traditional Chinese medicine decoction pieces incorporating image recognition technology according to an embodiment of the present invention. In this embodiment, the intelligent screening method for traditional Chinese medicine decoction pieces incorporating image recognition technology includes: S1. Extract texture features from the surface image of Chinese herbal medicine slices, and compare the extraction results with the standard surface texture features of Chinese herbal medicine slices to obtain the surface texture difference areas of Chinese herbal medicine slices. In this embodiment of the invention, the process of obtaining the surface texture difference region of the Chinese herbal medicine slices is as follows: Multispectral imaging of Chinese medicinal herbs was performed to obtain the original surface images of the herbs. Place the Chinese herbal medicine slices stably on a fixed imaging support surface that is free of reflection and impurities. Adjust the lens focal length and shooting distance of the multispectral imaging device to ensure that the Chinese herbal medicine slices are completely within the imaging field of view of the device. Start the multispectral imaging device and output different wavelengths of light to illuminate the surface of the slices in sequence. Each time a beam of light is illuminating the surface, an image is acquired. After all wavelengths have been acquired, all single-wavelength images are precisely aligned and fused according to their spatial positions to obtain the original surface image containing all the basic information of the color and texture of the slices.

[0022] Based on preset feature bands, the original surface image is separated and reconstructed in the frequency domain to obtain a texture highlighting image of Chinese herbal medicine slices; The brightness information of each pixel in the original surface image is converted into a frequency distribution state. The texture-related frequency signals are locked according to the frequency range corresponding to the preset feature band. All signals in the frequency distribution that do not belong to the feature band range are removed. Then, the remaining texture-related frequency signals are converted into a spatial pixel distribution state. The processed image retains only the texture-related information of the slice surface and has no other irrelevant interference, forming a texture-highlighting image.

[0023] Neighborhood relation quantization is performed on the texture-highlighting image to obtain the local texture coding value of the texture-highlighting image; On the texture-enhancing image, take any pixel as the center, select a fixed number of surrounding pixels to form a local analysis range, and read the brightness values ​​of the center pixel and each surrounding pixel in turn. According to the positional relationship between the center pixel and the surrounding pixels, record the distribution of brightness. Integrate the brightness positional relationship of all pixels in this local range into a unique numerical expression to obtain the local texture encoding value.

[0024] Obtain the standard texture coding value of Chinese herbal medicine slices, and perform a deviation space mapping between the local texture coding value and the standard texture coding value to obtain the texture difference localization map of Chinese herbal medicine slices; From the pre-built qualified Chinese herbal medicine decoction piece texture database, the standard texture code value that is consistent with the decoction piece variety being tested is retrieved. The local texture code value and the standard texture code value at each position on the texture highlighting image are compared bit by bit. The number of inconsistent contents is counted. The position where the number of inconsistent contents reaches a fixed threshold is determined as the texture difference position. All texture difference positions are marked on the blank background map according to the image coordinates to form a texture difference positioning map.

[0025] Perform connected component analysis on the texture difference localization map to obtain the surface texture difference regions of Chinese herbal medicine slices; Starting from the top left corner of the texture difference localization map, traverse all pixels row by row and column by column. Group pixels with adjacent coordinates that belong to the texture difference position into the same group. Calculate the spatial range covered by each group of pixels. Remove isolated pixel groups with a spatial range smaller than a fixed standard. Keep continuous pixel groups with a spatial range that meets the fixed standard. The surface range of the Chinese herbal medicine slice corresponding to this continuous pixel group is the surface texture difference region.

[0026] S2. Perform sequential illumination imaging on the surface texture difference region to obtain multi-angle images of the surface texture difference region, and construct three-dimensional morphology data of the surface texture difference region based on the multi-angle images; In this embodiment of the invention, the process of constructing three-dimensional topographic data of surface texture difference regions based on multi-angle images is as follows: Discretize the surface texture difference region to obtain the imaging block of the surface texture difference region; First, the complete boundary outline of the surface texture difference area is locked. Then, along the boundary, the difference area is evenly divided into multiple independent small blocks of fixed spatial size. These small blocks completely cover the entire surface texture difference area. After the division is completed, the imaging block of the surface texture difference area is obtained.

[0027] According to the preset illumination angle sequence, the imaging block is illuminated sequentially from multiple angles to obtain the reflected light field distribution of the imaging block; Multiple different lighting directions are pre-set and a fixed lighting sequence is formed. The light source is used to illuminate each imaging block one by one in this sequence. During the illumination process, all light information reflected from the surface of the block is continuously collected. Then, this light information is organized into a complete light distribution pattern according to the spatial position and lighting direction. This distribution pattern is the reflected light field distribution of the imaging block.

[0028] By spatially stitching together the reflected light field distribution of the imaging blocks under the same illumination angle, multi-angle images of areas with different surface textures are obtained. The reflected light field distributions of all imaging blocks acquired under the same illumination angle are sequentially stitched together according to the original spatial position of the imaging blocks in the surface texture difference region. After stitching, a complete image covering the entire difference region is formed. After performing this stitching operation for each illumination angle, a multi-angle image of the surface texture difference region is obtained.

[0029] Height field reconstruction is performed on multi-angle images to obtain depth information fields of surface texture difference regions; Based on the characteristics of light change at the same spatial location under different lighting angles in multi-angle images, the vertical height value of that location relative to a fixed reference plane is determined. The vertical height values ​​of all locations in the difference area are arranged in an orderly manner according to spatial coordinates to form a complete height distribution set. This set is the depth information field of the surface texture difference area.

[0030] Based on the pixel mapping relationship between the depth information field and the multi-angle image, texture mapping is performed on the multi-angle image to obtain the texture attachment depth map of the surface texture difference region; First, establish a one-to-one correspondence between each spatial coordinate of the depth information field and each pixel of the multi-angle image. Then, attach the texture information of each pixel in the multi-angle image to the same coordinate position in the depth information field according to the correspondence, so that each position in the depth information field carries the corresponding texture information. After integrating the depth and texture information of all positions, a complete image is formed. This is the texture attachment depth map of the surface texture difference region.

[0031] The texture attachment depth map is reconstructed into a point cloud according to spatial coordinates to obtain the three-dimensional morphological data of the surface texture difference region. The depth and texture information corresponding to each spatial coordinate in the texture attachment depth map are extracted. Each set of coordinates, depth and texture information is combined into an independent spatial data point. All data points are arranged in an orderly manner into a dense point set according to the spatial layout of the texture attachment depth map. This point set fully displays the three-dimensional morphology and surface texture features of the difference region, and finally obtains the three-dimensional morphological data of the surface texture difference region.

[0032] S3. Perform morphological parameter analysis on the three-dimensional morphology data to obtain the surface attachment area of ​​the Chinese herbal medicine slices; In this embodiment of the invention, the process of obtaining the surface attachment area of ​​the Chinese herbal medicine slices is as follows: Feature extraction is performed on the three-dimensional topography data to obtain the topography parameter set of the surface texture difference region; Three-dimensional topography data is a set of three-dimensional point clouds containing spatial coordinates, surface height, and texture information, formed by reconstructing the texture attachment depth map into a point cloud. All data points in this three-dimensional point cloud set are traversed, and the spatial coordinate values, surface vertical height values, and surface texture distribution status of each data point are extracted one by one. All extracted information is integrated and summarized according to the spatial arrangement order of the three-dimensional point cloud to form a set of topography parameters that can comprehensively reflect the overall shape and height changes of the surface texture difference area.

[0033] The topographic parameter set is compared with the elevation to generate an elevation difference map of the surface texture difference area; The topographic parameter set is a data set that carries the morphological and height information of the surface texture difference area. The vertical height values ​​of all points are extracted from the topographic parameter set. All points within a fixed range around each point are selected as comparison references. The height value of a single point is calculated to be different from the height values ​​of all points within the reference range. The calculated height difference is mapped to the spatial coordinates of the point. The height difference values ​​of all points are filled into the blank image according to the spatial coordinates to form an elevation difference map that can clearly show the height undulation differences in various parts of the area.

[0034] Perform region growing on the elevation difference map to obtain candidate regions for attachments in areas with surface texture differences; An elevation difference map is an image that marks the distribution of height differences in areas with different surface textures. First, points whose height differences meet the preset criteria are selected in the elevation difference map as initial growth base points. Starting from each initial growth base point, the map is gradually extended to adjacent points in the surrounding space. Points in adjacent locations whose height differences also meet the preset criteria are continuously grouped into the same area until all points that meet the height difference conditions are aggregated, forming a candidate area for attachments that covers all abnormally high protrusions.

[0035] Morphological identification is performed on candidate areas of attachments to obtain the surface attachment areas of Chinese herbal medicine slices; The candidate area for attachments is an aggregated area with abnormally high elevation in the elevation difference map. The external contour shape, internal undulation state, and regional continuity of each candidate area are extracted one by one. The contour and undulation state of the candidate area are checked against the preset attachment morphology standards. The parts with complete and closed contours, continuous internal undulations, and regional range that meet the requirements are retained, while the interference areas with scattered and fragmented contours and irregular internal undulations are eliminated. Finally, the actual distribution range of attachments on the surface of Chinese herbal medicine slices is determined, i.e., the surface attachment area.

[0036] S4. Perform dynamic displacement field tracking on the surface attachment area, and analyze the deformation displacement vector field of the surface attachment area based on the displacement tracking results. Construct a temporal deformation feature map of the surface attachment area based on the deformation displacement vector field. In this embodiment of the invention, the process of obtaining the temporal deformation feature map of the surface attachment region is as follows: Multi-temporal sampling imaging of the surface attachment area yields a temporal image sequence of the surface attachment area; The surface attachment area is the complete area on the surface of Chinese herbal medicine slices that has real attachments, as determined by morphological identification. Images of this area are continuously acquired at fixed time intervals. During the acquisition process, the relative position of the imaging device and the Chinese herbal medicine slices remains unchanged. This ensures that each acquired image only records the appearance of the surface attachment area at that specific time point. All images acquired in chronological order are arranged sequentially according to the acquisition time to form a time-series image sequence that can fully reflect the appearance changes of the surface attachment area at different time points.

[0037] Based on the temporal image sequence, edge contour tracking is performed on the surface attachment region to obtain the edge contour information of the surface attachment region; A time-series image sequence is a collection of images of surface attachment regions arranged in chronological order. Starting from the first image in the time-series image sequence, the boundary between the surface attachment region and the normal surface of the medicinal slice is traversed image by image. Pixels with abrupt changes in brightness and texture at the boundary are identified. These pixels are then connected sequentially according to their spatial position to form closed contour lines. The closed contour lines of the surface attachment region are extracted from all images one by one. The contour lines corresponding to all images are integrated to form edge contour information that can completely represent the boundary morphology of the attachment at each time point.

[0038] Based on the edge contour information, deformation vector mapping is performed on the surface attachment region to obtain the deformation displacement vector field of the surface attachment region; Edge contour information is a set of closed contour lines of the surface attachment area at each time node; the attachment contour lines of the subsequent frame in the time-series image are compared point by point with the corresponding contour lines of the previous frame; the spatial movement direction and distance of each point on the contour from the previous frame to the next frame are determined; the movement direction and distance of each point are integrated into independent displacement data; the displacement data of all points are arranged in an orderly manner according to the spatial coordinates of the surface attachment area; a deformation displacement vector field that can completely characterize the overall deformation trend and local displacement changes of the attachment is formed.

[0039] By arranging the spatiotemporal distribution of the deformation displacement vector field, a temporal deformation characteristic map of the surface attachment region is obtained; The deformation displacement vector field is a complete set of displacement data at various points in the surface attachment area. The deformation displacement vector fields corresponding to each time node are arranged in chronological order of acquisition time. At the same time, the spatial coordinate distribution corresponding to each vector field is preserved. The deformation data in the time dimension and the spatial dimension are integrated to form a time-series deformation feature map that can simultaneously show the complete change law of attachment deformation in time and space.

[0040] S5. Deconstruct the temporal deformation feature map in space-time, identify the abnormal deformation region based on the deconstruction result, and locate the mold precursor region of the Chinese herbal medicine slices based on the persistence and expansion characteristics of the abnormal deformation region. In this embodiment of the invention, the process of obtaining the mold precursor region of traditional Chinese medicine decoction pieces is as follows: The deformation rate hierarchy of the time-series deformation feature map is obtained by dividing the map into rate intervals. The temporal deformation feature map is a data carrier that integrates the temporal and spatial dimensions of deformation data of surface attachment regions. It can simultaneously display the complete change law of attachment deformation in time and space. It extracts the total deformation displacement of each spatial point in the map over a continuous time period. The deformation rate of each point is obtained by dividing the total deformation displacement by the corresponding time span. The deformation rates of all points are arranged in ascending order of numerical value. The arranged deformation rates are evenly divided into multiple continuous intervals. Each interval corresponds to a fixed rate level. The deformation rate of each point is assigned to the corresponding interval and marked with its level. The rate level information of all points is arranged according to the spatial coordinates of the map, forming a deformation rate level distribution covering all spatial points.

[0041] Based on the hierarchical distribution of deformation rate, the increasing trend of the time-series deformation feature map is identified to obtain the distribution of the rate increasing region of the time-series deformation feature map; The deformation rate hierarchy distribution is a set of rate levels corresponding to each spatial point in the time-series deformation feature map. The rate level of each spatial point in the time-series deformation feature map is compared continuously over time. The rate level change of each point at previous and subsequent time nodes is recorded. Points with continuously increasing rate level values ​​are identified as rate-increasing points. All rate-increasing points are labeled according to the spatial coordinates of the time-series deformation feature map. The labeled rate-increasing points are then integrated to form a complete rate-increasing region distribution.

[0042] Spatial proximity aggregation is performed on the distribution of rate-increasing regions to obtain the rate-increasing connected domain distribution of the temporal deformation feature map; The rate-increasing region distribution is a set of spatial labels for all rate-increasing points in the temporal deformation feature map. Starting from any rate-increasing point in the rate-increasing region distribution, we search for the four adjacent spatial locations above, below, left, and right of that point. We group the searched adjacent rate-increasing points into the same aggregation unit. We continue to expand the search in all directions until there are no adjacent rate-increasing points. We organize all the formed aggregation units according to their spatial locations. Each aggregation unit forms an independent connected domain. All connected domains are integrated to form the rate-increasing connected domain distribution.

[0043] The process of obtaining the mold precursor region of traditional Chinese medicine decoction pieces by performing duration-based weighting on the distribution of rate-increasing connected components is as follows: Perform duration-continuous frame tracking on the rate-increasing connected component distribution to obtain a set of continuous frame sequences of the rate-increasing connected component distribution; The rate-increasing connected component distribution is the spatial morphology set of all rate-increasing connected components. For each rate-increasing connected component, the existence state of the connected component is tracked frame by frame according to the time sequence of image acquisition. The number of all consecutive image frames from the first appearance to the disappearance of the connected component is recorded. The consecutive image frame numbers corresponding to each connected component are arranged in chronological order to form a unique continuous frame sequence for each rate-increasing connected component. The continuous frame sequences of all connected components are summarized to obtain a complete set of continuous frame sequences.

[0044] A time-varying trend analysis of the rate was performed on a continuous frame sequence set to obtain a rate growth feature set of the rate-increasing connected component distribution; A continuous frame sequence set is a set of consecutive image frame numbers corresponding to each rate-increasing connected component. For each rate-increasing connected component corresponding to a continuous frame sequence set, the deformation rate of all points within that connected component is extracted frame by frame. The deformation rates of all points within the connected component are summed and divided by the total number of points to obtain the average deformation rate of each frame in that connected component. The average deformation rates of consecutive frames are arranged in chronological order. The magnitude and stability of the change in the average deformation rate over time are statistically analyzed. Feature information showing a continuously increasing magnitude and stable change state is extracted. The above feature information of each connected component is summarized to form a rate-increasing feature set.

[0045] By performing regional expansion pattern identification on the rate growth feature set, the mold precursor region of Chinese herbal medicine pieces is obtained; The rate growth feature set is the set of rate growth feature information corresponding to each rate-increasing connected domain; the rate growth features of each rate-increasing connected domain are checked one by one against the preset mold precursor expansion features; connected domains with a stable rate growth amplitude and a gradually uniform outward expansion of the spatial range are identified as target regions; all connected domains that meet the judgment criteria are integrated according to spatial coordinates; forming the mold precursor region on the surface of Chinese herbal medicine slices.

[0046] S6. Assess the risk of mold growth in the precursor area to obtain intelligent screening results for Chinese herbal medicine pieces; In this embodiment of the invention, the process of obtaining the mold risk level and intelligent screening results of traditional Chinese medicine decoction pieces is as follows: Multidimensional feature parameter extraction was performed on the mold precursor region to obtain the area feature parameter, morphological feature parameter, and deformation rate feature parameter of the mold precursor region; The mold precursor region is the area on the surface of Chinese herbal medicine slices that is about to develop mold, identified through continuous frame sequence tracking of rate-increasing connected components, time-varying rate trend analysis, and region expansion pattern identification. All spatial coordinate points of the mold precursor region are systematically analyzed, and the complete spatial range covered by these points is statistically analyzed. The overall occupancy of this range is then compiled into area feature parameters, which are quantitative data representing the overall size of the mold precursor region. All coordinate points are traced along the outer boundary of the mold precursor region to clarify their arrangement, orientation, and closure. The state is then combined with the distribution density of coordinate points within the region to form morphological feature parameters. These morphological feature parameters are quantitative data used to represent the outline and internal distribution state of the mold precursor region. The temporal deformation feature map corresponding to the mold precursor region is retrieved, and the total deformation displacement of the region within consecutive time nodes is extracted. The total deformation displacement is divided by the corresponding time span to obtain the deformation change scale per unit time. This change scale is then organized into deformation rate feature parameters, which are quantitative data used to represent the speed of deformation change in the mold precursor region.

[0047] Based on area characteristic parameters, morphological characteristic parameters, and deformation rate characteristic parameters, risk factors are weighted and fused in the mold precursor region to obtain the comprehensive mold risk index of the mold precursor region. According to a predetermined fixed weight ratio, the influence of area characteristic parameters, morphological characteristic parameters, and deformation rate characteristic parameters is integrated. First, the area characteristic parameter is combined with its corresponding weight influence to obtain the risk value of the area dimension. Then, the morphological characteristic parameter is combined with its corresponding weight influence to obtain the risk value of the morphological dimension. Next, the deformation rate characteristic parameter is combined with its corresponding weight influence to obtain the risk value of the rate dimension. The risk values ​​of the three dimensions are then added together according to the spatial and temporal relationship. The overall value obtained is the comprehensive mold risk index, which is the only quantitative result that reflects the overall level of mold risk in the mold precursor area.

[0048] Based on the comprehensive index of mold risk, the risk level of the mold precursor area is mapped to obtain the mold risk level of the mold precursor area. A risk level classification standard corresponding to the comprehensive mold risk index is defined in advance. This standard clearly divides the index range corresponding to different risk levels. The comprehensive mold risk index is compared with the classification standard one by one to determine the risk level range to which the index belongs. The level corresponding to the range after comparison is the mold risk level of the mold precursor area. The mold risk level is the classification result that reflects the level of mold risk in the mold precursor area.

[0049] By making decisions based on the level of mold risk, intelligent screening results for Chinese herbal medicine slices are generated. According to the established criteria for judging qualified medicinal herbs, the criteria clearly define the range of mold risk levels corresponding to qualified medicinal herbs. The mold risk levels are compared with the qualification criteria. Medicinal herbs with mold risk levels that meet the qualification criteria are judged as qualified medicinal herbs, while those with mold risk levels that do not meet the qualification criteria are judged as unqualified medicinal herbs. The results of summarizing and organizing all qualified and unqualified medicinal herbs are the intelligent screening results of medicinal herbs. The intelligent screening results are the final judgment data reflecting whether medicinal herbs meet the usage requirements.

[0050] like Figure 2 The diagram shown is a functional block diagram of an intelligent screening system for traditional Chinese medicine decoction pieces that incorporates image recognition technology, provided by an embodiment of the present invention.

[0051] The intelligent screening system 100 for traditional Chinese medicine decoction pieces incorporating image recognition technology described in this invention can be installed in an electronic device. Depending on the functions implemented, the intelligent screening system 100 for traditional Chinese medicine decoction pieces incorporating image recognition technology may include a surface texture difference identification module 101, a three-dimensional morphology reconstruction module 102, a surface attachment analysis module 103, a temporal deformation feature analysis module 104, a mold precursor region clustering module 105, and a mold risk assessment and screening module 106. The modules described in this invention can also be called units, which refer to a series of computer program segments that can be executed by the processor of an electronic device and can perform fixed functions, and are stored in the memory of the electronic device.

[0052] In this embodiment, the functions of each module / unit are as follows: The surface texture difference identification module 101 is used to extract texture features from the surface image of Chinese herbal medicine slices and to identify the difference between the extraction results and the standard surface texture features of Chinese herbal medicine slices to obtain the surface texture difference region of Chinese herbal medicine slices. The three-dimensional topography reconstruction module 102 is used to perform sequential illumination imaging on the surface texture difference region to obtain multi-angle images of the surface texture difference region, and to construct three-dimensional topography data of the surface texture difference region based on the multi-angle images. The surface attachment analysis module 103 is used to perform morphological parameter analysis on the three-dimensional morphology data to obtain the surface attachment area of ​​the Chinese herbal medicine slices. The temporal deformation feature analysis module 104 is used to dynamically track the displacement field of the surface attachment region, and to analyze the deformation displacement vector field of the surface attachment region based on the displacement tracking result, and to construct the temporal deformation feature map of the surface attachment region based on the deformation displacement vector field. The mold precursor region clustering module 105 is used to perform spatiotemporal deconstruction of the temporal deformation feature map, identify abnormal deformation regions based on the deconstruction results, and locate the mold precursor region of the Chinese herbal medicine slices based on the persistence and expansion characteristics of the abnormal deformation regions. The mold risk assessment and screening module 106 is used to assess the mold risk of the mold precursor area and obtain intelligent screening results for Chinese herbal medicine pieces.

[0053] In the several embodiments provided by the present invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the division of modules is merely a logical functional division, and there may be other division methods in actual implementation.

[0054] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules may be selected to achieve the purpose of this embodiment according to actual needs.

[0055] In addition, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0056] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0057] This application embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0058] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for intelligent screening of traditional Chinese medicine decoction pieces incorporating image recognition technology, characterized in that, The method includes: S1, extracting texture features from the surface image of Chinese herbal medicine slices, and comparing the extraction results with the standard surface texture features of Chinese herbal medicine slices to obtain the surface texture difference region of Chinese herbal medicine slices; S2. Perform sequential illumination imaging on the surface texture difference region to obtain multi-angle images of the surface texture difference region, and construct three-dimensional morphology data of the surface texture difference region based on the multi-angle images. S3. Perform morphological parameter analysis on the three-dimensional morphology data to obtain the surface attachment area of ​​the Chinese herbal medicine slices; S4. Perform dynamic displacement field tracking on the surface attachment area, and analyze the deformation displacement vector field of the surface attachment area based on the displacement tracking results. Construct a temporal deformation feature map of the surface attachment area based on the deformation displacement vector field. S5. Deconstruct the temporal deformation feature map in space-time, identify the abnormal deformation region based on the deconstruction result, and locate the mold precursor region of the Chinese herbal medicine slices based on the persistence and expansion characteristics of the abnormal deformation region. S6. Assess the risk of mold growth in the precursor areas to obtain intelligent screening results for Chinese herbal medicine pieces.

2. The intelligent screening method for traditional Chinese medicine decoction pieces incorporating image recognition technology as described in claim 1, characterized in that, The process of obtaining the surface texture difference regions of Chinese herbal medicine slices is as follows: Multispectral imaging of Chinese medicinal herbs was performed to obtain the original surface images of the herbs. Based on preset feature bands, the original surface image is separated and reconstructed in the frequency domain to obtain a texture highlighting image of Chinese herbal medicine slices; Neighborhood relation quantization is performed on the texture-highlighting image to obtain the local texture coding value of the texture-highlighting image; Obtain the standard texture coding value of Chinese herbal medicine slices, and perform a deviation space mapping between the local texture coding value and the standard texture coding value to obtain the texture difference localization map of Chinese herbal medicine slices; Connectivity analysis was performed on the texture difference localization map to obtain the surface texture difference regions of Chinese herbal medicine slices.

3. The intelligent screening method for traditional Chinese medicine decoction pieces incorporating image recognition technology as described in claim 1, characterized in that, The process of obtaining multi-angle images of surface texture difference regions is as follows: Discretize the surface texture difference region to obtain the imaging block of the surface texture difference region; According to the preset illumination angle sequence, the imaging block is illuminated sequentially from multiple angles to obtain the reflected light field distribution of the imaging block; By spatially stitching together the reflected light field distribution of the imaging blocks under the same illumination angle, multi-angle images of areas with different surface textures are obtained.

4. The intelligent screening method for traditional Chinese medicine decoction pieces incorporating image recognition technology as described in claim 3, characterized in that, The process of constructing three-dimensional topographic data of surface texture difference regions based on multi-angle images is as follows: Height field reconstruction is performed on multi-angle images to obtain depth information fields of surface texture difference regions; Based on the pixel mapping relationship between the depth information field and the multi-angle image, texture mapping is performed on the multi-angle image to obtain the texture attachment depth map of the surface texture difference region; The texture attachment depth map is reconstructed into a point cloud according to spatial coordinates to obtain the three-dimensional morphological data of the surface texture difference region.

5. The intelligent screening method for traditional Chinese medicine decoction pieces incorporating image recognition technology as described in claim 1, characterized in that, The process of obtaining the surface deposits on the Chinese herbal medicine slices is as follows: Feature extraction is performed on the three-dimensional topography data to obtain the topography parameter set of the surface texture difference region; The topographic parameter set is compared with the elevation to generate an elevation difference map of the surface texture difference area; Perform region growing on the elevation difference map to obtain candidate regions for attachments in areas with surface texture differences; Morphological identification is performed on the candidate areas of the attachments to obtain the surface attachment areas of Chinese herbal medicine slices.

6. The intelligent screening method for traditional Chinese medicine decoction pieces incorporating image recognition technology as described in claim 1, characterized in that, The process of obtaining the temporal deformation feature map of the surface attachment region is as follows: Multi-temporal sampling imaging of the surface attachment area yields a temporal image sequence of the surface attachment area; Based on the temporal image sequence, edge contour tracking is performed on the surface attachment region to obtain the edge contour information of the surface attachment region; Based on the edge contour information, deformation vector mapping is performed on the surface attachment region to obtain the deformation displacement vector field of the surface attachment region; By arranging the spatiotemporal distribution of the deformation displacement vector field, a temporal deformation characteristic map of the surface attachment region is obtained.

7. The intelligent screening method for traditional Chinese medicine decoction pieces incorporating image recognition technology as described in claim 1, characterized in that, The process of obtaining the mold precursor region of traditional Chinese medicine decoction pieces is as follows: The deformation rate hierarchy of the time-series deformation feature map is obtained by dividing the map into rate intervals. Based on the hierarchical distribution of deformation rate, the increasing trend of the time-series deformation feature map is identified to obtain the distribution of the rate increasing region of the time-series deformation feature map; Spatial proximity aggregation is performed on the distribution of rate-increasing regions to obtain the rate-increasing connected domain distribution of the temporal deformation feature map; By performing continuous duration authentication on the distribution of rate-increasing connected regions, the mold precursor region of traditional Chinese medicine decoction pieces is obtained.

8. The intelligent screening method for traditional Chinese medicine decoction pieces incorporating image recognition technology as described in claim 7, characterized in that, The process of obtaining the mold precursor region of traditional Chinese medicine decoction pieces is as follows: Perform duration-continuous frame tracking on the rate-increasing connected component distribution to obtain a set of continuous frame sequences of the rate-increasing connected component distribution; A time-varying trend analysis of the rate was performed on a continuous frame sequence set to obtain a rate growth feature set of the rate-increasing connected component distribution; By performing regional expansion pattern identification on the rate growth feature set, the mold precursor region of Chinese herbal medicine slices is obtained.

9. The intelligent screening method for traditional Chinese medicine decoction pieces incorporating image recognition technology as described in claim 1, characterized in that, The process of obtaining the mold risk level and intelligent screening results of Chinese herbal medicine slices is as follows: Multidimensional feature parameter extraction was performed on the mold precursor region to obtain the area feature parameter, morphological feature parameter, and deformation rate feature parameter of the mold precursor region; Based on area characteristic parameters, morphological characteristic parameters, and deformation rate characteristic parameters, risk factors are weighted and fused in the mold precursor region to obtain the comprehensive mold risk index of the mold precursor region. Based on the comprehensive index of mold risk, the risk level of the mold precursor area is mapped to obtain the mold risk level of the mold precursor area. By screening and deciding on the level of mold risk, intelligent screening results for Chinese herbal medicine slices are generated.

10. A smart screening system for traditional Chinese medicine decoction pieces incorporating image recognition technology, characterized in that, The system is used to implement the intelligent screening method for traditional Chinese medicine decoction pieces incorporating image recognition technology according to any one of claims 1-9, the system comprising: The surface texture difference identification module is used to extract texture features from the surface image of Chinese herbal medicine slices and to identify the differences between the extraction results and the standard surface texture features of Chinese herbal medicine slices to obtain the surface texture difference areas of Chinese herbal medicine slices. The 3D topography reconstruction module is used to perform sequential illumination imaging on the surface texture difference region to obtain multi-angle images of the surface texture difference region, and to construct 3D topography data of the surface texture difference region based on the multi-angle images. The surface attachment analysis module is used to perform morphological parameter analysis on three-dimensional morphology data to obtain the surface attachment area of ​​Chinese herbal medicine slices; The temporal deformation feature analysis module is used to dynamically track the displacement field of the surface attachment region, and to analyze the deformation displacement vector field of the surface attachment region based on the displacement tracking results, and to construct the temporal deformation feature map of the surface attachment region based on the deformation displacement vector field. The mold precursor region clustering module is used to perform spatiotemporal deconstruction of the temporal deformation feature map, identify abnormal deformation regions based on the deconstruction results, and locate the mold precursor region of Chinese herbal medicine slices based on the persistence and expansion characteristics of the abnormal deformation regions. The mold risk assessment and screening module is used to assess the mold risk of the precursor area and obtain intelligent screening results for Chinese herbal medicine pieces.

Images