Multi-center storage tongue image fast indexing method and system and readable storage medium

Abstract

The application discloses a multi-center storage tongue diagnosis image quick indexing method, system and readable storage medium. The method comprises the following steps: extracting feature quantities of images to be diagnosed and converting the feature quantities into index data containing a plurality of keywords; based on the index data and according to an inverted index algorithm, calling target images from each storage center; wherein calling the target images comprises: counting the distribution position and concentration of each keyword in a record list in each storage center; changing the keyword order in the index data according to the distribution position and concentration; and calling the target images based on the new index data. According to the application, the order of the keywords in the index data is adjusted according to the position of the index keywords in the record list content in each storage center, the indexing efficiency can be improved, the reliability of each keyword in each storage center is assigned, the reliability of each index data is determined, the number of target images called from each storage center is adjusted based on the above values, and the accuracy and reliability of the index calling are improved.

Description

Technical Field

[0001] This invention relates to the field of image data search technology, and more specifically, to a method, system, and readable storage medium for fast indexing of tongue diagnosis images stored in a multi-center location. Background Technology

[0002] Tongue diagnosis is a simple and effective method that uses the color and shape of the tongue to assist medical staff in diagnosing and differentiating diseases.

[0003] With the continuous improvement of big data, artificial intelligence, and image processing technologies, intelligent tongue diagnosis technology is also developing rapidly. The current steps for intelligent tongue diagnosis typically involve: acquiring tongue images for comparison and storing them in association with various symptoms; then inputting the image to be diagnosed; comparing it with existing tongue images in the database; finding similar images; and thus obtaining the corresponding symptom data. As can be seen from the above process, a major guarantee of the accuracy and reliability of tongue diagnosis results comes from the richness and completeness of the comparison images in the database, as well as the search and recognition algorithms used during image retrieval.

[0004] Currently, to enrich the database with tongue diagnosis images for comparison, the database is typically made available to tongue diagnosis centers located in various regions for the collection of tongue diagnosis images and their corresponding disease data. The collected images and related data are then stored in the storage devices or cloud storage of each center. While this method facilitates the collection and storage of comparative images, it creates difficulties for later searching for similar tongue diagnosis images. The input image for diagnosis needs to find similar comparative images and their associated disease data within a short period. Indexing using conventional data search methods clearly cannot meet these requirements. Summary of the Invention

[0005] To address the problem of the inability to quickly identify and retrieve tongue diagnosis images stored in multiple centers in practical applications, this application aims to propose a fast indexing method for tongue diagnosis images stored in multiple centers. This method not only improves the indexing speed of comparison images but also significantly enhances search accuracy. To implement the aforementioned fast indexing method for tongue diagnosis images stored in multiple centers, this application also aims to propose a fast indexing system for these images, and a computer-readable storage medium. The specific solutions are as follows:

[0006] A fast indexing method for multi-center stored tongue diagnosis images includes:

[0007] Extract the features of the image to be diagnosed and convert each feature into index data containing multiple keywords based on the set transformation model;

[0008] Based on the above index data and according to the inverted index algorithm, the target images are retrieved from each storage center;

[0009] The step of retrieving the target image from each storage center includes:

[0010] Analyze the distribution and concentration of various keywords in the record list of each storage center;

[0011] Based on the above distribution location and concentration, adjust the index data for each storage center and change the sorting of keywords in the index data;

[0012] The target image is retrieved from each storage center based on the adjusted index data.

[0013] The above technical solution utilizes an inverted index method to quickly obtain a list of matching records from a large number of comparative images, thereby acquiring the target image and corresponding disease data. Furthermore, since the distribution and concentration of keywords vary across different storage centers, different index data can be used for different storage centers, allowing keywords with higher distribution to be indexed more quickly, further shortening the inverted indexing time.

[0014] Furthermore, retrieving the target image from each storage center also includes:

[0015] Based on the accuracy of the association between keywords and comparison images in each storage center, assign a reliability value to each keyword in each storage center;

[0016] Based on the reliability values ​​of each keyword in each storage center, calculate the sum of the reliability values ​​of the corresponding index data in each storage center;

[0017] Weight values ​​are assigned to the index data corresponding to each storage center based on the sum of the reliability values;

[0018] Set the number of target images to be retrieved, and retrieve the corresponding number of comparison images from each storage center according to the ratio of the weight values.

[0019] Since different storage centers use different methods to extract features from comparison images, and the correspondence between disease data and features is also different, the reliability of the comparison images retrieved from the same set of keywords is also different. The above technical solution can be used to process the search results based on the reliability values ​​of each keyword in each storage center, making the index results more reliable.

[0020] Furthermore, the statistical analysis of the distribution and concentration of various keywords in the record list across each storage center includes:

[0021] Retrieve a list of record IDs and set keywords associated with tongue diagnosis;

[0022] Use a word segmenter to perform word segmentation and retrieval on the contents of each record list;

[0023] Based on the inclusion relationship between the list IDs of the search results and each keyword, generate an association table between each list ID and each keyword in each data center;

[0024] Based on the association table, the distribution position and concentration of each keyword in the record list are determined.

[0025] The above technical solution associates the keywords contained in each content list with the list ID to form an association table. Based on the association table, the distribution of each keyword in the content list can be clearly known. Thus, the distribution of combinations of keywords in the content list can be further known. When retrieving relevant comparison images, i.e., searching for and outputting the corresponding list ID, the indexing time can be shortened and the indexing efficiency improved by adjusting the positional relationship of keywords in the index data.

[0026] Furthermore, retrieving the target image from each storage center also includes:

[0027] The number of comparison images retrieved from each storage center is determined based on the weight values.

[0028] The index order of each storage center is determined based on the stated quantity;

[0029] The indexing order is positively correlated with the number of comparison images retrieved.

[0030] The above technical solution improves the efficiency of early indexing by retrieving more comparison images in a more timely manner. At the same time, comparison images retrieved earlier are more reliable, making it easier for users to make more accurate diagnostic judgments based on the comparison images.

[0031] Furthermore, the method also includes:

[0032] When the number of comparison images retrieved from the storage center exceeds the requirement, a set number of comparison images are randomly selected.

[0033] When the number of comparison images retrieved from the storage center is less than the requirement, the comparison images retrieved from the storage center are directly output, or

[0034] Comparison images are selected from the comparison images output from the storage center with the highest weight value to supplement the comparison images until the total number of comparison images meets the set number of target images to be retrieved.

[0035] The above technical solution can balance the number of comparison images retrieved from various storage centers, ensuring the quality of the index while guaranteeing the number of retrieved images.

[0036] A multi-center storage system for fast indexing of tongue diagnosis images includes:

[0037] The input unit is configured to receive and store the image to be diagnosed input by the user;

[0038] A feature extraction unit is configured to receive the image to be diagnosed and extract and output the feature quantities in the image to be diagnosed.

[0039] The index data generation unit is configured to receive the feature data of the image to be diagnosed and generate intermediate index data containing multiple keywords based on a set transformation model before outputting it.

[0040] The index data adjustment unit is configured to receive intermediate index data and, based on the distribution and concentration of various keywords in the record list in each storage center, change the sorting of keywords in the index data to generate index data.

[0041] The inverted index unit is configured to receive the index data corresponding to the image to be diagnosed and retrieve the target image from each storage center using the inverted index algorithm.

[0042] The output unit is configured to receive the target image and output it.

[0043] The above technical solutions can improve the efficiency of data indexing by using inverted indexes. At the same time, changing the order of keywords in the index data can further improve the efficiency of retrieving target images from different storage centers.

[0044] Furthermore, the index data adjustment unit includes:

[0045] The keyword extraction module includes a word segmentation algorithm, which receives the keywords and performs word segmentation retrieval on the contents of the record list;

[0046] The association table generation module is configured to generate an association table between each list ID and each keyword in each data center based on the inclusion relationship between the list ID of the search results and each keyword.

[0047] The location and concentration output module is configured to determine the distribution location and concentration of each keyword in the record list based on the association table.

[0048] The position adjustment module is configured to adjust the sorting of each keyword in the index data based on the distribution and concentration of each keyword in the index data and then output the result.

[0049] Furthermore, the system also includes:

[0050] A reliability marking unit is configured to be data-connected to the index data generation unit and is used to assign a reliability value to each keyword in each storage center based on the accuracy of the association between keywords and comparison images in each storage center.

[0051] The index data weight generation unit is configured to calculate the sum of the reliability values ​​of the index data corresponding to each storage center based on the reliability values ​​of each keyword in each storage center, and to assign weight values ​​to the index data corresponding to each storage center according to the sum of the reliability values.

[0052] The output unit includes:

[0053] The image output module sets the number of target images to be retrieved, retrieves the corresponding number of comparison images from each storage center according to the ratio of the weight values, and outputs them.

[0054] The above technical solution allows for the assignment of values ​​to each index data based on the reliability of each keyword in each storage center. This enables the screening of target images retrieved from the index later, thereby improving the reliability and accuracy of the index data.

[0055] Furthermore, the image output module includes:

[0056] The quantity allocation submodule is configured to determine the number of comparison images to retrieve from each storage center based on the weight value.

[0057] The output order submodule is configured to determine the index order of each storage center based on the quantity.

[0058] The indexing order is positively correlated with the number of comparison images retrieved.

[0059] A computer-readable storage medium loaded with a program algorithm for implementing the multi-center storage tongue diagnosis image fast indexing method as described above.

[0060] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0061] This application employs an inverted index to retrieve target images from various storage centers, improving indexing efficiency. Furthermore, it adjusts the sorting of keywords in the index data based on their position within the record list in each storage center, further enhancing indexing efficiency. Finally, it assigns reliability values ​​to each keyword in each storage center, determining the reliability of each index data point for each storage center. Based on these values, it adjusts the number of target images retrieved from each storage center, further improving the accuracy and reliability of index retrieval. Attached Figure Description

[0062] Figure 1A schematic diagram of a fast indexing method for multi-center storage of tongue diagnosis images;

[0063] Figure 2 This is a schematic diagram illustrating the method for retrieving target images from various storage centers;

[0064] Figure 3 This is a diagram illustrating the distribution and concentration of keywords in the record list;

[0065] Figure 4 A schematic diagram illustrating the method for further retrieving target images from various storage centers;

[0066] Figure 5 A schematic diagram of the functional framework of a multi-center storage system for rapid indexing of tongue diagnosis images.

[0067] Figure reference numerals: 100, Input unit; 200, Feature extraction unit; 300, Index data generation unit; 400, Index data adjustment unit; 410, Keyword extraction module; 420, Association table generation module; 430, Location and concentration output module; 440, Location adjustment module; 500, Inverted index unit; 600, Output unit; 610, Image output module; 6101, Quantity allocation submodule; 6102, Output order submodule; 700, Reliability labeling unit; 800, Index data weight generation unit. Detailed Implementation

[0068] The present invention will be further described in detail below with reference to the embodiments and figures, but the implementation of this application is not limited thereto.

[0069] A fast indexing method for multi-center storage of tongue diagnosis images, such as Figure 1 As shown, the main steps include the following:

[0070] S100: Extract the feature quantities of the image to be diagnosed and convert each feature quantity into index data containing multiple keywords based on the set conversion model;

[0071] S200: Based on the above index data and according to the inverted index algorithm, retrieve the target image from each storage center.

[0072] In step S100 above, the features of the image to be diagnosed include tongue coating (LAB), tongue coating RGB, tongue body (LAB), tongue body RGB, thickness of tongue coating, and tongue body size. These features can be obtained through specific image recognition algorithms. After obtaining these features, they need to be converted into text, letters, numbers, or combinations thereof that can be stored in a database and used as index keywords, such as white coating, pale red tongue, and thick coating. Since a single image to be diagnosed contains multiple features, multiple keywords are extracted for each image. These keywords are then combined to form a text combination, i.e., index data, such as {white coating, pale red tongue, thick coating, diabetes}.

[0073] Since each storage center stores a large amount of target image data for comparison, directly using a forward index to find the internal features (i.e., the keywords) of the target image data would be time-consuming and laborious. Therefore, in step S200, this embodiment of the application uses an inverted index algorithm to retrieve the target images from each storage center. The process is roughly as follows: {white tongue coating, pale red tongue, thick tongue coating, diabetes} -> data center N -> record ID list, that is, finding the record ID list through keywords.

[0074] Since the target images are retrieved from different storage centers, in order to further improve the indexing speed, the retrieval of target images from each storage center, such as... Figure 2 As shown, it further includes:

[0075] S210, Analyze the distribution and concentration of various keywords in the record list of each storage center, such as... Figure 3 As shown;

[0076] S211, Based on the above distribution location and concentration, adjust the index data for each storage center and change the sorting of keywords in the index data;

[0077] S212, retrieve the target image from each storage center based on the adjusted index data.

[0078] like Figure 3 As shown, since the distribution and concentration of keywords in different storage centers are not the same, different index data are used for different storage centers. Keywords with higher distribution can be indexed more quickly, thereby further shortening the inverted index time.

[0079] In step S210 above, the distribution and concentration of various keywords in the record list of each storage center are statistically analyzed, including:

[0080] S2101, retrieve the list of record IDs and set keywords associated with tongue diagnosis;

[0081] S2102, Use a word segmenter to perform word segmentation retrieval on the contents of each record list;

[0082] S2103, based on the inclusion relationship between the list IDs of the search results and each keyword, generate a table relating each list ID to each keyword in each data center, such as... Figure 3 As shown;

[0083] S2104, determine the distribution position and concentration of each keyword in the record list based on the association table.

[0084] The above technical solution associates the keywords contained in each content list with the list ID to form an association table. Based on the association table, the distribution position of each keyword in the content list can be clearly known. Thus, the distribution position of the combination of each keyword in the content list can be further known. When retrieving relevant comparison images, that is, when searching for and outputting the corresponding list ID, the indexing time can be shortened and the indexing efficiency can be improved by adjusting the positional relationship of keywords in the index data.

[0085] Because different storage centers use different methods to extract features from comparison images (i.e., index target images), and the correspondence between disease data and features is also different, it means that the reliability of the target images retrieved from the same set of keywords is also different.

[0086] To make the indexing results more reliable, the process of retrieving target images from various storage centers is optimized, such as... Figure 4 As shown, it also includes:

[0087] S220, based on the accuracy of the association between keywords and comparison images in each storage center, assign a reliability value to each keyword in each storage center;

[0088] S221, Based on the reliability values ​​of each keyword in each storage center, calculate the sum of the reliability values ​​of the index data corresponding to each storage center;

[0089] S222, assign weight values ​​to the index data corresponding to each storage center based on the sum of reliability values;

[0090] S223, set the number of target images to be retrieved, and retrieve the corresponding number of comparison images from each storage center according to the weight ratio.

[0091] In step S221 above, each keyword in each different storage center is assigned a reliability value. This reliability value is a positive integer and is stored in association with the keyword, for example, "yellow tongue coating -2". After this assignment, each keyword in each storage center has a corresponding value. After generating the relevant index data, as described in step S221, the sum of the reliability values ​​of the index data is calculated. For example, for a given set of index data, the sum of reliability values ​​for storage center A is 8, for storage center B it is 10, and for storage center C it is 4. In step S222, the ratio of these reliability sums is used as the weight value for the corresponding index data in each storage center. The weight value for the index data corresponding to storage center A is 8 / 22 = 0.36 (rounded to two decimal places, the same below), the weight value for the index data corresponding to storage center B is 10 / 22 = 0.46, and similarly, the weight value for the index data corresponding to storage center C is 4 / 22 = 0.18.

[0092] Since an index data may correspond to multiple images in a storage center, it is not necessary to retrieve all identical images in practical applications. Therefore, in step S223, the number of target images to be retrieved is first set, and then the corresponding number of comparison images are retrieved from each storage center according to the weight ratio mentioned above. For example, if the number of target images to be retrieved is 100, then 36, 46, and 18 images are retrieved from storage centers A, B, and C respectively. Thus, the storage center with higher reliability provides more images, which can greatly improve the reliability of the indexing results.

[0093] In practical applications, when the number of comparison images retrieved from the storage center exceeds the requirement, a set number of comparison images are randomly selected. For example, if 36 target images need to be retrieved from storage center A, but there are 50 target images in storage center A that meet the index data, then 36 images are randomly selected for output.

[0094] When the number of comparison images retrieved from the storage center is less than the requirement, the comparison images retrieved directly from the storage center or the comparison images output from the storage center with the highest weight value are selected to supplement the comparison images until the total number of comparison images meets the set number of target images to be retrieved.

[0095] The above technical solution can balance the number of comparison images retrieved from various storage centers, ensuring the quality of the index while guaranteeing the number of retrieved images.

[0096] Furthermore, the target image is retrieved from each storage center, such as... Figure 4 As shown, it also includes:

[0097] S224, determine the number of comparison images to be retrieved from each storage center based on the weight values;

[0098] S225, determine the index order of each storage center according to the quantity, wherein the index order is positively correlated with the number of comparison images retrieved, that is, the more comparison images there are, the earlier they are retrieved.

[0099] The above approach can improve the efficiency of early indexing, and the comparison images obtained from earlier indexing are more reliable, making it easier for users to make more accurate judgments on diagnostic results based on the comparison images.

[0100] To implement the aforementioned method for fast indexing of tongue diagnosis images stored in multiple centers, this application also proposes a fast indexing system for tongue diagnosis images stored in multiple centers, such as... Figure 5 As shown, it mainly includes: input unit 100, feature extraction unit 200, index data generation unit 300, index data adjustment unit 400, inverted index unit 500, and output unit 600.

[0101] The input unit 100 is configured to receive and store the image to be diagnosed input by the user. In specific embodiments, it can be configured as a human-computer interaction data input interface, such as an image acquisition camera, a data transmission interface module, etc.

[0102] The feature extraction unit 200 is configured to receive the image to be diagnosed and extract and output the feature quantities in the image to be diagnosed. The feature quantities include tongue coating LAB, tongue coating RGB, tongue body LAB, tongue body RGB, as well as textual features such as the thickness of the tongue coating and the size of the tongue body, which can be extracted by a specific image recognition algorithm.

[0103] The index data generation unit 300 is configured to receive feature values ​​from the image to be diagnosed and generate intermediate index data containing multiple keywords based on a set conversion model before outputting it. The index data generation unit 300 mainly plays the role of data conversion, converting each feature value into text, letters, numbers, or combinations thereof that can be used as index keywords.

[0104] The index data adjustment unit 400 is configured to receive intermediate index data and, based on the distribution and concentration of various keywords in the record list in each storage center, change the sorting of keywords in the index data to generate index data.

[0105] The inverted index unit 500 is configured to receive the index data corresponding to the image to be diagnosed and retrieve the target image from each storage center using the inverted index algorithm. Since the inverted index algorithm has been widely used in practice, its specific working principle will not be elaborated here.

[0106] The output unit 600 is configured to receive and output the target image. In practical applications, it is configured as a related human-computer interaction component, such as a display.

[0107] In detail, the index data adjustment unit 400 includes a keyword extraction module 410, a relational table generation module 420, a position and concentration output module 430, and a position adjustment module 440. The keyword extraction module 410 includes a word segmentation algorithm, such as the IK word segmentation algorithm, which receives keywords and performs word segmentation retrieval on the content of the record list. The relational table generation module 420 is configured to generate a relational table between each list ID and each keyword in each data center based on the inclusion relationship between the list IDs marked in the search results and each keyword.

[0108] The location and concentration output module 430 is configured to determine the distribution location and concentration of each keyword in the record list based on the association table. The location adjustment module 440 is configured to adjust the sorting of each keyword in the index data based on the distribution location and concentration of each keyword in the index data and then output the result.

[0109] To improve the reliability of the system's output, the multi-center storage tongue diagnosis image fast indexing system of this application also includes: a reliability marking unit 700 and an index data weight generation unit 800.

[0110] The reliability marking unit 700 is configured to be data-connected to the index data generation unit 300. It assigns a reliability value to each keyword in each storage center based on the accuracy of the association between the keywords and the comparison images in each storage center. In practical applications, these reliability values ​​can be pre-set by the user. The index data weight generation unit 800 is configured to calculate the sum of the reliability values ​​of the index data corresponding to each storage center based on the reliability values ​​of each keyword in each storage center, and assign weight values ​​to the index data corresponding to each storage center according to the sum of the reliability values. For ease of calculation, all reliability values ​​are positive integers. Correspondingly, an adder is configured in the index data weight generation unit 800 to calculate the sum.

[0111] The output unit 600 includes an image output module 610, which is used to set the number of target images to be retrieved, retrieve the corresponding number of comparison images from each storage center according to the weight ratio, and output them. Correspondingly, a divider is configured in the image output module 610.

[0112] In a further optimized version, the image output module 610 includes a quantity allocation submodule 6101 and an output order submodule 6102.

[0113] The quantity allocation submodule 6101 is configured to determine the number of comparison images retrieved from each storage center based on the weight value, and the output order submodule 6102 is configured to determine the index order of each storage center based on the quantity. The index order is positively correlated with the number of comparison images retrieved, that is, the higher the weight value, the earlier the output order.

[0114] To facilitate the widespread application of the multi-center storage tongue diagnosis image fast indexing method described in this application, this application also proposes a computer-readable storage medium loaded with a program algorithm for implementing the aforementioned multi-center storage tongue diagnosis image fast indexing method. In a specific embodiment, the computer-readable storage medium can be configured as a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0115] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principle of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for fast indexing of tongue diagnosis images stored in a multi-center environment, characterized in that, include: Extract the features of the image to be diagnosed and convert each feature into index data containing multiple keywords based on the set transformation model; Based on the above index data and according to the inverted index algorithm, the target images are retrieved from each storage center; The step of retrieving the target image from each storage center includes: Analyze the distribution and concentration of various keywords in the record list of each storage center; Based on the above distribution location and concentration, adjust the index data for each storage center and change the sorting of keywords in the index data; Target images are retrieved from various storage centers based on the adjusted index data; The statistics include the distribution and concentration of various keywords in the record list of each storage center, including: Retrieve a list of record IDs and set keywords associated with tongue diagnosis; Use a word segmenter to perform word segmentation and retrieval on the contents of each record list; Based on the inclusion relationship between the list IDs of the search results and each keyword, generate an association table between each list ID and each keyword in each data center; Based on the association table, determine the distribution position and concentration of each keyword in the record list; The process of retrieving target images from various storage centers also includes: Based on the accuracy of the association between keywords and comparison images in each storage center, assign a reliability value to each keyword in each storage center; Based on the reliability values ​​of each keyword in each storage center, calculate the sum of the reliability values ​​of the corresponding index data in each storage center; Weight values ​​are assigned to the index data corresponding to each storage center based on the sum of the reliability values; Set the number of target images to be retrieved, and retrieve the corresponding number of comparison images from each storage center according to the ratio of the weight values; The process of retrieving target images from various storage centers also includes: The number of comparison images retrieved from each storage center is determined based on the weight values. The index order of each storage center is determined based on the stated quantity; The index order is positively correlated with the number of comparison images retrieved. The method further includes: When the number of comparison images retrieved from the storage center exceeds the requirement, a set number of comparison images are randomly selected. When the number of comparison images retrieved from the storage center is less than the requirement, the comparison images retrieved from the storage center are directly output, or Comparison images are selected from the comparison images output from the storage center with the highest weight value to supplement the comparison images until the total number of comparison images meets the set number of target images to be retrieved.

2. A multi-center storage tongue diagnosis image fast indexing system, characterized in that, include: The input unit (100) is configured to receive and store the image to be diagnosed input by the user; The feature extraction unit (200) is configured to receive the image to be diagnosed and extract and output the feature quantities in the image to be diagnosed; The index data generation unit (300) is configured to receive the feature data of the image to be diagnosed and generate intermediate index data containing multiple keywords based on a set transformation model before outputting it. The index data adjustment unit (400) is configured to receive intermediate index data and, based on the distribution and concentration of various keywords in the record list in each storage center, change the sorting of keywords in the index data to generate index data. The inverted index unit (500) is configured to receive the index data corresponding to the image to be diagnosed and retrieve the target image from each storage center using the inverted index algorithm; The output unit (600) is configured to receive the target image and output it. The index data adjustment unit (400) includes: The keyword extraction module (410) includes a word segmentation algorithm, which receives the keywords and performs word segmentation retrieval on the contents of the record list; The association table generation module (420) is configured to generate an association table of each list ID and each keyword in each data center based on the inclusion relationship between the list ID of the search results tag list and each keyword; The location and concentration output module (430) is configured to determine the distribution location and concentration of each keyword in the record list based on the association table; The position adjustment module (440) is configured to adjust the sorting of each keyword in the index data according to the distribution position and concentration of each keyword in the index data and output the result. The system also includes: The reliability marking unit (700) is configured to be data connected to the index data generation unit (300) and is used to assign a reliability value to each keyword in each storage center based on the accuracy of the association between the keywords and the comparison images in each storage center. The index data weight generation unit (800) is configured to calculate the sum of the reliability values ​​of the index data corresponding to each storage center based on the reliability values ​​of each keyword in each storage center, and to assign weight values ​​to the index data corresponding to each storage center according to the sum of the reliability values. The output unit (600) includes: The image output module (610) sets the number of target images to be retrieved, retrieves the corresponding number of comparison images from each storage center according to the ratio of the weight values, and outputs them. The image output module (610) includes: The quantity allocation submodule (6101) is configured to determine the number of comparison images to be retrieved from each storage center based on the weight value; The output order submodule (6102) is configured to determine the index order of each storage center based on the quantity. The indexing order is positively correlated with the number of comparison images retrieved.

3. A computer-readable storage medium, characterized in that, The program algorithm is loaded to implement the fast indexing method for multi-center stored tongue diagnosis images as described in claim 1.

Images