WO2017000109A1 - Procédé de recherche, appareil de recherche, équipement utilisateur, et produit de programme informatique - Google Patents

Procédé de recherche, appareil de recherche, équipement utilisateur, et produit de programme informatique Download PDF

Info

Publication number
WO2017000109A1
WO2017000109A1 PCT/CN2015/082628 CN2015082628W WO2017000109A1 WO 2017000109 A1 WO2017000109 A1 WO 2017000109A1 CN 2015082628 W CN2015082628 W CN 2015082628W WO 2017000109 A1 WO2017000109 A1 WO 2017000109A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
target
character information
target image
information
Prior art date
Application number
PCT/CN2015/082628
Other languages
English (en)
Chinese (zh)
Inventor
姚聪
周舒畅
周昕宇
吴育昕
Original Assignee
北京旷视科技有限公司
北京小孔科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 北京旷视科技有限公司, 北京小孔科技有限公司 filed Critical 北京旷视科技有限公司
Priority to PCT/CN2015/082628 priority Critical patent/WO2017000109A1/fr
Priority to CN201580000313.XA priority patent/CN105518678B/zh
Publication of WO2017000109A1 publication Critical patent/WO2017000109A1/fr

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/951—Indexing; Web crawling techniques

Definitions

  • the present disclosure relates to the field of information technology, and more particularly, to a search method, a search device, a user equipment, and a computer program product.
  • the Internet provides a new trading platform and entertainment platform. For example, you can buy goods on the Internet, download music, watch videos online, and more.
  • the search can be performed on the Internet based on the keywords of the object to be searched.
  • the keyword-based object search system relies on the character representation entered by the user. However, when the keyword input by the user is inaccurate or there is an error, it is difficult to obtain a satisfactory search result.
  • e-services such as e-commerce continues to expand and the number and variety of goods or services grows rapidly, consumers may need to spend more time browsing to find objects or products of interest.
  • Embodiments of the present disclosure provide a search method, a search device, a user device, and a computer program product, which enable accurate and convenient searching of related object information of a target object, thereby improving a user's use experience.
  • a search method which is applied to a server, the search method may include: receiving a search request, the search request including a target image of a target object to be searched; and extracting the target from the target image Character information and image features associated with the object; searching for related object information associated with the target object based on the character information and the image feature; transmitting the related object information.
  • the extracting the character information and the image feature associated with the target object from the target image may include: utilizing optical character recognition The OCR identifies a character and a symbol from the target image; and selects an identification character for identifying the target object from the recognized characters and symbols as character information associated with the target object.
  • the searching for related object information associated with the target object based on the character information and the image feature may include: The character information and the image feature search for the related object information from a pre-established object database, wherein the object database includes image features, character information, and associated information of each candidate object.
  • the searching for the related object information from the pre-established object database based on the character information and the image feature may include: Calculating image feature similarity between the target object and each candidate object by image features of the target image and image features of the respective candidate objects; calculating the target object based on character information of the target image and character information of each candidate object Character information similarity with each candidate object; searching for related object information associated with the target object from the plurality of candidate objects based on the image feature similarity and the character information similarity.
  • the searching for the location and the location information based on the image feature similarity and the character information similarity may include: performing weighted averaging on the image feature similarity and the character information similarity to obtain an average similarity between the target object and each candidate object; according to the average similarity
  • the descending order of degrees selects a predetermined number of candidate objects from the plurality of candidate objects; information corresponding to the selected candidate objects is used as related object information associated with the target object.
  • the image feature based on the target image and the image feature of each candidate object are used to calculate between the target object and each candidate object.
  • the image feature similarity may include calculating a cosine similarity between the image feature of the target object and an image feature between the respective candidate objects as the image feature similarity.
  • the character information based on the target image and the character information of each candidate object are calculated between the target object and each candidate object.
  • the character information similarity may include: calculating an edit distance between the character information of the target object and the character information of each candidate object; based on the edit distance, the length of the character information of the target object, the character of the candidate object The length of the information is used to calculate the similarity of the character information.
  • the extracting the character information and the image feature associated with the target object from the target image may include at least one of the following operations One: calculating a color histogram feature of the target image as the image feature; and calculating a word bag model feature of the target image as the image feature.
  • the target image may satisfy a predetermined condition.
  • a search method for application to a user equipment.
  • the search method may include: collecting a target image of the target object to be searched; determining whether the target image satisfies a predetermined condition; and when the target image satisfies a predetermined condition, issuing a search request for the target object, the search request including The target image; receiving related object information associated with the target object, wherein the related object information is obtained based on character information and image feature search associated with the target object extracted from the target image.
  • the determining whether the target image meets the predetermined condition may include: determining an illumination parameter in the process of acquiring the target image; and when the illumination parameter is greater than or equal to When the illuminance is preset, it is determined that the target image satisfies a predetermined condition.
  • the determining whether the target image meets a predetermined condition may include: determining an average gradient of pixel points of an edge of the collected target image; When the average gradient of the pixel points of the edge of the target image is less than the preset gradient threshold, it is determined that the target image satisfies a predetermined condition.
  • a search device for use in a server.
  • the search device can include a transceiver that receives a search request, the search request including a target image of a target object to be searched, a processor, a memory, and computer program instructions stored in the memory. Performing the steps of: extracting, from the target image, character information and image features associated with the target object when the computer program instructions are executed by the processor; searching and said based on the character information and image features Relevant object information associated with the target object; the searched related object information is provided to the transceiver for transmission.
  • the extracting the character information and the image feature associated with the target object from the target image may include: using the optical character recognition OCR from the target Identifying characters and symbols in the image; selecting an identification character for identifying the target object from the recognized characters and symbols as character information associated with the target object.
  • Searching for related object information associated with the target object based on the character information and the image feature may include: searching for the related object information from a pre-established object database based on the character information and the image feature, wherein The object database includes image features, character information, and associated information of each candidate object.
  • the searching for the related object information from the pre-established object database based on the character information and the image feature may include: Calculating image feature similarity between the target object and each candidate object by image features of the target image and image features of the respective candidate objects; calculating the target object based on character information of the target image and character information of each candidate object Character information similarity with each candidate object; searching for related object information associated with the target object from the plurality of candidate objects based on the image feature similarity and the character information similarity.
  • the related object information associated with the target object may include: performing weighted averaging on the image feature similarity and the character information similarity to obtain an average similarity between the target object and each candidate object; according to the average similarity
  • the descending order of degrees selects a predetermined number of candidate objects from the plurality of candidate objects; information corresponding to the selected candidate objects is used as related object information associated with the target object.
  • the image feature based on the target image and the image feature of each candidate object are calculated between the target object and each candidate object.
  • the image feature similarity may include calculating a cosine similarity between the image feature of the target object and an image feature between the respective candidate objects as the image feature similarity.
  • the character information based on the target image and the character information of each candidate object are calculated between the target object and each candidate object.
  • the character information similarity may include: calculating an edit distance between the character information of the target object and the character information of each candidate object; based on the edit distance, the length of the character information of the target object, and the candidate objects The length of the character information is used to calculate the similarity of the character information.
  • extracting the character information and the image feature associated with the target object from the target image may include at least one of the following operations: Calculating a color histogram feature of the target image as the image feature; And calculating a word bag model feature of the target image as the image feature.
  • the target image may satisfy a predetermined condition.
  • a user equipment may include: an image collector for acquiring a target image of the target object to be searched; a processor for determining whether the target image satisfies a predetermined condition; and a transceiver, when the target image satisfies a predetermined condition, The issuing a search request for the target object, the search request including the target image, and receiving related object information associated with the target object, wherein the related object information is based on extraction from the target image Character information and image feature search associated with the target object are obtained.
  • the user equipment may further include an illuminometer for measuring an illumination parameter of the target object
  • the processor may instruct the illuminometer to be in image collection
  • the illumination parameter of the target object is measured during the process of acquiring the target image, and determining that the target image satisfies a predetermined condition when the illumination parameter is greater than or equal to the preset illumination.
  • the processor analyzes the target image to determine an average gradient of pixel points of the edge thereof, and a pixel point at an edge of the target image When the average gradient is less than the preset gradient threshold, it is determined that the target image satisfies a predetermined condition.
  • a computer program product for searching for an object can include a computer readable storage medium. Storing computer program instructions on the computer readable storage medium, the computer program instructions being executed by a processor to cause the processor to: receive a search request, the search request including a target image of a target object to be searched; Extracting character information and image features associated with the target object in the target image; searching for related object information associated with the target object based on the character information and the image feature; and transmitting the related object information.
  • a computer program product for searching for an object can include a computer readable storage medium.
  • Computer program instructions are stored on the computer readable storage medium.
  • the computer program instructions may be executed by a processor to cause the processor to: acquire an object image of a target object to be searched using an image collector; determine whether the target image satisfies a predetermined condition; and when the target image satisfies a predetermined condition Transmitting, by the transceiver, a search request for the target object, the search request including the target image; and receiving, by the transceiver, related object information associated with the target object, wherein the related object information is based on the The character information and image feature associated with the target object extracted in the target image are searched.
  • character information and image features associated with the target object are extracted from a target image of a target object to be searched for
  • the search is performed based on the character information and the image feature, and the related object information of the target object can be searched accurately and conveniently, thereby improving the user experience.
  • a search request is issued based on the target image, so that It can accurately and conveniently search related object information of the target object, thereby improving the user experience.
  • FIG. 1(a) schematically illustrates an application scenario according to an embodiment of the present disclosure
  • Figure 1 (b) schematically illustrates a schematic diagram of a target image taken by a user device
  • FIG. 2 is a flow chart that schematically illustrates a search method for a server in accordance with an embodiment of the present disclosure
  • FIG. 3 is a flow chart schematically illustrating related object information of a target object based on image features and character information in the search method of FIG. 2;
  • FIG. 4 is a flow chart that schematically illustrates a search method for a user equipment in accordance with an embodiment of the present disclosure
  • FIG. 5 is a block diagram schematically illustrating a first search device according to an embodiment of the present disclosure
  • FIG. 6 is a block diagram schematically illustrating a second search device for a server according to an embodiment of the present disclosure
  • FIG. 7 is a block diagram schematically illustrating a user equipment in accordance with an embodiment of the present disclosure.
  • FIG. 1(a) schematically illustrates an application scenario in accordance with an embodiment of the present disclosure.
  • the user equipment 10 is communicatively coupled to the search server 20 via a network.
  • the user device 10 is, for example, a smart phone, a tablet computer, a notebook computer, or the like.
  • the search server 20 is a cloud server, a web server, or the like. Communication between user device 10 and search server 20 may be implemented using a variety of techniques including, but not limited to, the Internet, local area networks, third generation mobile communication technologies, and the like. For example, a user of a user device browses a Taobao web page to expect to purchase a particular item, ie, a target object.
  • the user equipment is connected to the search server of Taobao via the Internet.
  • the user inputs a keyword of the product to be purchased in the Taobao webpage of the user equipment, and the user equipment transmits the keyword to the search server of Taobao via the Internet, and the latter performs a search based on the keyword and via the Internet. Sending the search result to the user equipment.
  • the keyword input by the user is inaccurate or there is an error, it is difficult to obtain a satisfactory search result.
  • a plurality of items associated with keywords may be included in the search results, which may make it impossible for the user to find the target object to be purchased from the search results.
  • the user equipment 10 performs image acquisition on a target object to be purchased using a camera or the like, and transmits the collected target image to the search server 20.
  • the search server 20 extracts character information and/or image information from the target image, and performs a search based on the extracted information, and transmits the search result to the user device via the Internet.
  • the target image it usually carries rich information about the target object, such as the appearance, name, trademark, manufacturer, date of manufacture, and the like of the target object.
  • the search server can more accurately search for the target object of the user, thereby improving the accuracy of the search.
  • the search server can automatically extract information in the target image without requiring the user to manually input keywords or the like, which makes the user's search operation more convenient.
  • FIG. 1(b) schematically illustrates a schematic diagram of a target image taken by the user device 10.
  • the captured target images (1), (2), and (3) are respectively Evian mineral water, calbee potato chips, and blue moon laundry liquid.
  • the target image (1) includes information on the appearance of the shape of the bottled water, the name of the evian, the shape of the mountain, the capacity of 550 ml, etc., based on the information, the search server 20 can accurately search for the target object of the user. However, if the user enters the keyword "Evian mineral water", it will search for Evian mineral water in different packaging, different series and different capacities.
  • the search server 20 can accurately search for each target object.
  • the search method 200 is applicable to a search server as shown in FIG. 1(a).
  • the search method 200 may include receiving a search request including a target image of a target object to be searched (S210); extracting characters associated with the target object from the target image Information and image features (S220); searching for related object information associated with the target object based on the character information and the image feature (S230); transmitting the related object information (S240).
  • the server receives a search request from the user device, the search request including a target image of the target object to be searched.
  • the target image is any one of the target images as shown in FIG. 1(b).
  • the target image contains various information of the target object to be searched, including but not limited to brand name, object content, series, appearance, capacity, production date, and the like.
  • the target image may be collected by the user equipment by using the image collection device, or may be received by the user equipment from other electronic devices.
  • the manner in which the user equipment acquires the target image does not constitute a limitation on the embodiments of the present disclosure.
  • the server extracts information from the target image to search for a target object. Accordingly, the image quality of the target image will directly affect the search results. For example, in the target image (1) of FIG. 1(b), if the target image is blurred and information such as the brand name evian, capacity, and the like cannot be extracted, it is difficult to accurately search for the target object. Therefore, a request can be made for the target image, for example, the target image satisfies a predetermined condition.
  • the predetermined condition may be a condition regarding the brightness of the target image or a condition regarding the sharpness of the target image.
  • the target image when the brightness of the target image is greater than or equal to the preset brightness threshold, it is determined that the target image satisfies a predetermined condition; when the brightness of the target image is less than the preset brightness threshold, it is determined that the target image does not satisfy the predetermined condition.
  • the sharpness of the target image when the sharpness of the target image is greater than or equal to the preset sharpness threshold, it is determined that the target image satisfies a predetermined condition; when the brightness of the target image is less than the preset sharpness threshold, it is determined that the target image does not satisfy the predetermined condition.
  • the preset brightness threshold or the preset definition threshold may be adjusted according to the processing capability of the server.
  • the preset brightness threshold or the preset definition threshold may be set to a lower value; when the processing capability of the server is weak, the preset brightness threshold or the preset definition threshold may be set. Is a higher value.
  • character information associated with the target object is extracted from the target image and Image features.
  • the character information included in the target image is, for example, a product name, a capacity, a brand name, a date of manufacture, and the like, and the character information is a character or a symbol.
  • the image special diagnosis included in the target image is a color component of the image, a composition ratio of each color component, and the like. Typically, different techniques are employed to extract character information and image features in the target image.
  • the character information in the target image can be extracted by using Optical Character Recognition (OCR) technology.
  • OCR Optical Character Recognition
  • the server determines its shape by detecting the dark and bright patterns of the target image, and then uses the character recognition method to translate the shape into computer text.
  • other techniques may be employed to perform character recognition on the target image to obtain character information therein.
  • Character information associated with the target object may be extracted from the target image by recognizing words and symbols from the target image using optical character recognition OCR; selecting for identification from the identified characters and symbols
  • the identification character of the target object is the character information associated with the target object.
  • rich information is included in the target image, and some of the information may be closely related to the search of the target object, such as product name, brand, capacity, and the like.
  • the target image may also include information that is not related to the search of the target object, such as components, security reminders, etc., which may be information related to all similar products, which cannot be used to identify the target object. Therefore, after performing character recognition on the target image, it is necessary to filter out information required for searching the target object, that is, an identification character for identifying the target object.
  • Image features are index-valued image feature representations, such as using vectors to represent image characteristics.
  • the image features of the target image may be represented in various ways that are present or appearing in the future.
  • a color histogram and a Bag of Words feature are taken as an example of an image feature. It is to be noted that, in the application, any one of the color histogram and the word bag model feature may be used to represent the image feature of the target image, and both the color histogram and the bag model feature may be used to represent the image feature of the target image.
  • the extracting the image object associated with the target object from the target image includes at least one of: calculating a color histogram feature of the target image as the image feature; and calculating a location
  • the word bag model feature of the target image is used as the image feature.
  • a color histogram is a statistical representation of the color characteristics of an image that is used to represent the proportion of different colors in the entire target image, without concern for the spatial location of each color.
  • Color histograms are closely related to how color space is represented. Common color histograms include RGB spatial color histograms, HSV spatial color histograms, and Lab space color histograms. In different color spaces The color histogram of the target image has different values.
  • the word bag model feature is a statistical representation of the texture features of an image that can effectively describe the overall and local characteristics of the image.
  • the word bag model feature of the target image can be obtained by extracting feature descriptors from the target image, such as Scale Invariant Feature Transform (SIFT), Directional Histogram (HOG, Histogram of Oriented Gradient). ); for each descriptor, search for the most similar cluster center in the pre-accurate codebook, and count the frequency of occurrence of different cluster centers in the target image to form a histogram; The processing is performed to obtain the word bag model feature of the target image.
  • SIFT Scale Invariant Feature Transform
  • HOG Directional Histogram of Oriented Gradient
  • the pre-accurate codebook can be obtained by randomly extracting a large number of image descriptors (for example, SIFT, HOG, etc.) from the set of training images, and clustering the image descriptors by using a clustering algorithm to obtain multiple Category, all the categories obtained by clustering constitute the codebook.
  • image descriptors for example, SIFT, HOG, etc.
  • related object information associated with the target object is searched based on the character information and the image feature obtained in S220. Specifically, the related object information is searched from a pre-established object database based on the character information and the image feature.
  • the object database includes image features, character information, and associated information of respective candidate objects.
  • f I (p j ) represents an image feature of the object p j , which may be a color histogram feature, or a bag model feature, or a vector composed of color histogram features and word bag model features.
  • f T (p j ) is character information of the object p j , which is typically a character string such as a name, a brand, a content, and the like.
  • a(p j ) represents other associated information associated with the object p j , such as price, sales volume, user rating, promotional video, and hyperlinks.
  • each object p j can also be represented by a binary group ⁇ f I (p j ), f T (p j ) ⁇ .
  • the image feature and the character information of the target image of the target object q to be searched are f I (q) and f T (q), respectively, and accordingly, the character information f T (q) and the image obtained in S220 can be obtained.
  • the feature f I (q) is compared with the character information f T (p j ) of each candidate object in the object database P and the image feature f I (p j ) to perform a search.
  • FIG. 3 is a flowchart schematically illustrating related object information (S230) of searching for a target object based on image features and character information in the search method of FIG. 2.
  • image feature similarity between the target object and each candidate object is calculated based on image features of the target image and image features of the respective candidate objects (S231); character information based on the target image and each device Selecting character information of the object to calculate a character information similarity between the target object and each candidate object (S232); performing weighted averaging on the image feature similarity and the character information similarity to obtain the target object and each device Selecting an average similarity between the objects (S233); selecting a predetermined number of candidate objects from the plurality of candidate objects in descending order of the average similarity (S234); corresponding to the selected candidate objects
  • the information is related object information associated with the target object (S235).
  • a cosine similarity s I (q, p j ) between the image feature f I (q) of the target object q and the image feature f I (p j ) between the respective candidate objects p j can be calculated. ) as the image feature similarity.
  • the cosine similarity s I (q, p j ) can be calculated by the following formula (1):
  • the character information similarity between the target object q and each candidate object p j may be calculated as follows: calculating character information f T (q) of the target object q and each candidate object p j An edit distance d(f T (q), f T (p j )) between the character information f T (p j ); based on the edit distance, the length of the character information f T (q) of the target object, The character information similarity is calculated by the length of the character information f T (p j ) of the candidate object.
  • Edit distance is the minimum number of edit operations required to convert from one string to another between two strings. The allowed editing operations include replacing one character with another, inserting a character, and deleting One character.
  • the edit distance d(f T (q), f T (p j )) is the minimum number of editing operations required to convert the character information f T (q) into the character information f T (p j ).
  • Character information length f T (q) for example, the number of characters and symbols included in the character information f T (q) in.
  • the length of the object character information f T (p j) for example, the number of characters and symbols included in the character information f T (p j) in.
  • the character information similarity s T (q, p j ) can be calculated by the following formula (2):
  • d(f T (q), f T (p j )) is the edit distance between the character information f T (q) and the character information f T (p j ), and L(f T (q)) is a character
  • the length of the information f T (q), L(f T (p j )), is the length of the character information f T (p j ).
  • the image feature similarity s I (q, p j ) and the character information similarity s T (q, p j ) are weighted and averaged to obtain an average between the target object and each candidate object. Similarity.
  • the average similarity s(q, p j ) can be calculated by the following formula (3):
  • ⁇ is the weight coefficient.
  • the image feature similarity s I (q, p j ) increases in the average similarity
  • the character information similarity s T (q, p j ) decreases in the average similarity.
  • the weight coefficient ⁇ decreases, the image feature similarity s I (q, p j ) decreases in the average similarity, and the character information similarity s T (q, p j ) increases in the average similarity.
  • a predetermined number of candidate objects are selected from the plurality of candidate objects in descending order of the average similarity s(q, p j ).
  • the average similarities may be arranged in descending order, and for example, a predetermined number of R candidate objects with an average degree of similarity are selected, and the R candidate objects are search results.
  • the average similarity between the R candidate objects and the target object is high, indicating that the R candidate objects are closer to the target object, so that there is a larger target object that the user desires.
  • R is a configurable parameter, and its typical value can be set to 10, 20, 100, and so on.
  • information corresponding to the selected R candidate objects is taken as related object information associated with the target object.
  • the picture, the character description, the related information, and the like of the R objects are used as related object information.
  • the related information is, for example, price, sales volume, user rating, promotional video, hyperlink, and the like.
  • related object information associated with the target object is searched from the plurality of candidate objects based on the image feature similarity and the character information similarity.
  • the related object information may be searched in such a manner that R1 candidate objects are selected from the plurality of candidate objects in descending order of image feature similarity; in descending order of similarity of character information R2 candidate objects are selected from the plurality of candidate objects; information corresponding to the selected R1 candidate objects and R2 candidate objects is used as related object information associated with the target object.
  • R1 is a natural number smaller than N.
  • R2 is also a natural number smaller than N.
  • the server sends the searched related object information as a search result to the user equipment.
  • the server can transmit the correlation by using various networks or communication technologies such as the Internet and a local area network.
  • Object information is, for example, a picture, a text description, and associated information of the R candidate objects, or a picture, a text description, and associated information of the R1 plus R2 candidate object described above.
  • the user equipment may display the related object information on the screen of the user equipment for the user to view.
  • character information and image features associated with the target object are extracted from a target image of a target object to be searched based on the character information and
  • the image feature performs a search, and can accurately and conveniently search related object information of the target object, thereby improving the user experience.
  • the step of manually inputting a keyword by the user is eliminated by automatically recognizing the character information contained in the target image.
  • the search method 400 may include: collecting a target image of a target object to be searched (S410); determining whether the target image satisfies a predetermined condition (S420); and when the target image satisfies a predetermined condition, issuing a search request for the target object, the search request including the target image (S430); receiving related object information associated with the target object (S440), wherein the related object information is based on from the target image
  • S410 target image of a target object to be searched
  • S420 determining whether the target image satisfies a predetermined condition
  • S440 receiving related object information associated with the target object
  • the related object information is based on from the target image
  • the extracted character information and image feature search associated with the target object are obtained.
  • the image capturing device in the user device may be utilized to collect the target image of the target object to be searched. For example, if the blue moon laundry liquid in the user's FIG. 1(b) is exhausted and it is desired to purchase the blue moon laundry liquid, the user utilizes an image capturing device built in the user device 10 or an image acquisition connected to the user device. The device performs image acquisition on the existing Blue Moon laundry detergent. The positional relationship between the image capture device and the user equipment does not constitute a limitation on the embodiments of the present disclosure.
  • the server is to extract information from the target image to search for the target object, the image quality of the target image will directly affect the search result.
  • the target image (1) of FIG. 1(b) as an example, if the target image is blurred and information such as the brand name evian, capacity, etc. cannot be extracted, it is difficult to accurately search for the target object.
  • a requirement may be made for the target image at S420, for example, the target image satisfies a predetermined condition.
  • the predetermined condition may be a condition regarding the brightness of the target image or a condition regarding the sharpness of the target image.
  • the target image acquired in S410 may be converted into image data of the HSL color space in which the luminance information is included in the image data of the HSL color space. Then, the average value of the illumination components (ie, L components) of all pixels in the image data of the HSL color space is counted Average value of the illumination component used for the target image When the predetermined brightness threshold T L is greater than or equal to, it may be determined that the target image satisfies a predetermined condition. Average value of the illumination component used for the target image When it is less than the predetermined brightness threshold T L , it may be judged that the target image does not satisfy the predetermined condition.
  • the predetermined brightness threshold T L is typically 64.
  • the quality of the target image can be indirectly determined by measuring the lighting conditions in the image acquisition environment. For example, the illumination parameter in the process of acquiring the target image may be determined; when the illumination parameter is greater than or equal to the preset illumination, determining that the target image satisfies a predetermined condition; when the illumination parameter is less than the preset illumination, determining the location The target image does not satisfy the predetermined condition.
  • an edge of the target image acquired in S410 may be extracted using a predetermined algorithm (for example, Canny algorithm) in S420, and each of the edges located in the target image is calculated.
  • the gradient G of the pixel and then further calculate the average of the gradients of all the pixel points at the edge in the target image.
  • the average of the gradients of all pixel points at the edge of the target image When it is greater than or equal to the preset gradient threshold T G , it may be determined that the target image satisfies a predetermined condition.
  • the average of the gradients of all pixel points at the edge of the target image When it is less than the preset gradient threshold T G , it may be determined that the target image does not satisfy the predetermined condition.
  • the preset gradient threshold T G is typically 100.
  • the predetermined brightness threshold T L or the preset gradient threshold T G described above may be adjusted according to the processing capability of the server performing the search. For example, when the processing capability of the server is strong, the predetermined brightness threshold T L or the preset gradient threshold T G may be set to a lower value; when the processing capability of the server is weak, the predetermined brightness threshold T L or a preset gradient may be used. The threshold T G is set to a higher value.
  • a search request for the target object is issued in S430, the search request including the target image.
  • the search server 20 extracts character information and image features associated with the target object from the target image, and performs a search based on the character information and the image feature, ie, The various steps of the search method described in connection with FIG. 2 are performed. Since the brightness or sharpness of the target image is good, the character information and the image feature can be accurately extracted in the server, thereby ensuring the accuracy of the search.
  • a retake prompt message may be output in the user equipment to prompt re-execution S410 to collect the target image of the target object to be searched.
  • the retake prompt message it is also possible to specifically list the reason why the target image does not satisfy the predetermined condition.
  • the average of the illumination components of the target image When less than the predetermined brightness threshold T L , the brightness may be indicated in the replay prompt message; the average of the gradients of all the pixel points located at the edge of the target image
  • the preset gradient threshold T G is smaller than the preset gradient threshold T G , it is indicated that the sharpness is insufficient in the replay prompt message.
  • the shooting of the target image can be adjusted according to the replay prompt message until the target image that satisfies the predetermined condition is acquired.
  • the setting parameters of the image pickup device may be automatically adjusted in accordance with the determination result of S420 until the target image satisfying the predetermined condition is acquired.
  • the server After the user device issues a search request to the server in S430, the server performs the search method described in connection with FIGS. 2 and 3 and obtains related object information associated with the target object. That is, the related object information is obtained based on character information and image feature search associated with the target object extracted from the target image.
  • the user equipment receives relevant object information associated with the target object in S440.
  • the user equipment can receive the related object information from the server by using various networks or communication technologies such as the Internet and a local area network.
  • the related object information is, for example, a picture of a plurality of candidate objects, a text description, and associated information.
  • the associated information is, for example, price, sales volume, user rating, promotional video, hyperlinks, etc., which assists the user in performing selection operations among a plurality of candidate objects.
  • the user equipment may display the related object information on the screen of the user equipment for the user to view.
  • the user device can automatically calculate the illumination condition and the degree of clarity of the image. If the lighting conditions and clarity of the image meet the requirements, the user device is allowed to issue a search request based on the acquired target image. If the lighting conditions and clarity of the image do not meet the requirements, the user equipment is prompted or automatically instructed to re-shoot until the desired target image is obtained.
  • a search request is issued based on the target image, so that the target object can be searched accurately and conveniently Relevant object information, thereby improving the user experience.
  • FIG. 5 is a block diagram schematically illustrating a first search device 500 in accordance with an embodiment of the present disclosure.
  • the first search device 500 can be applied to a user equipment or server.
  • the first data processing apparatus 500 may include one or more processors 510, a storage unit 520, an input unit 530, an output unit 540, a communication unit 550, and an image acquisition unit 560. These components are interconnected by a bus system 570 and/or other form of connection mechanism (not shown).
  • the first search shown in Figure 5 The components and structures of the device 500 are merely exemplary and not limiting.
  • the first search device 500 may also have other components and structures as needed, and may, for example, not include the input unit 530, the output unit 540, and the image acquisition unit 560. Wait.
  • Processor 510 can be a central processing unit (CPU) or other form of processing unit with data processing capabilities and/or instruction execution capabilities, and can control other components in first search device 500 to perform desired functions.
  • CPU central processing unit
  • Processor 510 can be a central processing unit (CPU) or other form of processing unit with data processing capabilities and/or instruction execution capabilities, and can control other components in first search device 500 to perform desired functions.
  • Storage unit 520 can include one or more computer program products, which can include various forms of computer readable storage media, such as volatile memory and/or nonvolatile memory.
  • the volatile memory may include, for example, a random access memory (RAM) and/or a cache or the like.
  • the nonvolatile memory may include, for example, a read only memory (ROM), a hard disk, a flash memory, or the like.
  • One or more computer program instructions may be stored on the computer readable storage medium, and the processor 510 may execute the program instructions to implement various of the search methods described above in connection with FIGS. 2 and 3 of embodiments of the present disclosure. Step, at this time, the first search device 500 can be included in the server.
  • the processor 510 can execute the program instructions to implement the various steps of the search method described above in connection with FIG. 4 of the embodiments of the present disclosure, at which time the first search device 500 can be included in the user equipment.
  • Various applications and various data such as an operating state of the display screen, an operational state of the application, and the like can also be stored in the computer readable storage medium.
  • the input unit 530 may be a unit used by a user to input an instruction, and may include one or more of a keyboard, a mouse, a microphone, a touch screen, and the like.
  • the output unit 540 may output various information (such as an image or a sound) to the outside (for example, a user), and may include one or more of a display, a speaker, and the like.
  • Communication unit 550 can communicate with other units (e.g., personal computers, servers, mobile stations, base stations, etc.) via a network or other technology, which can be the Internet, a wireless local area network, a mobile communication network, and the like.
  • the character information and the image feature associated with the target object are extracted from the target image of the target object to be searched, based on the character information and the image feature.
  • the search can accurately and conveniently search for related object information of the target object, thereby improving the user experience.
  • FIG. 6 is a block diagram schematically illustrating a second search device 600 for a server in accordance with an embodiment of the present disclosure.
  • the second search device 600 is applicable to a search server as shown in FIG. 1(a).
  • the second search device 600 may include a first receiving unit 610, an extracting unit 620, and a search. Unit 630 and first transmitting unit 640.
  • the first receiving unit 610 receives a search request including a target image of a target object to be searched for.
  • the target image is any one of the target images as shown in FIG. 1(b).
  • the target image contains various information of the target object to be searched, including but not limited to brand name, object content, series, appearance, capacity, production date, and the like.
  • the target image may be collected by the user equipment by using the image collection device, or may be received by the user equipment from other electronic devices, and the manner in which the target image is acquired does not constitute a limitation on the embodiments of the present disclosure.
  • the first receiving unit 610 corresponds to the communication unit 550 in FIG. 5 and can be implemented by using a radio frequency circuit and a signal receiving circuit.
  • the target image preferably satisfies a predetermined condition.
  • the predetermined condition may be a condition regarding the brightness of the target image or a condition regarding the sharpness of the target image.
  • the sharpness of the target image is greater than or equal to the preset sharpness threshold, it is determined that the target image satisfies a predetermined condition; when the brightness of the target image is less than the preset sharpness threshold, it is determined that the target image does not satisfy the predetermined condition.
  • the preset brightness threshold or the preset definition threshold may be adjusted according to the processing capability of the server. For example, when the processing capability of the server is strong, the preset brightness threshold or the preset definition threshold may be set to a lower value; when the processing capability of the server is weak, the preset brightness threshold or the preset definition threshold may be set. Is a higher value.
  • the extracting unit 620 extracts character information and image features associated with the target object from the target image.
  • the character information included in the target image is, for example, a product name, a capacity, a brand name, a date of manufacture, and the like, and the character information is a character or a symbol.
  • the image special diagnosis included in the target image is a color component of the image, a composition ratio of each color component, and the like. Typically, different techniques are employed to extract character information and image features in the target image.
  • Extraction unit 620 can be implemented using the memory and processor of FIG.
  • the extracting unit 620 may extract the character information in the target image using OCR technology or other techniques.
  • OCR technology the server determines its shape by detecting the dark and bright patterns of the target image, and then uses the character recognition method to translate the shape into computer text.
  • the extracting unit 620 may include an OCR module, and may extract the target pair from the target image by the following operation
  • OCR module may extract the target pair from the target image by the following operation
  • Linked character information Rich information is included in the target image, and some of the information may be closely related to the search of the target object, such as product name, brand, capacity, and the like.
  • the target image may also include information that is not related to the search of the target object, such as components, security reminders, etc., which may be information related to all similar products, which cannot be used to identify the target object. Therefore, after performing the character recognition on the target image, the extracting unit 620 needs to filter out information required for searching the target object, that is, the identification character for identifying the target object.
  • Image features are index-valued image feature representations, such as using vectors to represent image characteristics.
  • the image features of the target image may be represented in various ways that are present or appearing in the future.
  • the extracting unit 620 may include an image feature extraction module, and the image feature extraction module may perform at least one of: extracting an image feature: calculating a color histogram feature of the target image as the image feature; and calculating the image
  • the word bag model feature of the target image is used as the image feature. That is, the extracting unit 620 may represent the image features of the target image using at least one of a color histogram and a bag model feature.
  • a color histogram is a statistical representation of the color characteristics of an image that is used to represent the proportion of different colors in the entire target image, without concern for the spatial location of each color. Common color histograms include RGB spatial color histograms, HSV spatial color histograms, and Lab space color histograms. In different color spaces, the color histogram of the target image has different values.
  • the word bag model feature is a statistical representation of the texture features of an image that can effectively describe the overall and local characteristics of the image.
  • the extracting unit 620 can obtain the word bag model feature of the target image by extracting feature descriptors such as SIFT, HOG, etc.
  • the clustering center counts the frequency of occurrence of different clustering centers in the target image to form a histogram; normalizes the histogram to obtain the word bag model feature of the target image.
  • the pre-accurate codebook can be obtained by randomly extracting a large number of image descriptors from a set of training images, and clustering the image descriptors by using a clustering algorithm to obtain a plurality of categories, and all the clusters are obtained.
  • the category is the codebook.
  • the search unit 630 searches for related object information associated with the target object based on the character information and the image feature. For example, the search unit 630 searches the related object information from a pre-established object database based on the character information and the image feature.
  • each object p j can also be represented by a binary group ⁇ f I (p j ), f T (p j ) ⁇ . It is assumed that the image features and character information of the target image of the target object q to be searched are f I (q) and f T (q), respectively, and accordingly, the search unit 630 can pass the character information f T extracted by the extracting unit 620 ( q) The image feature f I (q) is compared with the character information f T (p j ) of each candidate object in the object database P and the image feature f I (p j ) to perform a search. Search unit 630 can be implemented using the memory and processor of FIG.
  • the searching unit 630 is operable to search for related object information associated with the target object: calculating image features similar to the target object and each candidate object based on the image features of the target image and the image features of the respective candidate objects Calculating a similarity of character information between the target object and each candidate object based on the character information of the target image and the character information of each candidate object; and based on the image feature similarity and the character information similarity Searching for related object information associated with the target object from the plurality of candidate objects.
  • the search unit 630 may calculate a cosine similarity s I (q, between the image feature f I (q) of the target object q and the image feature f I (p j ) between the respective candidate objects p j , p j ) as the image feature similarity.
  • the search unit 630 can calculate the cosine similarity s I (q, p j ) according to the above formula (1), and can be specifically referred to the description above in connection with the formula (1).
  • the search unit 630 may also take the Pearson correlation coefficient between the image feature of the target object and the image feature between the respective candidate objects as the image feature similarity.
  • the search unit 630 may be calculated as the similarity of the character information: calculating the edit distance d between the target object character information q f T (q) of each candidate character information of the object p j f T (p j) (f T (q), f T (p j )); based on the edit distance, the length of the character information f T (q) of the target object, the length of the character information f T (p j ) of the candidate object To calculate the similarity of the character information.
  • the edit distance d(f T (q), f T (p j )) is the minimum number of editing operations required to convert the character information f T (q) into the character information f T (p j ).
  • the search unit 630 can calculate the character information similarity s T (q, p j ), for example, by the above formula (2). Alternatively, the search unit 630 may also use the edit distance d(f T (q), f T (p j )) as the character information similarity.
  • the searching unit 630 may search for related object information by selecting R1 candidate objects from the plurality of candidate objects in descending order of image feature similarity; from the plurality of devices in descending order of character information similarity R2 candidate objects are selected among the selected objects; information corresponding to the selected R1 candidate objects and R2 candidate objects is used as related object information associated with the target object.
  • R1 is a natural number smaller than N.
  • R2 is also a natural number smaller than N.
  • the searching unit 630 may further search for related object information based on the image feature similarity and the character information similarity in such a manner that the image feature similarity and the character information similarity are weighted and averaged to obtain the target.
  • R is a configurable parameter, and its typical value can be set to 10, 20, 100, and so on.
  • the first sending unit 640 sends the related object information, that is, the searched related object information is sent to the user equipment as a search result.
  • the first transmitting unit 640 can transmit the related object information by using various networks or communication technologies such as the Internet and a local area network.
  • the related object information is, for example, a picture, a text description, and associated information of the R candidate objects, or a picture, a text description, and associated information of the R1 plus R2 candidate object described above.
  • the user equipment may display the related object information on the screen of the user equipment for the user to view.
  • the first transmitting unit 640 may correspond to the communication unit 550 in FIG. 5 and may be implemented by using a radio frequency circuit and a signal transmitting circuit.
  • the character information and the image feature associated with the target object are extracted from the target image of the target object to be searched, based on the character
  • the information and image feature performs a search, and can accurately and conveniently search for related object information of the target object, thereby improving the user experience.
  • the step of manually inputting a keyword by the user is eliminated by automatically recognizing the character information contained in the target image.
  • FIG. 7 is a block diagram that schematically illustrates a user device 700 in accordance with an embodiment of the present disclosure.
  • the user equipment 700 corresponds to the user equipment shown in FIG. 1(a).
  • the user equipment 700 may include an image acquisition unit 710, a determination unit 720, a second transmission unit 730, and a second reception unit 740.
  • the image acquisition unit 710 collects a target image of the target object to be searched for.
  • Image acquisition unit 710 is typically disposed in the user device. For example, if the user's blue moon laundry liquid is exhausted and it is desired to purchase the blue moon laundry liquid, the user uses the image acquisition unit 710 to perform image acquisition on the existing blue moon laundry liquid.
  • the image acquisition unit 710 is illustrated as being included in the user equipment in FIG. 7, but the image acquisition unit 710 may also be external to the user equipment, coupled to the user equipment, and capable of receiving instructions of the user equipment, And transmitting the acquired target image to the user equipment.
  • the positional relationship between the image capture device and the user equipment does not constitute a limitation on the embodiments of the present disclosure.
  • the image acquisition unit 710 can be a camera, a camera, or the like.
  • the image acquisition unit 710 corresponds to the image acquisition unit 560 of FIG.
  • the judging unit 720 judges whether or not the target image satisfies a predetermined condition. Since the server is to extract information from the target image to search for the target object, the image quality of the target image will directly affect the search result.
  • the determining unit 720 can make a request for the target image using a predetermined condition.
  • the predetermined condition may be a condition regarding the brightness of the target image or a condition regarding the sharpness of the target image.
  • the determining unit 720 can be implemented using the memory and processor in FIG.
  • the determination unit 720 may convert the acquired target image into image data of an HSL color space in which the luminance information is included in the image data of the HSL color space. Then, the judging unit 720 counts the average value of the illumination components (ie, the L component) of all the pixels in the image data of the HSL color space. And comparing it to a predetermined brightness threshold T L . Average value of the illumination component used for the target image When the predetermined brightness threshold T L is greater than or equal to, the determination unit 720 may determine that the target image satisfies a predetermined condition.
  • the determination unit 720 may determine that the target image does not satisfy the predetermined condition.
  • the predetermined brightness threshold T L is typically 64.
  • the determining unit 720 can also indirectly determine the quality of the target image by measuring the lighting conditions in the image capturing environment by means of the illuminometer.
  • the user equipment 700 may further include an illuminometer 750 for measuring an illumination parameter of the target object, the determination unit 720 communicating with the illuminometer to determine an illumination parameter in the process of acquiring the target image; When the parameter is greater than or equal to the preset illuminance, it is determined that the target image satisfies a predetermined condition; when the illumination parameter is less than the preset illuminance, it is determined that the target image does not satisfy the predetermined condition.
  • the determination unit 720 may extract an edge of the acquired target image using a predetermined algorithm (for example, the Canny algorithm), and calculate the edge located in the target image.
  • the gradient G of each pixel and then further calculate the average of the gradients of all the pixel points at the edge in the target image
  • the average of the gradients of all pixel points at the edge of the target image When it is greater than or equal to the preset gradient threshold T G , the determining unit 720 may determine that the target image satisfies a predetermined condition.
  • the average of the gradients of all pixel points at the edge of the target image When it is less than the preset gradient threshold T G , the determination unit 720 may determine that the target image does not satisfy the predetermined condition.
  • the preset gradient threshold T G is typically 100.
  • the predetermined brightness threshold T L or the preset gradient threshold T G described above may be adjusted according to the processing capability of the server performing the search. For example, when the processing capability of the server is strong, the predetermined brightness threshold T L or the preset gradient threshold T G may be set to a lower value; when the processing capability of the server is weak, the predetermined brightness threshold T L or a preset gradient may be used. The threshold T G is set to a higher value.
  • the second transmitting unit 730 issues a search request for the target object when the target image satisfies a predetermined condition, the search request including the target image.
  • a search device as shown in FIG. 5 or FIG. 6 extracts character information and image features associated with the target object from the target image, and performs a search based on the character information and the image features. For example, after receiving the search request, the first receiving unit 610 in FIG.
  • the COR module extracts character information associated with the target object from the target image; the search unit 630 searches for the related object information associated with the target object from the object database based on the character information and the image feature;
  • the two transmitting unit 640 transmits the searched related object information to the user equipment. Since the brightness or sharpness of the target image is good, the character information and the image feature can be accurately extracted in the server, thereby ensuring the accuracy of the search.
  • the second transmitting unit 730 corresponds to the transceiver unit 550 in FIG. 5, and can be implemented by using a radio frequency circuit and a signal transmitting circuit.
  • the user equipment 700 may further include an output unit for outputting a retake prompt message to prompt the user to operate the image collection device to collect the target image of the target object to be searched.
  • a retake prompt message it is also possible to specifically list the reason why the target image does not satisfy the predetermined condition.
  • the average of the illumination components of the target image When less than the predetermined brightness threshold T L , it may be indicated that the brightness is insufficient in the re-scuing message; the average of the gradients of all the pixel points located at the edge of the target image
  • the preset gradient threshold T G is smaller than the preset gradient threshold T G , it is indicated that the sharpness is insufficient in the replay prompt message.
  • the shooting of the target image can be adjusted according to the replay prompt message until the target image that satisfies the predetermined condition is acquired.
  • the determination unit 720 determines that the target image does not satisfy the predetermined condition
  • the setting parameters of the image acquisition unit 710 may be automatically adjusted until the target image satisfying the predetermined condition is acquired.
  • the second receiving unit 740 receives related object information associated with the target object. After the second transmitting unit 730 issues a search request to the server, the server performs the search method described in connection with FIGS. 2 and 3, and obtains related object information associated with the target object. Correspondingly, the second receiving unit 740 receives related object information associated with the target object. The related object information is obtained based on character information and image feature search associated with the target object extracted from the target image. The second receiving unit 740 can receive the related object information from the server through various networks or communication technologies such as the Internet, a local area network, and the like.
  • the related object information is, for example, a picture of a plurality of candidate objects, a text description, and associated information.
  • the associated information is, for example, price, sales volume, user rating, promotional video, hyperlinks, etc., which assists the user in performing selection operations among a plurality of candidate objects.
  • the second receiving unit 740 may display the related object information on the screen of the user equipment for the user to view.
  • the second receiving unit 740 corresponds to the transceiver unit 550 in FIG. 5 and can be implemented by using a radio frequency circuit and a signal receiving circuit.
  • the user device can automatically calculate the illumination condition and the degree of clarity of the image. If the lighting conditions and clarity of the image meet the requirements, the user device is allowed to issue a search request based on the acquired target image. If the lighting conditions and clarity of the image do not meet the requirements, the user equipment is prompted or automatically instructed to re-shoot until the desired target image is obtained.
  • the search request is issued based on the target image, so that the related object information of the target object can be accurately and conveniently searched, Thereby improving the user experience.
  • an electronic device or server including any of the first search device and the second search device is also within the scope of the present disclosure.
  • the disclosed apparatus and method may be implemented in other manners.
  • the device embodiments described above are merely illustrative.
  • the division of the unit is only a logical function division.
  • there may be another division manner for example, multiple units or components may be combined or Can be integrated into another device, or some features can be ignored or not executed.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • Image Analysis (AREA)

Abstract

L'invention concerne un procédé de recherche, un appareil de recherche, un équipement utilisateur, et un produit de programme informatique. Le procédé de recherche utilisé pour un serveur consiste à : recevoir une demande de recherche, la demande de recherche comprenant une image cible d'un objet cible à rechercher (S210) ; extraire de l'image cible des informations sur des caractères et une caractéristique d'image associée à l'objet cible (S220) ; rechercher des informations d'objet associées à l'objet cible selon les informations des caractères et la caractéristique d'image (S230); et envoyer des informations d'objet associées (S240). Des informations d'objet associées d'un objet cible peuvent être recherchées avec précision et de manière pratique, ce qui permet d'améliorer l'expérience des utilisateurs.
PCT/CN2015/082628 2015-06-29 2015-06-29 Procédé de recherche, appareil de recherche, équipement utilisateur, et produit de programme informatique WO2017000109A1 (fr)

Priority Applications (2)

Application Number Priority Date Filing Date Title
PCT/CN2015/082628 WO2017000109A1 (fr) 2015-06-29 2015-06-29 Procédé de recherche, appareil de recherche, équipement utilisateur, et produit de programme informatique
CN201580000313.XA CN105518678B (zh) 2015-06-29 2015-06-29 搜索方法、搜索装置和用户设备

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2015/082628 WO2017000109A1 (fr) 2015-06-29 2015-06-29 Procédé de recherche, appareil de recherche, équipement utilisateur, et produit de programme informatique

Publications (1)

Publication Number Publication Date
WO2017000109A1 true WO2017000109A1 (fr) 2017-01-05

Family

ID=55725016

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2015/082628 WO2017000109A1 (fr) 2015-06-29 2015-06-29 Procédé de recherche, appareil de recherche, équipement utilisateur, et produit de programme informatique

Country Status (2)

Country Link
CN (1) CN105518678B (fr)
WO (1) WO2017000109A1 (fr)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116737982A (zh) * 2023-08-11 2023-09-12 拓锐科技有限公司 一种基于数据分析的图片搜索结果智能筛选管理系统

Families Citing this family (19)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107315181B (zh) * 2016-04-26 2020-11-06 天津远度科技有限公司 一种避障方法及装置、飞行器
CN107403128B (zh) * 2016-05-20 2020-12-29 株式会社理光 一种物品识别方法及装置
CN107403179B (zh) * 2016-05-20 2020-10-23 株式会社理光 一种物品包装信息的注册方法及装置
CN106777177A (zh) * 2016-12-22 2017-05-31 百度在线网络技术（北京）有限公司 检索方法和装置
WO2018145577A1 (fr) 2017-02-08 2018-08-16 腾讯科技（深圳）有限公司 Procédé et dispositif de recommandation d'expression faciale
CN108401005B (zh) * 2017-02-08 2021-05-14 腾讯科技（深圳）有限公司 一种表情推荐方法和装置
TWI650656B (zh) * 2017-05-26 2019-02-11 虹光精密工業股份有限公司 於電腦系統搜尋影像檔案之方法、影像檔案搜尋裝置以及電腦系統
CN109213332B (zh) * 2017-06-29 2022-11-08 北京搜狗科技发展有限公司 一种表情图片的输入方法和装置
CN107679070B (zh) * 2017-08-22 2021-10-01 科大讯飞股份有限公司 一种智能阅读推荐方法与装置、电子设备
CN107569848B (zh) * 2017-08-30 2020-08-04 武汉斗鱼网络科技有限公司 一种游戏分类方法、装置及电子设备
CN107862239A (zh) * 2017-09-15 2018-03-30 广州唯品会研究院有限公司 一种结合文本与图片进行图片识别的方法及其装置
CN107798070A (zh) * 2017-09-26 2018-03-13 平安普惠企业管理有限公司 一种网页数据获取方法及终端设备
CN107633086A (zh) * 2017-09-27 2018-01-26 北京萌哥玛丽科技有限公司 一种通过图片实现视频或音频查找的控制方法及系统
CN107944022A (zh) * 2017-12-11 2018-04-20 努比亚技术有限公司 图片分类方法、移动终端及计算机可读存储介质
CN108563663A (zh) * 2018-01-04 2018-09-21 出门问问信息科技有限公司 图片推荐方法、装置、设备及存储介质
CN108960234A (zh) * 2018-06-13 2018-12-07 山东师范大学 一种基于词袋模型的Logo识别方法及系统
CN110532413B (zh) * 2019-07-22 2023-08-08 平安科技（深圳）有限公司 基于图片匹配的信息检索方法、装置、计算机设备
CN111782841A (zh) * 2019-11-27 2020-10-16 北京沃东天骏信息技术有限公司 图像搜索方法、装置、设备和计算机可读介质
CN117216308B (zh) * 2023-11-09 2024-04-26 天津华来科技股份有限公司 基于大模型的搜索方法、系统、设备及介质

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101388023A (zh) * 2008-09-12 2009-03-18 北京搜狗科技发展有限公司 电子地图兴趣点数据冗余检测方法和系统
CN102625937A (zh) * 2009-08-07 2012-08-01 谷歌公司 用于对视觉查询作出响应的体系结构
CN102682091A (zh) * 2012-04-25 2012-09-19 腾讯科技（深圳）有限公司 基于云服务的视觉搜索方法和系统
CN103473327A (zh) * 2013-09-13 2013-12-25 广东图图搜网络科技有限公司 图像检索方法与系统
CN103493069A (zh) * 2010-12-01 2014-01-01 谷歌公司 响应于视觉查询识别匹配的规范文档
CN104572905A (zh) * 2014-12-26 2015-04-29 小米科技有限责任公司 照片索引创建方法、照片搜索方法及装置

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101593278B (zh) * 2008-05-27 2013-01-16 佳能株式会社 文档图像的语言判别方法和系统

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101388023A (zh) * 2008-09-12 2009-03-18 北京搜狗科技发展有限公司 电子地图兴趣点数据冗余检测方法和系统
CN102625937A (zh) * 2009-08-07 2012-08-01 谷歌公司 用于对视觉查询作出响应的体系结构
CN103493069A (zh) * 2010-12-01 2014-01-01 谷歌公司 响应于视觉查询识别匹配的规范文档
CN102682091A (zh) * 2012-04-25 2012-09-19 腾讯科技（深圳）有限公司 基于云服务的视觉搜索方法和系统
CN103473327A (zh) * 2013-09-13 2013-12-25 广东图图搜网络科技有限公司 图像检索方法与系统
CN104572905A (zh) * 2014-12-26 2015-04-29 小米科技有限责任公司 照片索引创建方法、照片搜索方法及装置

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116737982A (zh) * 2023-08-11 2023-09-12 拓锐科技有限公司 一种基于数据分析的图片搜索结果智能筛选管理系统
CN116737982B (zh) * 2023-08-11 2023-10-31 拓锐科技有限公司 一种基于数据分析的图片搜索结果智能筛选管理系统

Also Published As

Publication number Publication date
CN105518678A (zh) 2016-04-20
CN105518678B (zh) 2018-07-31

Similar Documents

Publication Publication Date Title
WO2017000109A1 (fr) Procédé de recherche, appareil de recherche, équipement utilisateur, et produit de programme informatique
US10706094B2 (en) System and method for customizing a display of a user device based on multimedia content element signatures
CN109284729B (zh) 基于视频获取人脸识别模型训练数据的方法、装置和介质
CN104239408B (zh) 基于由移动设备记录的图像的内容的数据访问
US9875303B2 (en) System and process for building a catalog using visual objects
EP3028184B1 (fr) Procédé et système pour rechercher des images
US11461386B2 (en) Visual recognition using user tap locations
WO2017045443A1 (fr) Procédé et système d'extraction d'image
US10380461B1 (en) Object recognition
WO2015051629A1 (fr) Procédé et dispositif de mise en ordre d'images
US10380267B2 (en) System and method for tagging multimedia content elements
TWI781554B (zh) 物品名稱確定方法、裝置、電腦設備及儲存媒體
CN110020093A (zh) 视频检索方法、边缘设备、视频检索装置及存储介质
JP2019520662A (ja) 商標画像のコンテンツ・ベースの検索及び取得
CN111291765A (zh) 用于确定相似图片的方法和装置
JP6310599B1 (ja) 検索装置、検索方法および検索プログラム
CN109189960A (zh) 信息获取方法、信息获取装置及计算机存储介质
KR20220039578A (ko) 사용자 선택 의류를 기반으로 의류 추천 정보를 제공하는 방법, 이를 이용하는 서버 및 프로그램
US20150052155A1 (en) Method and system for ranking multimedia content elements
KR20150101846A (ko) 스케치를 기반으로 하는 영상 분류 서비스 시스템, 사용자 장치, 서비스 제공 장치, 그 서비스 방법 및 컴퓨터 프로그램이 기록된 기록매체
US20200074218A1 (en) Information processing system, information processing apparatus, and non-transitory computer readable medium
CN109766052B (zh) 菜品图片上传方法、装置、计算机设备及可读存储介质
WO2016155537A1 (fr) Procédé et dispositif de classification de résultats de recherche d'objets image
KR20200024540A (ko) 이미지 검색을 위한 요약 정보 제공 방법 및 이를 지원하는 서비스 장치
US12038977B2 (en) Visual recognition using user tap locations

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 15896640

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC , EPO FORM 1205A DATED 12.04.18.

122 Ep: pct application non-entry in european phase

Ref document number: 15896640

Country of ref document: EP

Kind code of ref document: A1