A method and system for extracting target objects based on image recognition

By constructing a three-dimensional model of the target object and performing state adjustment interception, comparative judgment, the problem of low target object recognition accuracy in the prior art is solved, and high-precision extraction under different conditions is achieved.

CN114998663BActive Publication Date: 2024-11-08ZHENGZHOU UNIVERSITY OF AERONAUTICS
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210802230.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-07-07
Publication Date
2024-11-08
Estimated Expiration
2042-07-07

AI Technical Summary

Technical Problem

The existing image recognition technology is difficult to accurately identify and extract target objects under complex circumstances, especially when the target objects are at different times, different states, and different angles, resulting in low recognition accuracy.

Method used

By obtaining multiple azimuth images captured at the same time, the target three-dimensional model of the target object is constructed, the image to be processed is divided into the background image and the recognition image, initial recognition and marking are performed, the three-dimensional model is adjusted according to the state characteristics, and comparison and judgment are performed to extract the target object.

Benefits of technology

When the target object is at different times, different states, and different angles, it can identify and extract the target object with high accuracy, improving the accuracy of image recognition.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114998663B_ABST
    Figure CN114998663B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of image recognition technology, and specifically discloses a method and system for extracting a target object based on image recognition. The present invention constructs a target three-dimensional model by acquiring images shot in multiple directions; divides them into a background image and multiple recognition images; performs initial recognition; adjusts and intercepts the target three-dimensional model; and performs comparison, judgment, and target object extraction. The target three-dimensional model of the target object can be constructed to initially recognize the image to be processed, mark the suspicious image, and then adjust and intercept the target three-dimensional model according to the state characteristics of the suspicious image to generate a state-adjusted intercepted image. By comparing and judging the state-adjusted intercepted image with the suspicious image, when it is determined that the suspicious image is the image of the target object, the target object is extracted, so that the target object can be identified and extracted when the target object is shot at different times, different states, and different angles. The image recognition accuracy is high, and the target object can be accurately extracted.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of image recognition, and in particular relates to a method and system for extracting a target object based on image recognition. Background Art

[0002] Image recognition refers to the technology of using computers to process, analyze and understand images to identify targets and objects of various different patterns. It is a practical application of deep learning algorithms. Image recognition is an important field of artificial intelligence. The generation and updating of image recognition technology has become a very important development direction at present, and it has shown good development prospects. Image recognition technology has been widely used in information collection, medical treatment and product safety, and has played a very important role.

[0003] Image recognition technology is used to identify multiple objects in an image, and then the target object is extracted from the multiple objects. Since the target objects are photographed at different times, in different states, and at different angles and have different shapes, the existing image recognition technology is difficult to identify and extract targets in complex situations, resulting in low image recognition accuracy and inability to accurately extract the target object. Summary of the invention

[0004] The purpose of the embodiments of the present invention is to provide a method and system for extracting a target object based on image recognition, aiming to solve the problems raised in the background technology.

[0005] To achieve the above objectives, the embodiments of the present invention provide the following technical solutions:

[0006] A method for extracting a target object based on image recognition, the method specifically comprising the following steps:

[0007] Acquire multiple azimuth images of the target object, extract basic features of the target object based on the multiple azimuth images, and construct a target three-dimensional model of the target object;

[0008] Acquire an image to be processed, and divide the image to be processed into a background image and a plurality of recognition images;

[0009] Performing initial recognition on the plurality of recognition images according to the basic features, and marking the recognition images that pass the recognition as suspicious images;

[0010] Performing state analysis on the suspicious image to obtain state features of the suspicious image, adjusting and intercepting the target three-dimensional model according to the state features, and generating a state-adjusted intercepted image;

[0011] The state-adjusted captured image is compared with the suspicious image and judged. When the suspicious image is an image of a target object, the target object is extracted according to the suspicious image.

[0012] As a further limitation of the technical solution of the embodiment of the present invention, the acquiring of images taken from multiple orientations of the target object, extracting basic features of the target object based on the images taken from multiple orientations, and constructing a target three-dimensional model of the target object specifically comprises the following steps:

[0013] Acquire multiple azimuth shooting images of the target object;

[0014] Performing feature analysis on the images taken in the plurality of said directions to extract basic features of the target object;

[0015] A target three-dimensional model of the target object is constructed based on the basic features.

[0016] As a further limitation of the technical solution of the embodiment of the present invention, the obtaining of the image to be processed and dividing the image to be processed into a background image and a plurality of recognition images specifically include the following steps:

[0017] Get the image to be processed;

[0018] Performing basic recognition on the image to be processed to generate a basic recognition result;

[0019] According to the basic recognition result, generating a division boundary in the image to be processed;

[0020] According to the division boundary, the image to be processed is divided into a background image and a plurality of recognition images.

[0021] As a further limitation of the technical solution of the embodiment of the present invention, the initial recognition of the plurality of recognition images according to the basic features and marking the recognition images that have passed the recognition as suspicious images specifically includes the following steps:

[0022] Performing feature analysis on the plurality of recognition images to obtain a plurality of recognition features;

[0023] Performing fuzzy comparison between the basic feature and the plurality of identification features to generate a fuzzy comparison result;

[0024] According to the fuzzy comparison result, determining whether there is a recognition image that passes the comparison;

[0025] If there is a recognized image that passes the comparison, the corresponding recognized image is marked as a suspicious image.

[0026] As a further limitation of the technical solution of the embodiment of the present invention, the state analysis of the suspicious image, obtaining the state characteristics of the suspicious image, adjusting and intercepting the target three-dimensional model according to the state characteristics, and generating the state-adjusted intercepted image specifically include the following steps:

[0027] Performing effect state analysis on the suspicious image to obtain effect state features;

[0028] Performing an occlusion state analysis on the suspicious image to obtain an occlusion state feature;

[0029] According to the effect state characteristics and the occlusion state characteristics, the target three-dimensional model is adjusted and captured to generate a state-adjusted captured image.

[0030] As a further limitation of the technical solution of the embodiment of the present invention, the comparing and judging the state-adjusted captured image with the suspicious image, when the suspicious image is an image of a target object, extracting the target object according to the suspicious image specifically comprises the following steps:

[0031] Performing a similarity evaluation between the state-adjusted captured image and the suspicious image to obtain a comparative similarity;

[0032] Determining whether the comparison similarity is greater than a preset standard similarity;

[0033] If the comparison similarity is greater than a preset standard similarity, determining that the suspicious image is an image of the target object;

[0034] Target objects are extracted according to the suspicious image.

[0035] A target object extraction system based on image recognition, the system comprises a three-dimensional model construction unit, a processed image division unit, an initial recognition marking unit, a model adjustment interception unit and a comparison judgment extraction unit, wherein:

[0036] A basic model building unit is used to obtain images taken from multiple orientations of a target object, extract basic features of the target object based on the images taken from the multiple orientations, and build a target three-dimensional model of the target object;

[0037] A processing image division unit, used for acquiring an image to be processed, and dividing the image to be processed into a background image and a plurality of recognition images;

[0038] An initial recognition marking unit, used to perform initial recognition on the plurality of recognition images according to the basic features, and mark the recognition images that have passed the recognition as suspicious images;

[0039] A model adjustment and interception unit, configured to perform a state analysis on the suspicious image, obtain state features of the suspicious image, adjust and intercept the target three-dimensional model according to the state features, and generate a state-adjusted intercepted image;

[0040] The comparison, judgment and extraction unit is used to compare and judge the state-adjusted intercepted image with the suspicious image, and when the suspicious image is an image of a target object, extract the target object according to the suspicious image.

[0041] As a further limitation of the technical solution of the embodiment of the present invention, the basic model building unit specifically includes:

[0042] A shooting acquisition module is used to acquire images shot in multiple directions for shooting a target object in multiple directions;

[0043] A feature analysis module, used to perform feature analysis on the images taken in the plurality of orientations to extract basic features of the target object;

[0044] The model building module is used to build a target three-dimensional model of the target object according to the basic features.

[0045] As a further limitation of the technical solution of the embodiment of the present invention, the image processing division unit specifically includes:

[0046] An image acquisition module, used for acquiring an image to be processed;

[0047] A basic recognition module, used to perform basic recognition on the image to be processed and generate a basic recognition result;

[0048] A partition analysis module, used to generate a partition boundary in the image to be processed according to the basic recognition result;

[0049] The image division module is used to divide the image to be processed into a background image and a plurality of recognition images according to the division boundary.

[0050] As a further limitation of the technical solution of the embodiment of the present invention, the model adjustment and interception unit specifically includes:

[0051] An effect analysis module, used to perform effect status analysis on the suspicious image to obtain effect status features;

[0052] An occlusion analysis module is used to analyze the occlusion state of the suspicious image to obtain occlusion state features;

[0053] The adjustment and capture module is used to adjust and capture the target three-dimensional model according to the effect state characteristics and the occlusion state characteristics to generate a state adjustment capture image.

[0054] Compared with the prior art, the present invention has the following beneficial effects:

[0055] The embodiment of the present invention acquires images taken from multiple directions to construct a three-dimensional model of the target; divides the images into a background image and multiple recognition images; performs initial recognition; adjusts and captures the target three-dimensional model; and performs comparison, judgment, and target object extraction. By constructing a target three-dimensional model of the target object, the image to be processed can be initially recognized, suspicious images can be marked, and then the target three-dimensional model can be adjusted and captured according to the state characteristics of the suspicious image to generate a state-adjusted captured image. By comparing and judging the state-adjusted captured image with the suspicious image, when it is determined that the suspicious image is an image of the target object, the target object can be extracted. Thus, the target object can be recognized and extracted when the target object is photographed at different times, different states, and different angles. The image recognition accuracy is high, and the target object can be accurately extracted. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention.

[0057] Figure 1 A flow chart of a method provided by an embodiment of the present invention is shown.

[0058] Figure 2 A flow chart of constructing a target three-dimensional model in the method provided in an embodiment of the present invention is shown.

[0059] Figure 3 A flow chart of dividing an image to be processed in the method provided by an embodiment of the present invention is shown.

[0060] Figure 4 A flow chart of initially identifying image tags in the method provided by an embodiment of the present invention is shown.

[0061] Figure 5 A flow chart of adjusting and capturing a target three-dimensional model in a method provided by an embodiment of the present invention is shown.

[0062] Figure 6 A flow chart of comparing and judging target object extraction in the method provided by an embodiment of the present invention is shown.

[0063] Figure 7 The application architecture diagram of the system provided by the embodiment of the present invention is shown.

[0064] Figure 8 A structural block diagram of a basic model building unit in a system provided by an embodiment of the present invention is shown.

[0065] Fig. 9 The structure block diagram of the image division processing unit in the system provided by the embodiment of the present invention is shown.

[0066] Fig.10 The structure block diagram of the model adjustment and interception unit in the system provided by the embodiment of the present invention is shown. DETAILED DESCRIPTION

[0067] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0068] It is understandable that through image recognition technology, multiple objects in an image are identified and then the target object is extracted from the multiple objects. Since the target objects are photographed at different times, in different states, and at different angles and have different forms, the existing image recognition technology is difficult to identify and extract targets in complex situations, resulting in low image recognition accuracy and inability to accurately extract the target object.

[0069] To solve the above problems, the embodiment of the present invention acquires images taken from multiple directions to construct a target three-dimensional model; divides the image to be processed into a background image and multiple recognition images; performs initial recognition; adjusts and captures the target three-dimensional model; and performs comparison, judgment, and target object extraction. By constructing a target three-dimensional model of the target object, the image to be processed can be initially recognized, suspicious images can be marked, and then the target three-dimensional model can be adjusted and captured according to the state characteristics of the suspicious image to generate a state-adjusted captured image. By comparing and judging the state-adjusted captured image with the suspicious image, when it is determined that the suspicious image is the image of the target object, the target object can be extracted. Therefore, the target object can be identified and extracted when the target object is photographed at different times, different states, and different angles. The image recognition accuracy is high, and the target object can be accurately extracted.

[0070] Figure 1 A flow chart of a method provided by an embodiment of the present invention is shown.

[0071] Specifically, a method for extracting a target object based on image recognition comprises the following steps:

[0072] Step S101, acquiring images of a target object taken from multiple directions, extracting basic features of the target object based on the images taken from multiple directions, and constructing a target three-dimensional model of the target object.

[0073] In an embodiment of the present invention, for a target object that requires image recognition and extraction, preparatory work needs to be performed before recognition and extraction. By photographing the target object at the same distance and in multiple directions, images of the target object taken in multiple directions are obtained. By performing a comprehensive analysis on the images taken in multiple directions, basic features of the target object's appearance are extracted, and then a three-dimensional model is constructed according to the basic features of the target object's appearance to obtain a target three-dimensional model of the target object.

[0074] It can be understood that by photographing the target object at the same distance and in multiple directions, the images obtained from multiple directions can reflect the entire shape of the target object. Therefore, by comprehensively analyzing the images taken from multiple directions, the basic features of the target object's shape can be obtained, and then according to the basic features, a target three-dimensional model consistent with the target object's shape can be constructed; the basic features include: the shape composition of the target object, the relative size of each component, the relative position of each component, the color of each component, etc.

[0075] Specifically, Figure 2 A flow chart of constructing a target three-dimensional model in the method provided in an embodiment of the present invention is shown.

[0076] Among them, in the preferred embodiment provided by the present invention, the acquiring of images taken from multiple orientations of the target object, extracting basic features of the target object according to the images taken from multiple orientations, and constructing a target three-dimensional model of the target object specifically comprises the following steps:

[0077] Step S1011, acquiring multiple azimuth shooting images of the target object.

[0078] Step S1012, performing feature analysis on the plurality of images taken in the orientations to extract basic features of the target object.

[0079] Step S1013: construct a target three-dimensional model of the target object according to the basic features.

[0080] Furthermore, the target object extraction method based on image recognition also includes the following steps:

[0081] Step S102: acquiring an image to be processed, and dividing the image to be processed into a background image and a plurality of recognition images.

[0082] In an embodiment of the present invention, by acquiring an image to be processed, basic recognition of specific objects in the image is performed on the image to be processed to generate a basic recognition result, and then according to the basic recognition result, a division boundary separating multiple specific objects is generated in the image to be processed, and according to the division boundary, the multiple specific objects are divided and extracted, so that the image to be processed is divided into a background image and multiple recognition images.

[0083] It can be understood that the recognition and division process of the image to be processed is to automatically recognize the specific objects in the image to be processed, and to divide and extract multiple existing specific objects from the image to be processed, so that the image to be processed becomes a background image and multiple recognition images. For example, if the specific objects recognized from the image to be processed include: cows, sheep, and dogs, then the multiple recognition images divided are images of cows, sheep, and dogs, and the background image is the image after removing the cows, sheep, and dogs.

[0084] Specifically, Figure 3 A flow chart of dividing an image to be processed in the method provided by an embodiment of the present invention is shown.

[0085] Among them, in the preferred embodiment provided by the present invention, the step of obtaining the image to be processed and dividing the image to be processed into a background image and a plurality of recognition images specifically comprises the following steps:

[0086] Step S1021, obtaining an image to be processed.

[0087] Step S1022, performing basic recognition on the image to be processed to generate a basic recognition result.

[0088] Step S1023: generating a division boundary in the image to be processed according to the basic recognition result.

[0089] Step S1024: dividing the image to be processed into a background image and a plurality of recognition images according to the division boundary.

[0090] Furthermore, the target object extraction method based on image recognition also includes the following steps:

[0091] Step S103: performing initial recognition on the plurality of recognition images according to the basic features, and marking the recognition images that pass the recognition as suspicious images.

[0092] In an embodiment of the present invention, feature analysis is performed on multiple divided and extracted recognition images to obtain multiple corresponding recognition features. Fuzzy comparisons are performed on the multiple recognition features with basic features to obtain fuzzy comparison results. According to the fuzzy comparison results, it is determined whether the recognition features pass the fuzzy comparison with the basic features. For recognition features that pass the fuzzy comparison, their corresponding recognition images are marked as suspicious images.

[0093] It can be understood that in the fuzzy comparison, as long as the identification feature and the basic feature are identical in at least one aspect, the fuzzy comparison passes, for example, the appearance is identical or the shape of a certain component is identical.

[0094] Specifically, Figure 4 A flow chart of initially identifying image tags in the method provided by an embodiment of the present invention is shown.

[0095] Among them, in the preferred embodiment provided by the present invention, the initial recognition of the plurality of recognition images according to the basic features and marking the recognition images that have passed the recognition as suspicious images specifically include the following steps:

[0096] Step S1031, performing feature analysis on the plurality of recognition images to obtain a plurality of recognition features.

[0097] Step S1032, fuzzy comparison is performed on the basic feature and the plurality of recognition features to generate a fuzzy comparison result.

[0098] Step S1033: judging whether there is a recognized image that passes the comparison according to the fuzzy comparison result.

[0099] Step S1034: if there is a recognition image that passes the comparison, the corresponding recognition image is marked as a suspicious image.

[0100] Furthermore, the target object extraction method based on image recognition also includes the following steps:

[0101] Step S104: performing a state analysis on the suspicious image to obtain state features of the suspicious image, adjusting and capturing the target three-dimensional model according to the state features, and generating a state-adjusted captured image.

[0102] In an embodiment of the present invention, by analyzing the effect state and occlusion state of the suspicious target corresponding to the suspicious image, the effect state characteristics and occlusion state characteristics are obtained respectively, and then the effect state characteristics and occlusion state characteristics are integrated to make corresponding state adjustments to the target three-dimensional model, so that when the target three-dimensional model is in the corresponding state, a state-adjusted captured image is obtained.

[0103] It can be understood that the effect state characteristics are the light and shadow effects, shooting angle effects, etc. of the suspicious target corresponding to the suspicious image when being photographed; the occlusion state characteristics are the relative occlusion position, relative occlusion area, relative occlusion angle, etc. of the suspicious target corresponding to the suspicious image when being photographed.

[0104] Specifically, Figure 5 A flow chart of adjusting and capturing a target three-dimensional model in a method provided by an embodiment of the present invention is shown.

[0105] Among them, in the preferred embodiment provided by the present invention, the state analysis of the suspicious image is performed to obtain the state characteristics of the suspicious image, and the target three-dimensional model is adjusted and intercepted according to the state characteristics, and the state adjustment intercepted image is generated, which specifically includes the following steps:

[0106] Step S1041, performing effect status analysis on the suspicious image to obtain effect status features.

[0107] Step S1042: Analyze the occlusion state of the suspicious image to obtain occlusion state features.

[0108] Step S1043: adjusting and capturing the target three-dimensional model according to the effect state characteristics and the occlusion state characteristics to generate a state-adjusted captured image.

[0109] Furthermore, the target object extraction method based on image recognition also includes the following steps:

[0110] Step S105 : comparing and judging the state-adjusted captured image with the suspicious image, and when the suspicious image is an image of a target object, extracting the target object according to the suspicious image.

[0111] In an embodiment of the present invention, a similarity comparison and evaluation is performed between the suspicious image and the state-adjusted captured image to generate a contrast similarity, which is then compared with a preset standard similarity to determine whether the contrast similarity is greater than the preset standard similarity. When the contrast similarity is greater than the preset standard similarity, it indicates that the similarity between the suspicious image and the state-adjusted captured image is high. At this time, it can be determined that the suspicious image is an image of the target object, thereby being able to extract the target object from the suspicious image.

[0112] It can be understood that extracting the target object from the suspicious image is to remove the occlusion of the target object, fill the removed part, etc., so that the target object extracted from the suspicious image can be more complete than in the suspicious image.

[0113] Specifically, Figure 6 A flow chart of comparing and judging target object extraction in the method provided by an embodiment of the present invention is shown.

[0114] Among them, in the preferred embodiment provided by the present invention, the state adjustment captured image is compared with the suspicious image for judgment, and when the suspicious image is an image of the target object, the target object extraction according to the suspicious image specifically includes the following steps:

[0115] Step S1051 : performing a similarity evaluation between the state-adjusted captured image and the suspicious image to obtain a comparative similarity.

[0116] Step S1052, determining whether the comparison similarity is greater than a preset standard similarity.

[0117] Step S1053: If the comparison similarity is greater than a preset standard similarity, the suspicious image is determined to be an image of a target object.

[0118] Step S1054: extracting a target object according to the suspicious image.

[0119] Furthermore, Figure 7 The application architecture diagram of the system provided by the embodiment of the present invention is shown.

[0120] Among them, in another preferred embodiment provided by the present invention, a target object extraction system based on image recognition includes:

[0121] The basic model building unit 101 is used to obtain multiple azimuth images of the target object, extract basic features of the target object based on the multiple azimuth images, and build a target three-dimensional model of the target object.

[0122] In an embodiment of the present invention, for a target object that requires image recognition and extraction, preparatory work needs to be performed before recognition and extraction. The basic model construction unit 101 obtains images of the target object from multiple orientations by photographing the target object at the same distance and in multiple orientations, and extracts basic features of the target object's appearance by performing a comprehensive analysis of the images taken from multiple orientations. Then, a three-dimensional model is constructed according to the basic features of the target object's appearance to obtain a target three-dimensional model of the target object.

[0123] Specifically, Figure 8 It shows a structural block diagram of the basic model building unit 101 in the system provided by the embodiment of the present invention.

[0124] Among them, in the preferred embodiment provided by the present invention, the basic model building unit 101 specifically includes:

[0125] The photographing and acquiring module 1011 is used to acquire multiple-directional photographing images of the target object.

[0126] The feature analysis module 1012 is used to perform feature analysis on the multiple images taken in the above directions to extract basic features of the target object.

[0127] The model building module 1013 is used to build a target three-dimensional model of the target object according to the basic features.

[0128] Furthermore, the target object extraction system based on image recognition also includes:

[0129] The processed image division unit 102 is used to obtain an image to be processed and divide the image to be processed into a background image and a plurality of recognition images.

[0130] In an embodiment of the present invention, the image processing division unit 102 obtains the image to be processed, performs basic recognition of specific objects in the image to be processed, generates a basic recognition result, and then generates a division boundary in the image to be processed according to the basic recognition result, and divides and extracts multiple specific objects according to the division boundary, so that the image to be processed is divided into a background image and multiple recognition images.

[0131] Specifically, Fig. 9 The structure block diagram of the image division processing unit 102 in the system provided by the embodiment of the present invention is shown.

[0132] In a preferred embodiment of the present invention, the image processing division unit 102 specifically includes:

[0133] The image acquisition module 1021 is used to acquire the image to be processed.

[0134] The basic recognition module 1022 is used to perform basic recognition on the image to be processed and generate a basic recognition result.

[0135] The segmentation analysis module 1023 is used to generate a segmentation boundary in the image to be processed according to the basic recognition result.

[0136] The image division module 1024 is used to divide the image to be processed into a background image and a plurality of recognition images according to the division boundary.

[0137] Furthermore, the target object extraction system based on image recognition also includes:

[0138] The initial recognition marking unit 103 is used to perform initial recognition on the plurality of recognition images according to the basic features, and mark the recognition images that have passed the recognition as suspicious images.

[0139] In an embodiment of the present invention, the initial recognition marking unit 103 performs feature analysis on multiple divided and extracted recognition images to obtain multiple corresponding recognition features, and obtains fuzzy comparison results by fuzzy comparing the multiple recognition features with the basic features respectively. According to the fuzzy comparison results, it is determined whether the recognition features pass the fuzzy comparison with the basic features. For the recognition features that pass the fuzzy comparison, the corresponding recognition images are marked as suspicious images.

[0140] The model adjustment and interception unit 104 is used to perform state analysis on the suspicious image, obtain state features of the suspicious image, adjust and intercept the target three-dimensional model according to the state features, and generate a state-adjusted intercepted image.

[0141] In an embodiment of the present invention, the model adjustment capture unit 104 analyzes the effect state and occlusion state of the suspicious target corresponding to the suspicious image to obtain effect state characteristics and occlusion state characteristics respectively, and then combines the effect state characteristics and occlusion state characteristics to perform corresponding state adjustments on the target three-dimensional model, so that when the target three-dimensional model is in a corresponding state, a state adjustment capture image is captured.

[0142] Specifically, Fig.10 It shows a structural block diagram of the model adjustment and interception unit 104 in the system provided by an embodiment of the present invention.

[0143] Among them, in the preferred embodiment provided by the present invention, the model adjustment interception unit 104 specifically includes:

[0144] The effect analysis module 1041 is used to perform effect status analysis on the suspicious image to obtain effect status features.

[0145] The occlusion analysis module 1042 is used to perform occlusion state analysis on the suspicious image to obtain occlusion state features.

[0146] The adjustment and capture module 1043 is used to adjust and capture the target three-dimensional model according to the effect state characteristics and the occlusion state characteristics to generate a state adjustment capture image.

[0147] Furthermore, the target object extraction system based on image recognition also includes:

[0148] The comparison, judgment and extraction unit 105 is used to compare and judge the state-adjusted captured image with the suspicious image, and to extract the target object according to the suspicious image when the suspicious image is an image of the target object.

[0149] In the embodiment of the present invention, the comparison judgment extraction unit 105 performs a similarity comparison evaluation between the suspicious image and the state-adjusted captured image, generates a comparison similarity, and then compares the comparison similarity with a preset standard similarity to determine whether the comparison similarity is greater than the preset standard similarity. When the comparison similarity is greater than the preset standard similarity, it indicates that the similarity between the suspicious image and the state-adjusted captured image is high. At this time, it can be determined that the suspicious image is an image of the target object, so that the target object can be extracted from the suspicious image.

[0150] In summary, the embodiments of the present invention can construct a target three-dimensional model of the target object, perform initial recognition on the image to be processed, mark the suspicious image, and then adjust and capture the target three-dimensional model according to the state characteristics of the suspicious image to generate a state-adjusted captured image. By comparing and judging the state-adjusted captured image with the suspicious image, when it is determined that the suspicious image is an image of the target object, the target object is extracted. Therefore, the target object can be recognized and extracted when the target object is photographed at different times, different states, and different angles. The image recognition accuracy is high, and the target object can be accurately extracted.

[0151] It should be understood that, although each step in the flow chart of each embodiment of the present invention is shown in sequence according to the indication of the arrow, these steps are not necessarily performed in sequence according to the order indicated by the arrow. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be performed in other orders. Moreover, at least a portion of the steps in each embodiment may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these sub-steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of other steps or sub-steps or stages of other steps.

[0152] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0153] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0154] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

[0155] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for extracting a target object based on image recognition, characterized in that: The method specifically comprises the following steps: Acquire multiple azimuth images of the target object, extract basic features of the target object based on the multiple azimuth images, and construct a target three-dimensional model of the target object; Acquire an image to be processed, and divide the image to be processed into a background image and a plurality of recognition images; Performing initial recognition on the plurality of recognition images according to the basic features, and marking the recognition images that pass the recognition as suspicious images; Performing state analysis on the suspicious image to obtain state features of the suspicious image, adjusting and intercepting the target three-dimensional model according to the state features, and generating a state-adjusted intercepted image; Comparing and judging the state-adjusted captured image with the suspicious image, and when the suspicious image is an image of a target object, extracting the target object according to the suspicious image; The performing of initial recognition on the plurality of recognition images according to the basic features and marking the recognition images that have passed the recognition as suspicious images specifically comprises the following steps: Performing feature analysis on the plurality of recognition images to obtain a plurality of recognition features; Performing fuzzy comparison between the basic feature and the plurality of identification features to generate a fuzzy comparison result; According to the fuzzy comparison result, determining whether there is a recognition image that passes the comparison; If there is a recognized image that passes the comparison, the corresponding recognized image is marked as a suspicious image; The state analysis of the suspicious image, obtaining the state characteristics of the suspicious image, adjusting and intercepting the target three-dimensional model according to the state characteristics, and generating the state-adjusted intercepted image specifically comprises the following steps: Performing effect state analysis on the suspicious image to obtain effect state features; Performing an occlusion state analysis on the suspicious image to obtain an occlusion state feature; According to the effect state characteristics and the occlusion state characteristics, the target three-dimensional model is adjusted and intercepted to generate a state-adjusted intercepted image; The comparing and judging the state-adjusted captured image with the suspicious image, and when the suspicious image is an image of a target object, extracting the target object according to the suspicious image specifically comprises the following steps: Performing a similarity evaluation between the state-adjusted captured image and the suspicious image to obtain a comparative similarity; Determining whether the comparison similarity is greater than a preset standard similarity; If the comparison similarity is greater than a preset standard similarity, determining that the suspicious image is an image of the target object; Target objects are extracted according to the suspicious image.

2. The target object extraction method based on image recognition according to claim 1, characterized in that: The step of acquiring images of the target object taken from multiple orientations, extracting basic features of the target object based on the images taken from multiple orientations, and constructing a target three-dimensional model of the target object specifically comprises the following steps: Acquire multiple azimuth shooting images of the target object; Performing feature analysis on the images taken in the plurality of said directions to extract basic features of the target object; A target three-dimensional model of the target object is constructed according to the basic features.

3. The target object extraction method based on image recognition according to claim 1, characterized in that: The step of obtaining an image to be processed and dividing the image to be processed into a background image and a plurality of recognition images specifically comprises the following steps: Get the image to be processed; Performing basic recognition on the image to be processed to generate a basic recognition result; According to the basic recognition result, generating a division boundary in the image to be processed; According to the division boundary, the image to be processed is divided into a background image and a plurality of recognition images.

4. A target object extraction system based on image recognition, used to implement any of the methods described in claims 1-3, characterized in that: The system includes a three-dimensional model building unit, a processed image division unit, an initial recognition marking unit, a model adjustment interception unit and a comparison judgment extraction unit, wherein: A basic model building unit is used to obtain images taken from multiple orientations of a target object, extract basic features of the target object based on the images taken from the multiple orientations, and build a target three-dimensional model of the target object; A processing image division unit, used for acquiring an image to be processed, and dividing the image to be processed into a background image and a plurality of recognition images; An initial recognition marking unit, used to perform initial recognition on the plurality of recognition images according to the basic features, and mark the recognition images that have passed the recognition as suspicious images; A model adjustment and interception unit, configured to perform a state analysis on the suspicious image, obtain state features of the suspicious image, adjust and intercept the target three-dimensional model according to the state features, and generate a state-adjusted intercepted image; The comparison, judgment and extraction unit is used to compare and judge the state-adjusted intercepted image with the suspicious image, and when the suspicious image is an image of a target object, extract the target object according to the suspicious image.

5. The target object extraction system based on image recognition according to claim 4 is characterized in that: The basic model building unit specifically includes: A shooting acquisition module is used to acquire images shot in multiple directions for shooting a target object in multiple directions; A feature analysis module, used to perform feature analysis on the images taken in the plurality of orientations to extract basic features of the target object; The model building module is used to build a target three-dimensional model of the target object according to the basic features.

6. The target object extraction system based on image recognition according to claim 4, characterized in that: The image processing division unit specifically includes: An image acquisition module, used for acquiring an image to be processed; A basic recognition module, used to perform basic recognition on the image to be processed and generate a basic recognition result; A partition analysis module, used to generate a partition boundary in the image to be processed according to the basic recognition result; The image division module is used to divide the image to be processed into a background image and a plurality of recognition images according to the division boundary.

7. The target object extraction system based on image recognition according to claim 4 is characterized in that: The model adjustment and interception unit specifically includes: An effect analysis module, used to perform effect status analysis on the suspicious image to obtain effect status features; An occlusion analysis module is used to analyze the occlusion state of the suspicious image to obtain occlusion state features; The adjustment and capture module is used to adjust and capture the target three-dimensional model according to the effect state characteristics and the occlusion state characteristics to generate a state adjustment capture image.