Biological material experiment identification system based on computer vision
Through a biomaterial experimental recognition system based on computer vision, the problems of inaccurate results and low efficiency in traditional methods are solved, intelligent identification and resource optimization are achieved, and the accuracy and efficiency of biomaterial experiments are improved.
Patent Information
- Application Number
- CN202510620066.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-14
- Publication Date
- 2025-08-19
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Traditional biomaterial experimental identification relies on manual observation and recording, resulting in inaccurate results and low efficiency, and lack of effective correlation target prediction and resource optimization mechanisms.
A biomaterial experimental recognition system based on computer vision is adopted, including a material database, requirements module, identification module and analysis module. Through screening and judgment models, correlation analysis models and image recognition feature combinations, intelligent recognition and resource optimization are achieved.
It improves the accuracy and efficiency of biomaterial experimental identification, optimizes resource allocation, reduces the number of identification features, and improves the feedback of identification results and the identification efficiency of associated targets.
Smart Images
Figure CN120510495A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of biomaterial experiment identification, and in particular is a biomaterial experiment identification system based on computer vision. Background Art
[0002] Accurately and efficiently acquiring and analyzing information about biological materials is crucial in fields such as biological research, biomedical engineering, and biomaterials science. Traditional methods rely on manual observation and recording, which is not only time-consuming and labor-intensive, but also limited by the observer's subjective judgment and experience, making it difficult to guarantee the accuracy and consistency of the results.
[0003] In terms of target association analysis, when screening and determining the targets to be identified, excessive reliance on manual judgment and experience is placed on this process. This not only increases the workload but also easily introduces subjective errors. At the same time, there is a lack of effective prediction and evaluation mechanisms for the associated targets that may be subsequently identified after each identified target. This makes it difficult to ensure the accuracy of the primary target while also rationally allocating computing power and other resources to effectively identify the associated targets. Furthermore, existing identification methods are often based on preset, fixed identification features, without sufficient analysis of the characteristics of biological materials, leaving room for improvement in identification efficiency.
[0004] Based on this, in order to solve the above problems, the present invention provides a biomaterial experiment identification system based on computer vision. Summary of the Invention
[0005] In order to solve the problems existing in the above solutions, the present invention provides a biomaterial experiment identification system based on computer vision.
[0006] The purpose of the present invention can be achieved through the following technical solutions:
[0007] A biomaterial experiment identification system based on computer vision, including a material database, a requirement module, an identification module and an analysis module;
[0008] The material database is used to store biomaterial data of various biomaterials, and the biomaterial data includes various candidate targets corresponding to the biomaterials and a set of identification features corresponding to each candidate target.
[0009] Furthermore, the method for establishing the material database includes:
[0010] obtaining an experimental range, and setting a biomaterial screening feature according to the experimental range;
[0011] Acquire each candidate material in real time, screen each candidate material according to the biomaterial screening characteristics, and obtain each reserve material;
[0012] Identifying each candidate target of each reserve material, performing identification feature analysis on each candidate target, and obtaining an identification feature set corresponding to each candidate target;
[0013] Organize the reserve materials, candidate targets and identification feature sets into biological material data;
[0014] A database is established, and the data of each biological material is input into the database for storage, and the database is marked as a material database.
[0015] Furthermore, the method of screening each candidate material according to the biomaterial screening characteristics includes:
[0016] A screening judgment model is established, and the expression of the screening judgment model is:
[0017]
[0018] Where: (CT, DX) is the input data, CT is the biomaterial screening feature; DX is the candidate material for screening evaluation; the output data is the screening evaluation value SP(CT, DX), which is 1 or 0; DX→CT indicates that the candidate material for screening evaluation meets the requirements of the biomaterial screening feature;
[0019] The screening characteristics of each candidate material and biomaterial are analyzed by the screening judgment model to obtain a screening evaluation value corresponding to each candidate material, and each candidate material with the screening evaluation value equal to 1 is marked as a reserve material.
[0020] The requirement module analyzes the user's experimental requirements, determines the biomaterial and identifies the target, and marks the biomaterial as the target material;
[0021] Matching corresponding biomaterial data from the material database according to the target material; determining a set of identification features of the identified target according to the biomaterial data; analyzing the identified target to determine associated targets and their priorities;
[0022] Determine each reference target based on the biomaterial data, the identification target, and the associated target, and identify a set of identification features of each reference target based on the biomaterial data;
[0023] Acquire experimental detail data in real time, analyze each of the identification feature sets based on the experimental detail data, and determine the feature unit sets corresponding to the identification target, the associated target, and the reference target; compare each feature unit set to determine the image recognition feature combination corresponding to the identification target and the associated target, respectively.
[0024] Furthermore, the method for analyzing the identification target includes:
[0025] Acquire experimental requirements, acquire historical experimental identification data according to the experimental requirements, integrate the experimental requirements and the historical experimental identification data into a training set, establish an association analysis model based on the training set, and perform analysis using the association analysis model to obtain associated targets corresponding to the identified target;
[0026] Collecting association data corresponding to each of the association targets, determining each identification target associated with the association target and corresponding association probabilities and experimental auxiliary effect data based on the association data; and evaluating corresponding auxiliary effect values based on the experimental auxiliary effect data;
[0027] Marking the upper limit of the auxiliary effect value as the upper limit value; marking the associated target with the auxiliary effect value of the upper limit value as the associated necessary target; sorting the associated necessary targets in descending order according to the associated probability corresponding to the upper limit value to obtain a first priority sequence;
[0028] According to the formula Calculate the priority value of each remaining associated target;
[0029] Where: PY is the priority value; i represents the identification target associated with the corresponding associated target, i = 1, 2, ..., n, n is a positive integer; gi is the association probability; e is a natural constant; FGi is the auxiliary effect value;
[0030] Adding each of the associated targets to the first sequence in descending order of priority;
[0031] The priority of each associated target is determined according to the first sequence.
[0032] Furthermore, the method for evaluating the corresponding auxiliary effect value based on the experimental auxiliary effect data includes:
[0033] Determine the best experimental results based on experimental requirements and historical experimental data, and determine the required data and necessary data for each experiment based on the best experimental results;
[0034] According to the necessary experimental data and identification targets, the necessity of each associated target is analyzed to determine the associated necessary targets; the auxiliary effect value of each associated necessary target is marked as 10; the value range of the auxiliary effect value is [1, 10];
[0035] The effect differences of each related target are simulated through experimental demand data to obtain the auxiliary effect value corresponding to each related target.
[0036] Furthermore, the method for comparing each feature unit set includes:
[0037] Identify each identification feature corresponding to each feature unit set, perform combination comparison and analysis on each identification feature, and determine each candidate image recognition combination corresponding to each feature unit set;
[0038] The recognition efficiency of each candidate image recognition combination is compared to determine the image recognition feature combinations corresponding to the recognition target and the associated target respectively.
[0039] The recognition module is used to perform experimental recognition, collect images of biological materials, obtain collected images, recognize the collected images according to the image recognition feature combination of each recognition target, and obtain target recognition data of each recognition feature;
[0040] Monitor service idle resources in real time, determine each associated service target based on the service idle resources and the priority of each associated target, identify the collected image based on the image recognition feature combination corresponding to the associated service target, and obtain target recognition data of each associated service target.
[0041] The analysis module is used to analyze each target recognition data to obtain corresponding experimental recognition analysis results.
[0042] Compared with the prior art, the present invention has the following beneficial effects:
[0043] Through the coordination of the material database, requirement module, identification module, and analysis module, intelligent identification of biomaterial experiments is achieved. The targets to be identified are selected from a large number of candidate identification targets, and correlation analysis is performed on them. The associated targets that may be identified later for each identification target are deduced, and the priority of these associated targets is evaluated to optimize resource allocation. It is crucial to ensure that the identification task can be completed efficiently under limited computing power and other resources. The identification features of each identification target are analyzed, and by comparing the identification features, the combination of each image identification feature is determined, reducing the number of identification features in the combination to improve identification efficiency and data processing volume. At the same time, by continuously optimizing the image recognition feature combination, the identification results can also be fed back to the identification service of the associated target, further improving the accuracy and efficiency of identification. BRIEF DESCRIPTION OF THE DRAWINGS
[0044] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0045] Figure 1 This is a principle block diagram of the present invention. DETAILED DESCRIPTION
[0046] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.
[0047] like Figure 1 As shown, a biomaterial experiment identification system based on computer vision includes a material database, a demand module, an identification module and an analysis module;
[0048] The material database is used to store various biomaterial data, which include various candidate targets of the biomaterial and a set of identification features of each candidate target; candidate targets such as cells, tissues, microorganisms, proteins, nucleic acids, polysaccharides and other possible identification targets, and each candidate target is determined specifically according to the identification target of the biomaterial; the identification feature set is the various changing identification features that the candidate target may have during the experiment. For example, cells are candidates for selection and are often in dynamic change. Their identification features are not single and fixed, but may include multiple features such as shape, size, color, texture, and motion trajectory, and may still be changing with time, resulting in changes in their identification features under different circumstances. Therefore, according to their possible changes, the identification feature set of the candidate target is formed, and the corresponding identification features can be determined subsequently according to the conditions in which they are located.
[0049] In one embodiment, the material database can be established based on existing technologies, such as directly having the platform staff summarize the corresponding biological material data and then establish the material database.
[0050] In one embodiment, the method for establishing a material database includes:
[0051] Obtain the experimental scope of the application equipment, that is, which biological materials may be experimentally analyzed by the experimental equipment of this system, and determine the biomaterial screening characteristics based on the experimental scope, that is, each biomaterial that meets the biomaterial screening characteristics belongs to the experimental scope.
[0052] Various biological materials are acquired in real time and marked as candidate materials. No corresponding analysis is required for the biological materials already stored in the material database. The candidate materials are screened according to the biomaterial screening characteristics to obtain the biological materials for storage and mark them as reserve materials.
[0053] Identify each candidate target of the reserve material, perform identification feature analysis on each candidate target, and obtain an identification feature set corresponding to each candidate target.
[0054] Organize the reserve materials, candidate targets and identification feature sets into biological material data;
[0055] A database is established, and the data of each biological material is input into the database for storage, and the current database is marked as a material database.
[0056] In one embodiment, each candidate material is screened according to the biomaterial screening characteristics, and each candidate material can be intelligently screened based on existing technologies.
[0057] In one embodiment, the method for screening each candidate material according to the biomaterial screening characteristics includes:
[0058] Establish a screening judgment model, the expression of the screening judgment model is:
[0059]
[0060] Where: (CT, DX) is the input data, CT is the biomaterial screening feature; DX is the candidate material for screening evaluation; the output data is the screening evaluation value SP(CT, DX), which is 1 or 0; DX→CT indicates that the candidate material for screening evaluation meets the requirements of the biomaterial screening feature;
[0061] The screening characteristics of each candidate material and biomaterial are analyzed through the screening judgment model to obtain the screening evaluation value corresponding to each candidate material, and each candidate material with a screening evaluation value equal to 1 is marked as a reserve material.
[0062] The requirement module analyzes the user's experimental requirements, which include experimental purpose, experimental materials, experimental objectives and other related data; determines the biological material and the identification target, which is the cell waiting to be identified and is set by the user; there can be multiple identification targets; marks the corresponding biological material as the target material,
[0063] Matching corresponding biomaterial data from a material database according to the target material; determining a set of identification features for the identification target based on the biomaterial data; analyzing the identification target to determine each associated target and the priority of each associated target;
[0064] Marking each candidate target of a non-identified target and a non-associated target in the biomaterial data as a reference target, and identifying a set of identification features of each reference target based on the biomaterial data;
[0065] Acquire experimental detail data in real time. The experimental detail data is used to represent the process, time, and other data related to the changes of each recognition target and associated target of the current experiment, and is used to determine an approximate range in which they are located, and then obtain the corresponding available recognition features and integrate them into a feature unit set; analyze each recognition feature set through the experimental detail data to determine the feature unit set corresponding to the recognition target, associated target, and reference target; compare each feature unit set to determine the image recognition feature combination corresponding to the recognition target and the image recognition feature combination corresponding to the associated target.
[0066] In one embodiment, a method for analyzing an identified target includes:
[0067] Acquire experimental requirements, and acquire a large amount of historical experimental recognition data that is the same as the experimental requirements based on the experimental requirements. The historical experimental recognition data includes the recognition target that the user initially needs to identify, the recognition targets subsequently added based on the experimental requirements, and the data of various candidate targets for data support for subsequent analysis of the recognition target; integrate the experimental requirements and historical experimental recognition data into a training set based on the experimental requirements, and the training set includes input data and output data. The input data is the experimental requirements and the recognition target, and the output data is the associated target corresponding to the corresponding recognition target; establish an association analysis model based on a neural network, train it through the training set, and analyze it through the association analysis model after successful training to obtain various associated targets corresponding to the corresponding recognition target; there may be a situation where different recognition targets correspond to the same associated target.
[0068] Collecting association data corresponding to each association target, including the associated identification target, experimental auxiliary effect data on the associated identification target, association probability and other data. The association probability can be statistically analyzed based on the user's historical experimental data, or using historical experimental data of the same situation to obtain the association probability of the association target being associated with the identification target; determining each identification target associated with the association target and the corresponding association probability and experimental auxiliary effect data based on the association data; and evaluating the corresponding auxiliary effect value based on the experimental auxiliary effect data;
[0069] The upper limit of the auxiliary effect value is marked as the upper limit value, such as 10 in [1, 10] is the upper limit value; the associated targets with the auxiliary effect value of the upper limit value are marked as the associated necessary targets; the associated necessary targets are sorted in descending order according to the associated probability corresponding to the upper limit value, that is, when all are upper limit values, the high or low association probability of the associated identification target is used as the priority evaluation standard; and the first priority sequence is obtained;
[0070] According to the formula Calculate the priority value of each remaining associated target;
[0071] Where: PY is the priority value; i represents the identification target associated with the corresponding associated target, i = 1, 2, ..., n, n is a positive integer; gi is the association probability; e is a natural constant; FGi is the auxiliary effect value;
[0072] Add the remaining associated targets to the first sequence in descending order of priority;
[0073] The corresponding priorities are determined according to the order of the associated targets in the first sequence.
[0074] For example, to establish an association analysis model, we collect historical experimental identification data, which includes behavioral records, experimental requirements, identification targets, and other information in different experimental scenarios. We then perform data preprocessing, including data cleaning, missing value processing, and outlier detection, to ensure data quality and consistency.
[0075] Extract key features from the preprocessed data. These features should reflect the user's experimental needs, identification objectives, and possible association targets. Feature extraction methods may include statistical methods, machine learning algorithms, etc. The specific choice depends on the characteristics of the data and the complexity of the problem.
[0076] Select the appropriate model type to build the demand analysis model. In this example, choose a neural network-based model.
[0077] Constructing a neural network model involves determining parameters such as the number of layers, the number of neurons in each layer, and the activation function. These parameters should be adjusted based on the characteristics of the data and the needs of the problem.
[0078] The neural network model is trained using preprocessed data. The network parameters are adjusted using the backpropagation algorithm to ensure that the model output is as close as possible to the desired output. During the training process, methods such as cross-validation can be used to evaluate the model's performance and optimize the model based on the evaluation results.
[0079] Use data that was not used in training to validate and test the model to evaluate its generalization ability and accuracy. If the model's performance does not meet the requirements, you can return to the previous steps to adjust and optimize the model.
[0080] Deploy the trained demand analysis model to actual applications to analyze the user's experimental needs and identification goals, and determine the associated goals. In actual applications, new data can be continuously collected for model updates and optimization.
[0081] In one embodiment, the corresponding auxiliary effect value is evaluated according to the experimental auxiliary effect data, and the evaluation can be performed based on an existing effect evaluation method.
[0082] In one embodiment, a method for evaluating a corresponding auxiliary effect value based on experimental auxiliary effect data includes:
[0083] The optimal experimental results corresponding to the experimental requirements are determined based on the experimental requirements and historical experimental data. The experimental requirement data and necessary experimental data are determined based on the optimal experimental results. The necessary experimental data are the experimental data required to achieve the corresponding experimental purpose, mainly referring to the data corresponding to the identifiable targets in the biological materials. Statistics are not required for data that are not obtained through experimental materials. The experimental requirement data refers to the required data corresponding to achieving the optimal experimental results.
[0084] According to the necessary experimental data and identification targets, the necessity of each associated target is analyzed to determine whether the identification data corresponding to the corresponding associated target is absolutely required. If it is absolutely required, it is regarded as an associated necessary target; each associated necessary target is determined, and the auxiliary effect value of each associated necessary target is marked as 10; the effect difference of each associated target is simulated through the experimental demand data to obtain the auxiliary effect value corresponding to each associated target. The auxiliary effect value range is [1, 10], and the minimum can only be 1; that is, according to the combination of the corresponding data of each associated target and the identification target, the experimental effect is determined according to the corresponding experimental historical data, and compared with the best experimental results to determine the effect difference between them. The auxiliary effect value 10 is discounted according to the reduced proportion to form its corresponding effect auxiliary value; it can also be set by presetting the standards corresponding to different auxiliary effect values and then combining them with the interpolation method.
[0085] In one embodiment, the method for comparing each feature unit set is:
[0086] Identify the identification features corresponding to each feature unit set, combine and compare the identification features under the corresponding feature unit set, determine the identification feature combination that is unique to the feature unit set compared with other feature unit sets, and use it as the candidate image recognition combination for the feature unit set; perform cyclic analysis to obtain the candidate image recognition combinations; or directly combine the identification features within each feature unit set to determine the various combinations it has, compare the combinations, and determine the combination that belongs only to itself, and use it as the candidate image recognition combination.
[0087] Perform recognition estimation and comparison on each candidate image recognition combination, that is, compare the recognition efficiency of each candidate image recognition combination according to the estimation, and select the highest efficiency as its corresponding image recognition feature combination based on the differences in recognition speed and time corresponding to different recognition features.
[0088] The recognition module is used to perform experimental recognition, collect images of biological materials, obtain collected images, recognize the collected images according to the image recognition feature combination of each recognition target, and obtain target recognition data of each recognition feature;
[0089] Understand the current remaining data service resources, such as computing power and other resources in real time, mark them as service idle resources, make service estimates for each associated target based on the service idle resources and the priority of each associated target, determine each associated target that can provide identification service, mark them as associated service targets, identify the collected images based on the image recognition feature combination corresponding to the associated service target, and obtain the target recognition data of each associated service target; dynamically adjust the identification of each associated target based on the situation of service idle resources; and display the recognition data of each target to the user.
[0090] The analysis module is used to analyze each target recognition data to obtain corresponding experimental recognition analysis results, specifically to analyze based on existing analysis technology to determine the corresponding experimental results; and to display the experimental recognition analysis results to the user.
[0091] The above formulas are all calculated by removing dimensions and taking their numerical values. The formula is a formula that is closest to the actual situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and preset thresholds in the formula are set by technicians in this field according to actual conditions or obtained by simulating a large amount of data.
[0092] The above embodiments are only used to illustrate the technical method of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.
Claims
1. A biomaterial experiment identification system based on computer vision, characterized in that: Includes material database, requirement module, identification module and analysis module; The material database is used to store biomaterial data of various biomaterials, wherein the biomaterial data includes various candidate targets corresponding to the biomaterials and a set of identification features corresponding to each candidate target; The requirement module analyzes the user's experimental requirements, determines the biomaterial and identifies the target, and marks the biomaterial as the target material; matching corresponding biological material data from the material database according to the target material; Determining a set of identification features of the identification target based on the biomaterial data; Analyze the identified target to determine the associated targets and the priority of each associated target; Determine each reference target based on the biomaterial data, the identification target, and the associated target, and identify a set of identification features of each reference target based on the biomaterial data; Acquire experimental detail data in real time, analyze each of the recognition feature sets based on the experimental detail data, determine feature unit sets corresponding to the recognition target, the associated target, and the reference target, respectively; compare each feature unit set to determine the image recognition feature combination corresponding to the recognition target and the associated target, respectively; The recognition module is used to perform experimental recognition, collect images of biological materials, obtain collected images, recognize the collected images according to the image recognition feature combination of each recognition target, and obtain target recognition data of each recognition feature; Real-time monitoring of service idle resources, determining each associated service target based on the service idle resources and the priority of each associated target, identifying the captured image based on the image recognition feature combination corresponding to the associated service target, and obtaining target recognition data of each associated service target; The analysis module is used to analyze each target recognition data to obtain corresponding experimental recognition analysis results.
2. A biomaterial experiment identification system based on computer vision according to claim 1, characterized in that: The methods for establishing a material database include: obtaining an experimental range, and setting a biomaterial screening feature according to the experimental range; Acquire each candidate material in real time, screen each candidate material according to the biomaterial screening characteristics, and obtain each reserve material; Identifying each candidate target of each reserve material, performing identification feature analysis on each candidate target, and obtaining an identification feature set corresponding to each candidate target; Organize the reserve materials, candidate targets and identification feature sets into biological material data; A database is established, and the data of each biological material is input into the database for storage, and the database is marked as a material database.
3. A biomaterial experiment identification system based on computer vision according to claim 2, characterized in that: Methods for screening candidate materials based on biomaterial screening characteristics include: A screening judgment model is established, and the expression of the screening judgment model is: Where: (CT, DX) is the input data, CT is the biomaterial screening feature; DX is the candidate material for screening evaluation; the output data is the screening evaluation value SP(CT, DX), which is 1 or 0; DX→CT indicates that the candidate material for screening evaluation meets the requirements of the biomaterial screening feature; The screening characteristics of each candidate material and biomaterial are analyzed by the screening judgment model to obtain a screening evaluation value corresponding to each candidate material, and each candidate material with the screening evaluation value equal to 1 is marked as a reserve material.
4. The computer vision-based biomaterial experiment identification system according to claim 1, characterized in that: Methods for analyzing identified targets include: Acquire experimental requirements, acquire historical experimental identification data according to the experimental requirements, integrate the experimental requirements and the historical experimental identification data into a training set, establish an association analysis model based on the training set, and perform analysis using the association analysis model to obtain associated targets corresponding to the identified target; Collecting association data corresponding to each of the association targets, determining each identification target associated with the association target and corresponding association probabilities and experimental auxiliary effect data based on the association data; and evaluating corresponding auxiliary effect values based on the experimental auxiliary effect data; Marking the upper limit of the auxiliary effect value as the upper limit value; marking the associated target with the auxiliary effect value of the upper limit value as the associated necessary target; sorting the associated necessary targets in descending order according to the associated probability corresponding to the upper limit value to obtain a first priority sequence; According to the formula Calculate the priority value of each remaining associated target; Where: PY is the priority value; i represents the identification target associated with the corresponding associated target, i = 1, 2, ..., n, n is a positive integer; gi is the association probability; e is a natural constant; FGi is the auxiliary effect value; Adding each of the associated targets to the first sequence in descending order of priority; The priority of each associated target is determined according to the first sequence.
5. The computer vision-based biomaterial experiment identification system according to claim 4, characterized in that: Methods for evaluating the corresponding auxiliary effect value based on experimental auxiliary effect data include: Determine the best experimental results based on experimental requirements and historical experimental data, and determine the required data and necessary data for each experiment based on the best experimental results; Conduct a necessity analysis on each related target based on the necessary experimental data and identification targets, and determine the related necessary targets; mark the auxiliary effect value of each related necessary target as the upper limit value; The effect differences of each related target are simulated through experimental demand data to obtain the auxiliary effect value corresponding to each related target.
6. The computer vision-based biomaterial experiment identification system according to claim 5, characterized in that: The auxiliary effect value range is [1, 10].
7. The computer vision-based biomaterial experiment identification system according to claim 1, characterized in that: Methods for comparing feature unit sets include: Identify each identification feature corresponding to each feature unit set, perform combination comparison and analysis on each identification feature, and determine each candidate image recognition combination corresponding to each feature unit set; The recognition efficiency of each candidate image recognition combination is compared to determine the image recognition feature combinations corresponding to the recognition target and the associated target respectively.