CN115858846A - Deep learning-based skier image retrieval method and system - Google Patents
Deep learning-based skier image retrieval method and system Download PDFInfo
- Publication number
- CN115858846A CN115858846A CN202310120051.7A CN202310120051A CN115858846A CN 115858846 A CN115858846 A CN 115858846A CN 202310120051 A CN202310120051 A CN 202310120051A CN 115858846 A CN115858846 A CN 115858846A
- Authority
- CN
- China
- Prior art keywords
- image
- target
- retrieval
- image set
- identification
- Prior art date
- Legal status (The legal status 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 status listed.)
- Granted
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Image Analysis (AREA)
Abstract
The invention provides a skier image retrieval method and system based on deep learning, which relate to the technical field of image retrieval, and are used for constructing a target detection image set and training a target detection model; configuring an image preprocessing space to carry out image preprocessing and generating a re-identification image set; the method comprises the steps of training and generating a feature extraction model based on an Object-reiD network framework, integrating and regularly generating an image feature processing module, carrying out retrieval and identification based on a retrieval request, obtaining target features, traversing a storage database to determine an image retrieval result, solving the technical problems that the image retrieval accuracy has certain deviation and is missed and mistakenly detected due to the fact that the image feature extraction is single and the matching degree is insufficient in the prior art due to the fact that the intelligence and the rigor of the retrieval method for the images of skiers are insufficient, achieving the complete efficient and accurate identification of the image features and completing the high-matching image retrieval by constructing the image feature processing module and embedding a plurality of functional models.
Description
Technical Field
The invention relates to the technical field of image retrieval, in particular to a skier image retrieval method and system based on deep learning.
Background
The popularization of electronic mobile devices has led to the generation of a huge amount of images, and image retrieval technology is in return, so that image retrieval is a main problem to be solved. For example, a large number of ski images are shot, and how to efficiently and accurately search the required images is the current solution direction. In order to facilitate the retrieval of related images of different skiers, at present, with the updating and iteration of the image retrieval technology, image retrieval is mainly performed by technologies such as image searching and pedestrian re-identification, and compared with the traditional retrieval method, the retrieval effect is improved to a certain extent in the direction of retrieval efficiency, but certain defects still exist, the expected retrieval requirements cannot be met, and further technical innovation is needed.
In the prior art, the method for retrieving the images of the skiers is insufficient in intelligence and rigor, so that the extraction of the image features is single and the matching degree is insufficient, and the image retrieval accuracy has certain deviation, and the missed detection and the false detection are caused.
Disclosure of Invention
The application provides a skier image retrieval method and system based on deep learning, which are used for solving the technical problems of missed detection and false detection caused by certain deviation of image retrieval accuracy due to the fact that the image characteristics are extracted only singly and the matching degree is not enough by the aid of the method for retrieving the images of skiers in the prior art, which is not enough in intelligence and rigor.
In view of the above, the present application provides a skier image retrieval method and system based on deep learning.
In a first aspect, the present application provides a deep learning-based skier image retrieval method, including:
connecting the Internet of things, collecting a skiing image set by taking a multi-dimensional scene as an index direction, wherein the skiing image set comprises an image storage position and a storage path, and constructing a target detection image set;
extracting an identification positioning image set based on the target detection image set, and training a target detection model, wherein the identification positioning image set is provided with a skier position framing mark;
configuring an image preprocessing space, and performing image preprocessing on the identification positioning image set based on the image preprocessing space to generate a re-identification image set;
based on the re-recognition image set, training and generating a feature extraction model based on an Object-reiD network framework;
constructing an image feature processing module based on the target detection model, the image preprocessing space and the feature extraction model;
acquiring a retrieval request of a target user, and retrieving and identifying a target image according to the image feature processing module to acquire target features;
traversing a storage database, and determining an image retrieval result based on the target feature, wherein the image retrieval result is expressed as a retrieval image-storage location-storage path, and the storage location and the storage path are additional output information.
In a second aspect, the present application provides a deep learning based skier image retrieval system, the system comprising:
the image set construction module is used for connecting the Internet of things, collecting a skiing image set by taking a multi-dimensional scene as an index direction, and constructing a target detection image set by using image storage positions and storage paths;
the target detection model training module is used for extracting an identification positioning image set based on the target detection image set and training a target detection model, wherein the identification positioning image set is provided with a skier position frame selection mark;
the image preprocessing module is used for configuring an image preprocessing space, and carrying out image preprocessing on the identification positioning image set based on the image preprocessing space to generate a re-identification image set;
a feature extraction model training module, wherein the feature extraction model training module is used for generating a feature extraction model based on Object-reiD network framework training based on the re-recognition image set;
a construction module for constructing an image feature processing module based on the target detection model, the image pre-processing space and the feature extraction model;
the characteristic acquisition module is used for acquiring a retrieval request of a target user, and retrieving and identifying a target image according to the image characteristic processing module to acquire target characteristics;
and the result determining module is used for traversing the storage database and determining an image retrieval result based on the target characteristics, wherein the image retrieval result is expressed as a retrieval image-storage position-storage path, and the storage position and the storage path are additional output information.
One or more technical solutions provided in the present application have at least the following technical effects or advantages:
according to the skier image retrieval method based on deep learning, the internet of things is connected, a multi-dimensional scene is used as an index direction to collect a ski image set, a target detection image set is constructed, an identification positioning image set is extracted to train a target detection model, and the identification positioning image set is provided with a skier position frame selection mark; configuring an image preprocessing space, and performing image preprocessing on the identification positioning image set to generate a re-identification image set; based on the re-recognition image set, training and generating a feature extraction model based on an Object-reiD network framework; integrating and normalizing the target detection model, the image preprocessing space and the feature extraction model to construct an image feature processing module; acquiring a retrieval request of a target user, and retrieving and identifying a target image based on the image feature processing module to acquire target features; the method comprises the steps of traversing a storage database, determining an image retrieval result based on the target feature, solving the technical problems of missing detection and false detection caused by certain deviation of image retrieval accuracy due to insufficient intelligence and rigor of the retrieval method for the images of skiers in the prior art, and realizing the complete, efficient and accurate identification of the image features and completing the high-matching image retrieval by constructing an image feature processing module and embedding a plurality of functional models.
Drawings
FIG. 1 is a schematic flow chart of a deep learning-based skier image retrieval method provided by the present application;
FIG. 2 is a schematic diagram illustrating a process of obtaining a re-identified image set in a deep learning-based skier image retrieval method according to the present application;
FIG. 3 is a schematic diagram illustrating an image retrieval result obtaining process in a deep learning-based skier image retrieval method according to the present application;
fig. 4 is a schematic structural diagram of a deep learning-based skier image retrieval system according to the present application.
Description of reference numerals: the system comprises an image set construction module 11, an object detection model training module 12, an image preprocessing module 13, a feature extraction model training module 14, a construction module 15, a feature acquisition module 16 and a result determination module 17.
Detailed Description
The method and the system for retrieving the images of the skiers based on the deep learning are provided, a target detection image set is constructed, and an identification positioning image set is extracted to train a target detection model; configuring an image preprocessing space to carry out image preprocessing and generating a re-identification image set; the method comprises the steps of training and generating a feature extraction model based on an Object-reiD network framework, integrating and regularly generating an image feature processing module, carrying out retrieval and identification based on a retrieval request, obtaining target features, traversing a storage database to determine an image retrieval result, and solving the technical problems of missed detection and false detection caused by certain deviation of image retrieval accuracy due to the fact that the image feature extraction is single and the matching degree is insufficient in the prior art due to the fact that the intelligence and the rigor of the retrieval method for images of skiers are insufficient.
Example 1
As shown in fig. 1, the present application provides a deep learning-based skier image retrieval method, including:
step S100: connecting the Internet of things, collecting a skiing image set by taking a multi-dimensional scene as an index direction, wherein the skiing image set comprises an image storage position and a storage path, and constructing a target detection image set;
specifically, mass images are generated at all times due to popularization of electronic mobile equipment, an image retrieval technology is developed at the same time, a large number of skiing images are shot, and in order to facilitate related image retrieval of different skiers and improve image retrieval accuracy and retrieval efficiency, the application provides a skier image retrieval method based on deep learning. Specifically, the method comprises the steps of connecting the Internet of things, determining a multi-dimensional acquisition scene comprising various weather conditions, indoor and outdoor time axes and the like, acquiring skiing images comprising skiing images of a plurality of different target characters, determining storage positions and storage paths of the images, identifying the acquired images, and generating the target detection image set. The acquisition of the target detection image set provides basic source data for subsequent model training.
Further, step S100 of the present application further includes:
step S110: connecting the Internet, and collecting a skiing image set under a multi-dimensional scene as a target image set;
step S120: performing multi-stage light color adjustment and receptive field adjustment on the target image set to determine an augmented image set;
step S130: and integrating and warping the augmented image set and the target image set to generate the target detection image set.
Specifically, the internet is an auxiliary system tool for acquiring a skiing image, is connected with the internet, determines a multidimensional acquisition scene, such as sunny days, snow days, fog days, evenings, outdoors and the like, and performs related image acquisition by taking the multidimensional acquisition scene as an index direction of the skiing image, and classifies and integrates the acquired image based on a target person, wherein the target person is a skier in the image, and the target image set is obtained and comprises a plurality of skiing types. And setting a multi-level light color adjustment scale and a light field adjustment scale, for example, adjusting a light source direction, brightness, a color gamut, a view field of a target person, and the like, adjusting the target image set for multiple times, and taking an adjusted image as the augmented image set. And further adding the augmented image set into the target image set to generate the target detection image set. By image augmentation, the generalization degree of the image can be effectively improved, and the completeness of the image information is improved.
Step S200: extracting an identification positioning image set based on the target detection image set, and training a target detection model, wherein the identification positioning image set is provided with a skier position frame selection mark;
further, said extracting a set of identification positioning images based on said set of target detection images, and training a target detection model, step S200 of the present application further includes:
step S210: extracting the identified positioning image set based on the target detection image set;
step S220: taking the target detection image set as hierarchical identification information, taking the identification positioning image set as hierarchical decision information, and constructing a sample data set, wherein the target detection image set corresponds to the identification positioning image set one by one;
step S230: dividing the sample data set into K groups, and acquiring K groups of sample data subsets;
step S240: and based on a K-fold intersection method, dividing the K groups of sample data subsets into a training set and a test set, and training the target detection model.
Specifically, the image location of the skier is identified based on the target detection image, the image area of the target person, i.e., the skier, is determined based on a predetermined framing ratio, the figure distribution ratio of the framing area is set, the framing area is determined based on the size of the image target person, and the skier location framing mark is performed as the marker positioning image set.
Further, the identification positioning image set and the target detection image set are extracted, the target detection image set is used as hierarchical identification information, the identification positioning image set is used as hierarchical decision information, the identification positioning image set and the target detection image set are mapped and correspond to each other, so that the frame selection region identification and the size positioning of the target person are carried out, and the sample data set is constructed. And equally dividing the sample data set into K groups, and determining K groups of sample data subsets. And further based on the K groups of sample data subsets, taking K-1 groups of sample data subsets as a training set, taking 1 group of sample data subsets as a test set, carrying out neural network training based on the training set, obtaining the trained target detection model, further inputting the test set into the target detection model, carrying out deviation comparison on an obtained model output result and the identification positioning image in the test set, and obtaining the positioning deviation degree. Similarly, the sample amount of the test set and the sample amount of the training set are guaranteed to be unchanged, samples in the test set and the training set are adjusted, each group of sample data subsets can be used as the test set to perform model training result testing, deviation degree positioning is performed respectively, a plurality of trained target detection models are determined, deviation degree comparison is further performed, the minimum deviation degree is determined, the corresponding training model is used as the finally determined target detection model, the operation mechanism of the model can be effectively guaranteed to be better, and detection accuracy is improved.
Step S300: configuring an image preprocessing space, and performing image preprocessing on the identification positioning image set based on the image preprocessing space to generate a re-identification image set;
further, as shown in fig. 2, in the step S300 of generating the re-identification image set, the method further includes:
step S310: identifying the skier position frame selection mark based on the identification positioning image to obtain a target identification result;
step S320: determining target pixel coordinates to cut the identification positioning image based on the target identification result, and acquiring a target demand image;
step S330: performing definition judgment on the target demand image to obtain an image meeting a definition threshold value as a primary preprocessing image;
step S340: and performing image enhancement on the primary preprocessing image to serve as the re-identification image set.
Specifically, the image preprocessing space, that is, the operation space for image adjustment is configured, and the image preprocessing process is determined and embedded in the image preprocessing space. Specifically, the identification positioning image is extracted, and skier position frame selection mark recognition is performed in the image preprocessing space, so as to determine an image clipping region as the target recognition result. And based on the target recognition result, carrying out target pixel coordinate positioning, improving positioning accuracy, carrying out image cutting on the basis, and reserving a frame selection area, namely the area where the target person is located, integrating the cut image to be used as the target demand image.
Furthermore, a definition threshold value, namely a critical limit value for qualified judgment of image definition, is set. And checking the target demand image and the definition threshold, providing an image with definition smaller than the definition threshold, wherein the image has lower definition to cause identification deviation, and taking the image with definition larger than or equal to the definition threshold as the primary preprocessed image. And further carrying out contour and contrast processing on the primary preprocessed image, carrying out color gamut adjustment, improving the visual effect of the image, improving the prominence of information, and taking the enhanced primary preprocessed image as the re-recognition image, thereby providing a foundation for the subsequent image feature recognition, extraction and tamping.
Step S400: based on the re-recognition image set, training and generating a feature extraction model based on an Object-reiD network framework;
further, the step S400 of generating a feature extraction model based on Object-reID network framework training based on the re-identified image set further includes:
step S410: performing target region edge detection based on the re-identification image set, and determining a region set to be identified;
step S420: performing image convolution analysis based on the region set to be identified, and determining an image convolution feature set;
step S430: taking the Circle loss function as the loss function of the Object-Reid network framework;
step S440: and inputting an Object-Reid network framework according to the image convolution characteristic set and the region set to be identified, and performing supervised training until the Circle loss function converges to generate the characteristic extraction model.
Specifically, the re-recognition image is extracted, edge detection is performed on a target area, the target area is a coverage area of the re-recognition image, edge detection is performed to determine image framing edges, so that area positioning recognition of the re-recognition image is determined, recognition errors are caused on the surface, and a detection result is used as the area set to be recognized. And performing image convolution analysis on the to-be-identified region set, determining corresponding convolution characteristics by calculating convolution kernels, further performing summary fusion on the convolution characteristics of the to-be-identified region set and the target person, and determining representative convolution characteristics to serve as an even image convolution characteristic set. The Circle loss function is further used as the loss function of the Object-Reid network framework, and the contribution of the positive and negative samples to respective gradients is controlled by adding a weight between the positive and negative samples, so that the model has higher discrimination on similarity discrimination, and the problem of low discrimination of images of skiers caused by the same clothing is effectively improved.
Further, inputting the image convolution feature set and the region set to be identified into an Object-Reid network framework for training, and performing supervised training until a Circle loss function converges to generate the feature extraction model. And image feature recognition is carried out based on the feature extraction model, so that the output feature map can extract high-dimensional semantic information and rich spatial information. For the input skiing image, the abstract high-dimensional features of the skiing image can be extracted, and the spatial information features determining the body form and the action posture of the skiing image can be reserved. The similarity characteristics between different images of the same person and the differentiation characteristics between different persons are effectively improved, and the accuracy and the representativeness of characteristic extraction are improved.
Step S500: constructing an image feature processing module based on the target detection model, the image preprocessing space and the feature extraction model;
step S600: acquiring a retrieval request of a target user, and retrieving and identifying a target image according to the image feature processing module to acquire target features;
specifically, the method comprises the steps of collecting a skiing image, detecting, processing and training, carrying out functional modeling to obtain the target detection model, the image preprocessing space and the feature extraction model, placing the image preprocessing space in the target detection model, placing the image preprocessing space in the feature extraction model, generating the image feature processing module by carrying out link combination, and using the image feature processing module as an auxiliary tool for image feature analysis.
Further, with the reception of the retrieval request of the target user, the retrieval request is image retrieval application information, and the retrieval request of the target user is a retrieval performer, and the retrieval request further includes retrieval additional information, which is an example detection image, that is, an image example of a skier exists. The retrieval request is input into the image feature processing module, skiers are detected and positioned based on the target detection model, a positioning result is transmitted into the image preprocessing space, image cutting, screening and enhancement processing are carried out, the processed image is further transmitted to the feature extraction model, image convolution feature recognition and extraction are carried out based on a model training mechanism, a feature extraction result is used as the target feature, the target feature is reference feature information of image retrieval to be carried out, the target feature is used as an index direction to carry out relevant image detection and identification, the target feature is matched with a retrieval person, and the acquisition of the target feature with retrieval features of representative discrimination provides a basis for subsequent image retrieval.
Step S700: traversing a storage database, and determining an image retrieval result based on the target feature, wherein the image retrieval result is expressed as a retrieval image-storage location-storage path, and the storage location and the storage path are additional output information.
Further, as shown in fig. 3, step S700 of the present application further includes:
step S710: traversing a storage database, and performing similarity matching based on the target characteristics to obtain a similar sequence;
step S720: configuring a primary similarity threshold and a retrieval expected value;
step S730: determining an image sequence meeting the primary similarity threshold as an image sequence to be evaluated based on the similar sequence;
step S740: judging whether the magnitude of the image sequence to be evaluated is larger than the retrieval expected value or not;
step S750: if so, adjusting the primary similarity threshold value, and judging the threshold value to finish the updating of the image sequence to be evaluated;
step S760: and outputting an image retrieval result until the magnitude of the image sequence to be evaluated is less than or equal to the retrieval expected value.
Specifically, the storage database is obtained, which stores abundant skiing images, and there are various storage locations and storage paths of the images. Traversing the storage database, taking the target characteristics as matching identification information, performing similarity matching on each image in the storage database and the target characteristics, determining the similarity of a plurality of images, further sequencing the similarity of the plurality of images from large to small, synchronously sequencing corresponding images, and generating the similar sequence. And further configuring the primary similarity threshold and the retrieval expected value, wherein the primary similarity threshold is a similarity critical value for screening images, and the retrieval expected value is a critical value limited by the number of the screened images. And intercepting the similarity sequence based on the primary similarity threshold, performing image reverse matching based on an interception result, and determining an image sequence meeting the similarity threshold as the image sequence to be evaluated.
Further, determining the number of the images of the image sequence to be evaluated, acquiring a sequence quantity value, and judging whether the sequence quantity value is greater than the retrieval expected value. When the retrieval expectation value is less than or equal to the retrieval expectation value, the similar sequence is shown to meet the retrieval requirement, and the similar sequence is used as the image sequence to be evaluated; when the number of the retrieval images is larger than the expected retrieval value, the number of the retrieval images meeting the primary similarity threshold is more, and the images with higher similarity are determined as far as possible on the basis of ensuring that the number of the retrieval images reaches the standard, so that the retrieval accuracy is further improved. Specifically, the primary similarity threshold is adjusted up, for example, the primary similarity threshold is adjusted up by one percentage point, the interception and magnitude determination of the similar sequence are further performed, when the primary similarity threshold is still greater than the retrieval expected value, the primary similarity threshold is adjusted up again, the adjustment of the primary similarity threshold is stopped until the finally determined magnitude of the image sequence is less than or equal to the retrieval expected value, the image sequence to be evaluated is updated, the finally adjusted and determined image sequence to be evaluated is used as the image retrieval result, the storage position and storage path of each image in the image retrieval result are determined at the same time, the image retrieval result is used as additional output information, the image retrieval result is labeled, and a retrieval image-storage position-storage path sequence is generated and output. The image retrieval method can effectively guarantee the contact degree of the image retrieval result within the retrieval requirement, and reduce the possibility of abnormal retrieval deviation.
Example 2
Based on the same inventive concept as the deep learning based skier image retrieval method in the foregoing embodiment, as shown in fig. 4, the present application provides a deep learning based skier image retrieval system, which includes:
the image set building module 11 is used for connecting the Internet of things, collecting a skiing image set by taking a multi-dimensional scene as an index direction, and building a target detection image set, wherein the skiing image set comprises an image storage position and a storage path;
a target detection model training module 12, wherein the target detection model training module 12 is configured to extract a marker positioning image set based on the target detection image set, and train a target detection model, wherein the marker positioning image set has a skier position framing mark;
the image preprocessing module 13 is configured to configure an image preprocessing space, and perform image preprocessing on the identifier positioning image set based on the image preprocessing space to generate a re-identification image set;
a feature extraction model training module 14, wherein the feature extraction model training module 14 is configured to generate a feature extraction model based on Object-reID network framework training and based on the re-recognition image set;
a construction module 15, wherein the construction module 15 is configured to construct an image feature processing module based on the target detection model, the image preprocessing space and the feature extraction model;
the characteristic acquisition module 16 is used for acquiring a retrieval request of a target user, and retrieving and identifying a target image according to the image characteristic processing module to acquire target characteristics;
a result determining module 17, wherein the result determining module 17 is configured to traverse a storage database, and determine an image retrieval result based on the target feature, wherein the image retrieval result is expressed as a retrieval image-storage location-storage path, and the storage location and the storage path are additional output information.
Further, the system further comprises:
the image acquisition module is used for connecting the Internet, acquiring a skiing image set under a multi-dimensional scene as a target image set;
the image adjusting module is used for carrying out multi-stage light color adjustment and receptive field adjustment on the target image set to determine an augmented image set;
and the image set generation module is used for integrating and normalizing the augmented image set and the target image set to generate the target detection image set.
Further, the system further comprises:
the edge detection module is used for carrying out edge detection on a target area based on the re-identification image set and determining an area set to be identified;
the characteristic determining module is used for performing image convolution analysis based on the to-be-identified region set to determine an image convolution characteristic set;
a function embedding module for taking the Circle loss function as the loss function of the Object-Reid network framework;
and the model generation module is used for inputting an Object-Reid network framework according to the image convolution characteristic set and the to-be-identified region set, performing supervised training until a Circle loss function is converged, and generating the characteristic extraction model.
Further, the system further comprises:
an image set extraction module for extracting the identified positioning image set based on the target detection image set;
the sample construction module is used for constructing a sample data set by taking the target detection image set as hierarchical identification information and the identification positioning image set as hierarchical decision information, wherein the target detection image set and the identification positioning image set are in one-to-one correspondence;
the sample dividing module is used for dividing the sample data set into K groups and acquiring K groups of sample data subsets;
and the model training module is used for dividing the K groups of sample data subsets into a training set and a test set based on a K-fold intersection method and training the target detection model.
Further, the system further comprises:
the mark identification module is used for identifying the skier position frame selection mark based on the mark positioning image to obtain a target identification result;
the image cutting module is used for determining target pixel coordinates to cut the identification positioning image based on the target recognition result to obtain a target demand image;
the image screening module is used for judging the definition of the target demand image, acquiring an image meeting a definition threshold value and taking the image as a primary preprocessing image;
an image enhancement module to image enhance the primary pre-processed image as the re-identified image set.
Further, the system further comprises:
the sequence acquisition module is used for traversing a storage database, performing similarity matching based on the target characteristics and acquiring a similar sequence;
the parameter configuration module is used for configuring a primary similarity threshold value and a retrieval expected value;
the sequence screening module is used for determining an image sequence meeting the primary similarity threshold value as an image sequence to be evaluated based on the similar sequence;
the magnitude judgment module is used for judging whether the magnitude of the image sequence to be evaluated is greater than the retrieval expected value or not;
the sequence updating module is used for adjusting the primary similarity threshold value upwards and judging the threshold value to finish the updating of the image sequence to be evaluated when the magnitude value of the image sequence to be evaluated is greater than the retrieval expected value;
and the result output module is used for outputting an image retrieval result until the magnitude of the image sequence to be evaluated is less than or equal to the retrieval expected value.
In the present specification, through the foregoing detailed description of the method for retrieving an image of a skier based on deep learning, it is clear to those skilled in the art that a method and a system for retrieving an image of a skier based on deep learning in the present embodiment are provided.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (7)
1. A deep learning-based skier image retrieval method, the method comprising:
connecting the Internet of things, collecting a skiing image set by taking a multi-dimensional scene as an index direction, wherein the skiing image set comprises an image storage position and a storage path, and constructing a target detection image set;
extracting an identification positioning image set based on the target detection image set, and training a target detection model, wherein the identification positioning image set is provided with a skier position framing mark;
configuring an image preprocessing space, and performing image preprocessing on the identification positioning image set based on the image preprocessing space to generate a re-identification image set;
based on the re-recognition image set, training and generating a feature extraction model based on an Object-reiD network framework;
constructing an image feature processing module based on the target detection model, the image preprocessing space and the feature extraction model;
acquiring a retrieval request of a target user, and retrieving and identifying a target image according to the image feature processing module to acquire target features;
traversing a storage database, and determining an image retrieval result based on the target feature, wherein the image retrieval result is expressed as a retrieval image-storage location-storage path, and the storage location and the storage path are additional output information.
2. The method of claim 1, wherein the method comprises:
connecting the Internet, and acquiring a skiing image set under a multi-dimensional scene as a target image set;
performing multi-stage light color adjustment and receptive field adjustment on the target image set to determine an augmented image set;
and integrating and warping the augmented image set and the target image set to generate the target detection image set.
3. The method of claim 1, wherein generating a feature extraction model based on Object-reID web framework training based on the re-identified image set comprises:
performing target region edge detection based on the re-identification image set, and determining a region set to be identified;
performing image convolution analysis based on the region set to be identified, and determining an image convolution feature set;
taking the Circle loss function as the loss function of the Object-Reid network framework;
and inputting an Object-Reid network framework according to the image convolution characteristic set and the region set to be identified, and performing supervised training until the Circle loss function converges to generate the characteristic extraction model.
4. The method of claim 1, wherein extracting a set of marker-bearing images based on the set of target-bearing images, training a target-bearing model, comprises:
extracting the identification positioning image set based on the target detection image set;
taking the target detection image set as hierarchical identification information, taking the identification positioning image set as hierarchical decision information, and constructing a sample data set, wherein the target detection image set corresponds to the identification positioning image set one by one;
dividing the sample data set into K groups, and acquiring K groups of sample data subsets;
and based on a K-fold intersection method, dividing the K groups of sample data subsets into a training set and a test set, and training the target detection model.
5. The method of claim 1, wherein generating the set of re-identified images comprises:
identifying the skier position frame selection mark based on the identification positioning image to obtain a target identification result;
determining target pixel coordinates to cut the identification positioning image based on the target identification result, and acquiring a target demand image;
performing definition judgment on the target demand image to obtain an image meeting a definition threshold value as a primary preprocessing image;
and performing image enhancement on the primary preprocessing image to serve as the re-identification image set.
6. The method of claim 1, wherein the method comprises:
traversing a storage database, and performing similarity matching based on the target characteristics to obtain a similar sequence;
configuring a primary similarity threshold and a retrieval expected value;
determining an image sequence meeting the primary similarity threshold as an image sequence to be evaluated based on the similar sequence;
judging whether the magnitude of the image sequence to be evaluated is larger than the retrieval expected value or not;
if the primary similarity threshold is larger than the primary similarity threshold, the primary similarity threshold is adjusted upwards, threshold judgment is carried out, and the updating of the image sequence to be evaluated is completed;
and outputting an image retrieval result until the magnitude of the image sequence to be evaluated is less than or equal to the retrieval expected value.
7. A deep learning based skier image retrieval system, the system comprising:
the image set building module is used for connecting the Internet of things, collecting a skiing image set by taking a multi-dimensional scene as an index direction, and building a target detection image set, wherein the skiing image set comprises an image storage position and a storage path;
the target detection model training module is used for extracting an identification positioning image set based on the target detection image set and training a target detection model, wherein the identification positioning image set is provided with a skier position frame selection mark;
the image preprocessing module is used for configuring an image preprocessing space, and carrying out image preprocessing on the identification positioning image set based on the image preprocessing space to generate a re-identification image set;
a feature extraction model training module, wherein the feature extraction model training module is used for generating a feature extraction model based on Object-reiD network framework training based on the re-recognition image set;
a construction module for constructing an image feature processing module based on the target detection model, the image pre-processing space and the feature extraction model;
the characteristic acquisition module is used for acquiring a retrieval request of a target user, and retrieving and identifying a target image according to the image characteristic processing module to acquire target characteristics;
and the result determining module is used for traversing a storage database and determining an image retrieval result based on the target characteristic, wherein the image retrieval result is expressed as a retrieval image-storage position-storage path, and the storage position and the storage path are additional output information.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310120051.7A CN115858846B (en) | 2023-02-16 | 2023-02-16 | Skier image retrieval method and system based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310120051.7A CN115858846B (en) | 2023-02-16 | 2023-02-16 | Skier image retrieval method and system based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115858846A true CN115858846A (en) | 2023-03-28 |
| CN115858846B CN115858846B (en) | 2023-04-21 |
Family
ID=85658123
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310120051.7A Active CN115858846B (en) | 2023-02-16 | 2023-02-16 | Skier image retrieval method and system based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115858846B (en) |
Citations (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105354548A (en) * | 2015-10-30 | 2016-02-24 | 武汉大学 | Surveillance video pedestrian re-recognition method based on ImageNet retrieval |
| CN106778464A (en) * | 2016-11-09 | 2017-05-31 | 深圳市深网视界科技有限公司 | A kind of pedestrian based on deep learning recognition methods and device again |
| US20180144209A1 (en) * | 2016-11-22 | 2018-05-24 | Lunit Inc. | Object recognition method and apparatus based on weakly supervised learning |
| CN108537136A (en) * | 2018-03-19 | 2018-09-14 | 复旦大学 | The pedestrian's recognition methods again generated based on posture normalized image |
| CN109214430A (en) * | 2018-08-15 | 2019-01-15 | 天津大学 | A kind of recognition methods again of the pedestrian based on feature space topology distribution |
| CN109255052A (en) * | 2018-08-29 | 2019-01-22 | 浙江工业大学 | A kind of three stage vehicle retrieval methods based on multiple features |
| CN109635634A (en) * | 2018-10-29 | 2019-04-16 | 西北大学 | A kind of pedestrian based on stochastic linear interpolation identifies data enhancement methods again |
| CN109886242A (en) * | 2019-03-01 | 2019-06-14 | 中国科学院重庆绿色智能技术研究院 | A kind of method and system that pedestrian identifies again |
| CN109902590A (en) * | 2019-01-30 | 2019-06-18 | 西安理工大学 | Pedestrian's recognition methods again of depth multiple view characteristic distance study |
| CN109948628A (en) * | 2019-03-15 | 2019-06-28 | 中山大学 | A kind of object detection method excavated based on identification region |
| CN110110601A (en) * | 2019-04-04 | 2019-08-09 | 深圳久凌软件技术有限公司 | Video pedestrian weight recognizer and device based on multi-space attention model |
| CN110321965A (en) * | 2019-07-10 | 2019-10-11 | 腾讯科技（深圳）有限公司 | The method and device that the training method of object weight identification model, object identify again |
| CN110942012A (en) * | 2019-11-22 | 2020-03-31 | 上海眼控科技股份有限公司 | Image feature extraction method, pedestrian re-identification method, device and computer equipment |
| CN111126249A (en) * | 2019-12-20 | 2020-05-08 | 深圳久凌软件技术有限公司 | Pedestrian re-identification method and device combining big data and Bayes |
| CN111460914A (en) * | 2020-03-13 | 2020-07-28 | 华南理工大学 | Pedestrian re-identification method based on global and local fine-grained features |
| CN112001251A (en) * | 2020-07-22 | 2020-11-27 | 山东大学 | Pedestrian re-identification method and system based on combination of human body analysis and clothing color |
| US20210231440A1 (en) * | 2020-01-14 | 2021-07-29 | Tata Consultancy Services Limited | Systems and methods for performing inclusive indoor navigation |
| CN113326738A (en) * | 2021-05-06 | 2021-08-31 | 南京信息工程大学 | Pedestrian target detection and re-identification method based on deep network and dictionary learning |
| CN114547365A (en) * | 2022-02-22 | 2022-05-27 | 青岛海信网络科技股份有限公司 | Image retrieval method and device |
-
2023
- 2023-02-16 CN CN202310120051.7A patent/CN115858846B/en active Active
Patent Citations (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105354548A (en) * | 2015-10-30 | 2016-02-24 | 武汉大学 | Surveillance video pedestrian re-recognition method based on ImageNet retrieval |
| CN106778464A (en) * | 2016-11-09 | 2017-05-31 | 深圳市深网视界科技有限公司 | A kind of pedestrian based on deep learning recognition methods and device again |
| US20180144209A1 (en) * | 2016-11-22 | 2018-05-24 | Lunit Inc. | Object recognition method and apparatus based on weakly supervised learning |
| CN108537136A (en) * | 2018-03-19 | 2018-09-14 | 复旦大学 | The pedestrian's recognition methods again generated based on posture normalized image |
| CN109214430A (en) * | 2018-08-15 | 2019-01-15 | 天津大学 | A kind of recognition methods again of the pedestrian based on feature space topology distribution |
| CN109255052A (en) * | 2018-08-29 | 2019-01-22 | 浙江工业大学 | A kind of three stage vehicle retrieval methods based on multiple features |
| CN109635634A (en) * | 2018-10-29 | 2019-04-16 | 西北大学 | A kind of pedestrian based on stochastic linear interpolation identifies data enhancement methods again |
| CN109902590A (en) * | 2019-01-30 | 2019-06-18 | 西安理工大学 | Pedestrian's recognition methods again of depth multiple view characteristic distance study |
| CN109886242A (en) * | 2019-03-01 | 2019-06-14 | 中国科学院重庆绿色智能技术研究院 | A kind of method and system that pedestrian identifies again |
| CN109948628A (en) * | 2019-03-15 | 2019-06-28 | 中山大学 | A kind of object detection method excavated based on identification region |
| CN110110601A (en) * | 2019-04-04 | 2019-08-09 | 深圳久凌软件技术有限公司 | Video pedestrian weight recognizer and device based on multi-space attention model |
| CN110321965A (en) * | 2019-07-10 | 2019-10-11 | 腾讯科技（深圳）有限公司 | The method and device that the training method of object weight identification model, object identify again |
| CN110942012A (en) * | 2019-11-22 | 2020-03-31 | 上海眼控科技股份有限公司 | Image feature extraction method, pedestrian re-identification method, device and computer equipment |
| CN111126249A (en) * | 2019-12-20 | 2020-05-08 | 深圳久凌软件技术有限公司 | Pedestrian re-identification method and device combining big data and Bayes |
| US20210231440A1 (en) * | 2020-01-14 | 2021-07-29 | Tata Consultancy Services Limited | Systems and methods for performing inclusive indoor navigation |
| CN111460914A (en) * | 2020-03-13 | 2020-07-28 | 华南理工大学 | Pedestrian re-identification method based on global and local fine-grained features |
| CN112001251A (en) * | 2020-07-22 | 2020-11-27 | 山东大学 | Pedestrian re-identification method and system based on combination of human body analysis and clothing color |
| CN113326738A (en) * | 2021-05-06 | 2021-08-31 | 南京信息工程大学 | Pedestrian target detection and re-identification method based on deep network and dictionary learning |
| CN114547365A (en) * | 2022-02-22 | 2022-05-27 | 青岛海信网络科技股份有限公司 | Image retrieval method and device |
Non-Patent Citations (1)
| Title |
|---|
| 郭迎春等: "基于自适应融合网络的跨域行人重识别方法" * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115858846B (en) | 2023-04-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110443143B (en) | Multi-branch convolutional neural network fused remote sensing image scene classification method | |
| CN111259786B (en) | Pedestrian re-identification method based on synchronous enhancement of appearance and motion information of video | |
| CN111931684B (en) | Weak and small target detection method based on video satellite data identification features | |
| CN113705478B (en) | Mangrove single wood target detection method based on improved YOLOv5 | |
| CN111783576B (en) | Pedestrian re-identification method based on improved YOLOv3 network and feature fusion | |
| CN111898736B (en) | Efficient pedestrian re-identification method based on attribute perception | |
| CN111709311B (en) | Pedestrian re-identification method based on multi-scale convolution feature fusion | |
| CN111931637A (en) | Cross-modal pedestrian re-identification method and system based on double-current convolutional neural network | |
| CN113033520B (en) | Tree nematode disease wood identification method and system based on deep learning | |
| CN106557740B (en) | The recognition methods of oil depot target in a kind of remote sensing images | |
| CN110390308B (en) | Video behavior identification method based on space-time confrontation generation network | |
| CN110688955A (en) | Building construction target detection method based on YOLO neural network | |
| CN112668375A (en) | System and method for analyzing tourist distribution in scenic spot | |
| CN114359702A (en) | Method and system for identifying building violation of remote sensing image of homestead based on Transformer | |
| CN115620090A (en) | Model training method, low-illumination target re-recognition method and device and terminal equipment | |
| CN113536946B (en) | Self-supervision pedestrian re-identification method based on camera relationship | |
| CN111898510B (en) | Cross-modal pedestrian re-identification method based on progressive neural network | |
| CN112446305A (en) | Pedestrian re-identification method based on classification weight equidistant distribution loss model | |
| CN116311357A (en) | Double-sided identification method for unbalanced bovine body data based on MBN-transducer model | |
| CN115858846B (en) | Skier image retrieval method and system based on deep learning | |
| CN113780066B (en) | Pedestrian re-recognition method and device, electronic equipment and readable storage medium | |
| Peltomäki et al. | Evaluation of long-term LiDAR place recognition | |
| CN114627493A (en) | Gait feature-based identity recognition method and system | |
| CN115880740A (en) | Face living body detection method and device, computer equipment and storage medium | |
| CN114882224B (en) | Model structure, model training method, singulation method, device and medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |