CN110569918A - A kind of sample classification method and related device - Google Patents

Abstract

The embodiment of the present application provides a sample classification method and a related device, and the method can cluster pedestrian images according to mutual neighbor distance, neighbor ranking distance and absolute distance. In the embodiment of the present application, the same people are aggregated together, and a sample classification result can be generated according to the clustered images, the clustering effect is good, and samples suitable for training a pedestrian re-identification model can be obtained.

Description

A kind of sample classification method and related device

technical field

The present application relates to the technical field of artificial intelligence, in particular to a sample classification method and a related device.

Background technique

Person re-identification (Person re-identification, Person re-ID) is a research focus of computer vision in recent years. Given a monitored pedestrian image, retrieve the image of the pedestrian across devices. Due to the differences among different camera devices, the appearance of pedestrians is easily affected by clothing, scale, occlusion, posture and viewing angle, etc. Pedestrian re-identification is a research topic that is both valuable and challenging. Pedestrian re-identification technology may also be called pedestrian re-identification technology, hereinafter referred to as ReID technology.

At present, ReID technology is being widely used in business, security, transportation, finance and other fields. Data plays an important role in improving the performance of ReID models. For researchers, mining labeled identities from massive human body pictures is an important task. . Clustering technology plays a central role in the mining process. Grouping people with the same identity together can reduce the amount of subsequent manual labeling.

Since the face recognition technology is very mature, in the face database (labeled faces in the wild, LFW), the accuracy of the face recognition model is higher than the accuracy of human eye recognition, and the effect of clustering using face features is very good. it is good. However, the development of ReID technology is relatively slow compared with face recognition technology, and the model accuracy is still relatively weak. The clustering algorithm that performs well in face recognition often performs poorly in ReID technology, and the obtained samples often fail to meet the requirements, such as bad files and multiple files for one person, or for example, the same person does not The situation of coming together. Therefore, it is difficult to obtain suitable samples by applying the clustering algorithm in face recognition technology to ReID technology.

Contents of the invention

The embodiment of the present application provides a sample classification method and a related device, which solve the technical problem of poor clustering effect and difficulty in obtaining suitable samples in the prior art.

In the first aspect, the embodiment of the present application provides a sample classification method, including:

Acquiring a first set to be processed and a second set to be processed, the first set to be processed includes at least one first sample image, and the second set to be processed includes at least one second sample image;

Acquiring a first image feature vector and a second image feature vector, the first image feature vector has a first corresponding relationship with the first sample image, and the second image feature vector has a first corresponding relationship with the second sample image Two correspondences;

Obtaining the distance between the first set to be processed and the second set to be processed according to the first image feature vector and the second image feature vector, the distance includes absolute distance, mutual neighbor distance and neighbor sorting distance;

If the absolute distance satisfies the first set condition, the mutual neighbor distance satisfies the second set condition and the neighbor ranking distance satisfies the third set condition, then the first set to be processed and the second set to be processed are generated. The sample classification result corresponding to the set to be processed.

In the second aspect, the embodiment of the present application provides a device for classifying samples, including:

An acquiring unit, configured to acquire a first set to be processed and a second set to be processed, the first set to be processed includes at least one first sample image, and the second set to be processed includes at least one second sample image;

The acquiring unit is further configured to acquire a first image feature vector and a second image feature vector, the first image feature vector has a first corresponding relationship with the first sample image, and the second image feature vector and the The second sample image has a second corresponding relationship;

A processing unit, configured to obtain a distance between the first set to be processed and the second set to be processed according to the first image feature vector and the second image feature vector, the distance includes an absolute distance, Neighbor distance and neighbor sorting distance;

The processing unit is further configured to generate the first to-be-processed Set the sample classification results corresponding to the second set to be processed.

In a possible design, in an implementation manner of the second aspect of the embodiment of the present application, the processing unit is further configured to:

determining a cosine distance between the first sample image and the second sample image according to the first image feature vector and the second image feature vector;

An absolute distance between the first set to be processed and the second set to be processed is determined according to the cosine distance, and the absolute distance is a minimum value of the cosine distance.

In a possible design, in an implementation manner of the second aspect of the embodiment of the present application, the processing unit is further configured to:

Obtain a first absolute distance corresponding to the first set to be processed;

Sorting according to the size of the first absolute distance to obtain a first nearest neighbor sequence corresponding to the first set to be processed;

Obtain a second absolute distance corresponding to the second set to be processed;

Sorting according to the size of the second absolute distance to obtain a second nearest neighbor sequence corresponding to the second set to be processed;

Determine the first set to be processed and the second set to be processed according to the sequence number of the first set to be processed in the second nearest neighbor sequence and the sequence number of the second set to be processed in the first nearest neighbor sequence Handles mutual proximity distances between collections.

In a possible design, in an implementation manner of the second aspect of the embodiment of the present application, the processing unit is further configured to:

Determine a neighbor ranking distance between the first set to be processed and the second set to be processed according to the first nearest neighbor sequence and the second nearest neighbor sequence.

In a possible design, in an implementation manner of the second aspect of the embodiment of the present application, the processing unit is further configured to:

Acquiring a target set, where the target set includes the first to-be-processed set and the second to-be-processed set;

determining a single feature vector based on the first image feature vector and the second image feature vector;

calculating the similarity between the target sets according to the single feature vector;

Sending the target set whose similarity is smaller than a set threshold to the terminal device, so that the terminal device displays the target set.

In a possible design, in an implementation manner of the second aspect of the embodiment of the present application, the processing unit is further configured to:

selecting a sample image in said target set;

sending the sample image to a terminal device, so that the terminal device displays the sample image.

In a possible design, in an implementation manner of the second aspect of the embodiment of the present application, the processing unit is further configured to:

Acquiring annotation information, where the annotation information has an association relationship with the target set;

A secondary sample classification result corresponding to the target set is determined according to the annotation information.

In a third aspect, the embodiment of the present application provides a server, including:

One or more central processing units, memory, input and output interfaces, wired or wireless network interfaces, power supply;

The memory is a temporary storage memory or a persistent storage memory;

The central processing unit is configured to communicate with the memory, and the following steps are performed on the server:

Acquiring a first set to be processed and a second set to be processed, the first set to be processed includes at least one first sample image, and the second set to be processed includes at least one second sample image;

Acquiring a first image feature vector and a second image feature vector, the first image feature vector has a first corresponding relationship with the first sample image, and the second image feature vector has a first corresponding relationship with the second sample image Two correspondences;

Obtaining the distance between the first set to be processed and the second set to be processed according to the first image feature vector and the second image feature vector, the distance includes absolute distance, mutual neighbor distance and neighbor sorting distance;

If the absolute distance satisfies the first set condition, the mutual neighbor distance satisfies the second set condition and the neighbor ranking distance satisfies the third set condition, then the first set to be processed and the second set to be processed are generated. The sample classification result corresponding to the set to be processed.

In a possible design, in an implementation manner of the third aspect of the embodiment of the present application, the central processing unit is further configured to perform the following steps:

determining a cosine distance between the first sample image and the second sample image according to the first image feature vector and the second image feature vector;

An absolute distance between the first set to be processed and the second set to be processed is determined according to the cosine distance, and the absolute distance is a minimum value of the cosine distance.

In a possible design, in an implementation manner of the third aspect of the embodiment of the present application, the central processing unit is further configured to perform the following steps:

Determine the first nearest neighbor sequence corresponding to the first set to be processed and the second nearest neighbor sequence corresponding to the second set to be processed according to the absolute distance, the first nearest neighbor sequence being the first nearest neighbor sequence The two sample images are sorted according to the cosine distance, and the second nearest neighbor sequence is the sorted queue of the first sample image according to the cosine distance;

The first set to be processed and the second set to be processed are determined according to the sequence number of the first absolute sample image in the second nearest neighbor sequence and the sequence number of the first absolute sample image in the second nearest neighbor sequence. Handles mutual proximity distances between collections.

In a possible design, in an implementation manner of the third aspect of the embodiment of the present application, the central processing unit is further configured to perform the following steps:

Determine a neighbor ranking distance between the first set to be processed and the second set to be processed according to the first nearest neighbor sequence and the second nearest neighbor sequence.

In a possible design, in an implementation manner of the third aspect of the embodiment of the present application, the central processing unit is further configured to perform the following steps:

Acquiring a target set, where the target set includes the first to-be-processed set and the second to-be-processed set;

determining a single feature vector based on the first image feature vector and the second image feature vector;

calculating the similarity between the target sets according to the single feature vector;

Sending the target set whose similarity is smaller than a set threshold to the terminal device, so that the terminal device displays the target set.

In a possible design, in an implementation manner of the third aspect of the embodiment of the present application, the central processing unit is further configured to perform the following steps:

selecting a sample image in said target set;

sending the sample image to a terminal device, so that the terminal device displays the sample image.

In a possible design, in an implementation manner of the third aspect of the embodiment of the present application, the central processing unit is further configured to perform the following steps:

Acquiring annotation information, where the annotation information has an association relationship with the target set;

A secondary sample classification result corresponding to the target set is determined according to the annotation information.

In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, where instructions are stored in the computer-readable storage medium, and when the computer-readable storage medium is run on a computer, it causes the computer to execute the methods in the above aspects.

It can be seen from the above technical solutions that the embodiments of the present application have the following advantages:

The embodiment of the present application provides a sample classification method and a related device, which can cluster images of pedestrians by using mutual neighbor distance, neighbor sorting distance and absolute distance. In the embodiment of the present application, the same people are aggregated together, and a sample classification result can be generated according to the clustered images, the clustering effect is good, and samples suitable for training a pedestrian re-identification model can be obtained.

Description of drawings

FIG. 1 is a scene example diagram of the pedestrian re-identification technology in the embodiment of the present application;

Fig. 2 is an example figure of outliers in the embodiment of the present application;

Fig. 3 is a schematic diagram of a sample classification method provided in the embodiment of the present application;

Figure 4 is an example diagram of the first nearest neighbor sequence and the second nearest neighbor sequence in the embodiment of the present application;

FIG. 5 is an example diagram of the server calculating the neighbor sorting distance in the embodiment of the present application;

FIG. 6 is a diagram showing an example of a target set in the embodiment of the present application;

FIG. 7 is a schematic flow diagram of an optional embodiment in the embodiments of the present application;

FIG. 8 is a schematic flowchart of another optional embodiment in the embodiments of the present application;

FIG. 9 is an example diagram of a device for classifying samples in the embodiment of the present application;

FIG. 10 is a schematic structural diagram of a server provided by an embodiment of the present application.

Detailed ways

The embodiment of the present application provides a sample classification method and a related device, which solve the technical problem of poor clustering effect and difficulty in obtaining suitable samples in the prior art.

The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above drawings are used to distinguish similar objects, and not necessarily Used to describe a specific sequence or sequence. It is to be understood that the data so used are interchangeable under appropriate circumstances such that the embodiments of the application described herein can be practiced in sequences other than those illustrated or described herein. Furthermore, the terms "comprising" and "corresponding to" and any variations thereof, are intended to cover a non-exclusive inclusion, for example, a process, method, system, product or device comprising a sequence of steps or elements need not be limited to the expressly listed Instead, other steps or elements not explicitly listed or inherent to the process, method, product or apparatus may be included.

In the embodiments of the present application, words such as "exemplary" or "for example" are used as examples, illustrations or illustrations. Any embodiment or design scheme described as "exemplary" or "for example" in the embodiments of the present application shall not be interpreted as being more preferred or more advantageous than other embodiments or design schemes. Rather, the use of words such as "exemplary" or "such as" is intended to present related concepts in a concrete manner.

In order to make the description of the following embodiments clear and concise, a brief introduction of related technologies is first given:

Artificial Intelligence (AI) is a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technique of computer science that attempts to understand the nature of intelligence and produce a new kind of intelligent machine that can respond in a similar way to human intelligence. Artificial intelligence is to study the design principles and implementation methods of various intelligent machines, so that the machines have the functions of perception, reasoning and decision-making.

Artificial intelligence technology is a comprehensive subject that involves a wide range of fields, including both hardware-level technology and software-level technology. Artificial intelligence basic technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation/interaction systems, and mechatronics. Artificial intelligence software technology mainly includes several major directions such as computer vision technology, speech processing technology, natural language processing technology, and machine learning/deep learning.

Computer Vision Technology (Computer Vision, CV) Computer vision is a science that studies how to make machines "see". More specifically, it refers to machine vision that uses cameras and computers instead of human eyes to identify, track and measure targets. And further graphics processing, so that the computer processing becomes an image that is more suitable for human observation or sent to the instrument for detection. As a scientific discipline, computer vision studies related theories and technologies, trying to build artificial intelligence systems that can obtain information from images or multidimensional data. Computer vision technology usually includes image processing, image recognition, image semantic understanding, image retrieval, OCR, video processing, video semantic understanding, video content/behavior recognition, 3D object reconstruction, 3D technology, virtual reality, augmented reality, simultaneous positioning and maps Construction and other technologies also include common face recognition, fingerprint recognition, pedestrian re-identification and other biometric recognition technologies.

The solutions provided in the embodiments of the present application relate to technologies such as artificial intelligence pedestrian re-identification, and are specifically described through the following embodiments:

FIG. 1 is an example diagram of a scene of the pedestrian re-identification technology in the embodiment of the present application. The camera device may include but not limited to a camera, a monitoring device, a video recorder, etc., which is not limited in this embodiment of the present application. The camera device is connected to the server through a network, so that the camera device can transmit the captured images or videos to the server. After receiving these images or videos, the server can identify pedestrians in the images or videos through pedestrian re-identification technology (hereinafter referred to as ReID technology). The specific process is roughly as follows:

The server acquires image or video data in the camera device. If the server receives the video data transmitted by the camera device, the server may extract several frames of images in a sampling manner. Then, the server can cluster the acquired images, and classify all images corresponding to the same pedestrian into the same category. Exemplarily, if camera device A captures three images of pedestrian A, and camera device B also captures three images of pedestrian A, the server may classify these six images into the same category. Finally, the server identifies the identity of the pedestrian corresponding to each type of image based on recognition algorithms such as face recognition.

In the above process, the clustering algorithm for the server to cluster the images plays an important role in the whole process. The current hierarchical clustering algorithm and rank order clustering algorithm are difficult to obtain suitable samples, or it is difficult to obtain accurate samples. The following will briefly introduce the hierarchical clustering algorithm and the rank-order clustering algorithm:

The hierarchical clustering algorithm is implemented in the following way: first treat each object as a cluster, and then merge these atomic clusters into larger and larger clusters until all objects are in one cluster, or a certain terminal condition is satisfied. The feature expression ability of the ReID model using the hierarchical clustering algorithm is weak, the similarity of the same person crossing the camera, the front and back of the same person is low, and the similarity of different people's clothing is high. Using the hierarchical clustering method will generate a large number of bad files (different people's images are assigned to the same category) and multiple files for one person (images of the same person are assigned to multiple categories). Furthermore, termination conditions are not easy to set.

The clustering method based on rank order sorting distance is widely used in mobile phone photo album clustering. The rank order is based on a common phenomenon: two faces of the same person have many shared neighbors, but the neighbors of faces from different people are usually very different. When the rank order clustering algorithm is applied to ReID technology, the appearance of outliers will enlarge the sorting distance, resulting in the same person not being clustered together.

FIG. 2 is an example diagram of outliers in the embodiment of the present application. It can be seen that samples a, b, c, where a, b belong to the same person, c and a, b do not belong to the same person, if c is very close to one of a, b, and far away from the other, then c is an outlier point. For example, c is the first neighbor of a and is the 155th neighbor of b.

Therefore, the above-mentioned clustering algorithm fails to meet the requirements and cannot obtain reasonably classified samples.

In view of this, an embodiment of the present application provides a sample classification method to solve the above technical problem. FIG. 3 is a schematic diagram of a method for classifying samples provided by an embodiment of the present application. The method includes:

301. Acquire a first set to be processed and a second set to be processed, the first set to be processed includes at least one first sample image, and the second set to be processed includes at least one second sample image;

In the embodiment of the present application, the server first acquires the first set to be processed and the second set to be processed. The first set to be processed and the second set to be processed can also be referred to as classes in the clustering algorithm. For the convenience of description, the following unified The first to-be-processed set and the second to-be-processed set are referred to as class A and class B. It is understandable that the server not only obtains classes A and B, but also obtains other classes, such as classes C and D. In this embodiment of the present application, for the convenience of description, only the cases of class A and class B are described, and it should not be understood that the server only obtains class A and class B. Exemplarily, the collections to be processed acquired by the server include Class A (the first collection to be processed), Class B (the second collection to be processed), Class C, and Class D, etc. In this embodiment of the application, the server may determine the Whether it can be merged with Class B. If class A and class B can be merged, the server classifies the sample images in class A and class B into one large class. By analogy, the server can combine various categories to obtain several major categories, each of which includes several sample images, so as to realize the aggregation of sample images and reasonably classify the sample images, so that the server can Classified sample images for pedestrian identity recognition. For example, if the sample images in a large category obtained by the server are images of pedestrian A, then when the server recognizes pedestrians in the sample images under this large category, the recognition accuracy can be improved, and pedestrians under this category can be identified. The images are all images of pedestrian A.

Each category acquired by the server may include at least one sample image. In some embodiments, after the server acquires the image transmitted by the camera device or the image determined by the video, each image is processed as a sample image. In some embodiments, the server may treat each sample image as a class, and a class may include only one sample image. Exemplarily, when the server obtains the A sample image and the B sample image, the server initially classifies the A sample image into the category A, and preliminarily classifies the B sample image into the B category. It can be understood that, in the embodiment of the present application, the sample images in category A are the first sample images, and the sample images in category B are the second sample images.

302. Acquire a first image feature vector and a second image feature vector, the first image feature vector has a first corresponding relationship with the first sample image, and the second image feature vector has a second corresponding relationship with the second sample image;

In this embodiment of the present application, the server may extract features according to the first sample image to obtain the first image feature vector. Algorithms or models for extracting features may include, but are not limited to, convolutional neural network (CNN), human body pose (Pose) and skeleton key point (Skeleton) models, etc., which are not limited here. Similarly, the server may extract features according to the second sample image to obtain the second image feature vector.

It can be understood that the first image feature vector and the second image feature vector are used to represent the information of the pedestrian in the sample image.

303. Obtain the distance between the first set to be processed and the second set to be processed according to the first image feature vector and the second image feature vector, where the distance includes an absolute distance, a mutual neighbor distance, and a neighbor ranking distance;

In the embodiment of the present application, the server needs to calculate multiple distances, including absolute distance (also called the shortest distance), mutual neighbor distance and neighbor ranking distance. The three distances will be described in detail below.

In some embodiments, the server may calculate the absolute distance by first determining the cosine distance between the first sample image and the second sample image according to the first image feature vector and the second image feature vector. Then, the server determines the absolute distance between the first set to be processed and the second set to be processed according to the cosine distance, and the absolute distance is the minimum value of the cosine distance. The method for the server to calculate the cosine distance is not specifically limited in this embodiment of the application. The absolute distance calculated by the server can be calculated through the absolute distance calculation formula. The absolute distance calculation formula is:

d(C _i ,C _j )=min(f(x _m ,x _n ));

Among them, d(C _i , C _j ) is the absolute distance between class C _i and class C _j , x _m is the sample image in class C _i , x _n is the sample image in class C _j , f(x _m , x _n ) is a function to calculate the cosine distance between sample images.

Exemplarily, class A includes sample image a and sample image b, class B includes sample image c and sample image d, the calculated cosine distance between sample image a and sample image c is 1, and the cosine distance between sample image a and sample image d is If the distance is 2, the cosine distance between sample image b and sample image c is 3, and the cosine distance between sample image b and sample image d is 4, then the server determines that the absolute distance between class A and class B is 1.

In some embodiments, class A has only one sample image, and class B has only one sample image, then the absolute distance between class A and class B may be the cosine distance between the two sample images.

In some embodiments, the server may also determine the first absolute sample image and the second absolute sample image associated with the absolute distance. Exemplarily, if the cosine distance calculated by the server based on sample image a and sample image c is the smallest, the server uses the cosine distance as the absolute distance, and the sample image a associated with the absolute distance can be used as the first absolute sample image, and the cosine distance The distance-associated sample image c can be used as the second absolute sample image.

In some embodiments, the way in which the server can calculate the mutual neighbor distance is that the server first obtains the first absolute distance corresponding to the first set to be processed, and sorts the first absolute distance corresponding to the first set to be processed according to the size of the first absolute distance. adjacent sequence. Then, the server obtains the second absolute distance corresponding to the second to-be-processed set; sorts according to the size of the second absolute distance to obtain the second nearest neighbor sequence corresponding to the second to-be-processed set. Finally, the server determines the mutual neighbor distance between the first set to be processed and the second set to be processed according to the sequence number of the first set to be processed in the second nearest neighbor sequence and the sequence number of the second set to be processed in the first nearest neighbor sequence.

Exemplarily, the collection to be processed acquired by the server may include category A (first collection to be processed), category B (second collection to be processed), category C, category D, and so on. Then, the server can calculate the absolute distance (also called the first absolute distance) between class A and all classes, and sort all classes according to the absolute distance to obtain the first nearest neighbor sequence, as shown in FIG. 4 . FIG. 4 is an example diagram of a first nearest neighbor sequence and a second nearest neighbor sequence in an embodiment of the present application. Among them, _OA is the first nearest neighbor sequence, _OB is the second nearest neighbor sequence, and the classes in Figure 4 can include class A, class B, class C, class D, etc. Exemplarily, suppose the server calculates that the absolute distance between class A and class A is 1, the absolute distance between class A and class B is 0.1, the absolute distance between class A and class C is 0.8, and the absolute distance between class A and class D is 0.3, the server can obtain the first nearest neighbor sequence by sorting these absolute distances from large to small, as shown in Figure 4. In some other embodiments, the server may also sort by other methods, which are not limited here.

In some embodiments, the server may calculate the mutual neighbor distance between classes through a mutual neighbor distance calculation formula, and the mutual neighbor distance calculation formula is:

MN(C _i ,C _j )=m+n;

Among them, MN is the mutual neighbor distance, m is the serial number of the nearest neighbor sequence corresponding to class C _i in class C _j , and n is the serial number of the nearest neighbor sequence corresponding to class C _j in class C _i .

Exemplarily, as shown in FIG. 4 , the order of the first nearest neighbor sequence is class A, class C, class D, and class B, and the sequence number of class B in the first nearest neighbor sequence is 3. Similarly, if the serial number of class A in the second nearest neighbor sequence is 5, then the mutual neighbor distance between class A and class B is 3+5=8.

In this embodiment of the present application, the method for the server to calculate the neighbor sorting distance may be: determining the neighbor sorting distance between the first to-be-processed set and the second to-be-processed set according to the first nearest neighbor sequence and the second nearest neighbor sequence.

In some embodiments, the server may first calculate the sort sum (also called asymmetric rank order distance) of all classes whose distance to class A is closer or equal to that of class B to class A in the second nearest neighbor sequence. The server can calculate by sorting and formula, which is:

Among them, D(A,B) is the asymmetric rank order distance between class A and class B, O _A (B) is the serial number of class B in the first nearest neighbor sequence, O _B (f _A (i)) is the In the two-nearest neighbor sequence, the serial number corresponding to class i.

Specifically, the server can first calculate the absolute distance between each class and class A. If the absolute distance between a class and class A is greater than or equal to the absolute distance between class A and class B, it means that the class is closer to class A. The server can record the sequence number of the class in the second nearest neighbor sequence. After the server traverses all classes, it can add up the recorded serial numbers to obtain the sorted sum. Exemplarily, as shown in FIG. 5 , FIG. 5 is an example diagram of the server calculating the neighbor ranking distance in the embodiment of the present application. In Figure 5, Class A, Class C, and Class D are all ranked before Class B, indicating that Class A, Class C, and Class D are all closer to Class A than Class B. It can also be said that the absolute distance between Class A and Class A is greater than The absolute distance between class A and class B, the absolute distance between class C and class A is greater than the absolute distance between class A and class B, the absolute distance between class D and class A is also greater than the absolute distance between class A and class B distance. The absolute distance between class B and class A is equal to the absolute distance between class A and class B. Therefore, the server records the sequence numbers of class A, class C, class D, and class B in the second nearest neighbor sequence, which are 5, 2, 4, and 0, respectively. Therefore, the server can calculate 5+2+4+0=11. This process can be represented by the following formula:

In the embodiment of this application, the server can also calculate the asymmetric rank order distance between class B and class A as D(B,A). The calculation process is similar to the aforementioned process of calculating D(A,B), and will not repeat.

After the server calculates D(A,B) and D(B,A), they can be added together to obtain the neighbor sorting distance. The formula for calculating the neighbor sorting distance is:

RO(A,B)=D(A,B)+D(B,A);

Among them, RO(A,B) is the neighbor ranking distance between class A and class B.

304. If the absolute distance satisfies the first set condition, the mutual neighbor distance satisfies the second set condition, and the neighbor ranking distance satisfies the third set condition, then generate sample classification results corresponding to the first set to be processed and the second set to be processed .

In this embodiment of the application, the server can pre-set the first setting condition, the second setting condition and the third setting condition, so that when the absolute distance satisfies the first setting condition, the distance between adjacent neighbors satisfies the second setting condition And when the neighbor ranking distance satisfies the third set condition, the server generates sample classification results corresponding to the first set to be processed and the second set to be processed. In some embodiments, the sample classification result may be that the first set to be processed and the second set to be processed are merged into a large target set, and the first sample image and the second set in the first set to be processed and the second set to be processed Both sample images belong to the large target set. In this embodiment of the present application, a target set may also be referred to as a category.

Exemplarily, if the absolute distance between class A and class B is less than or equal to 3, the mutual neighbor distance is greater than t, and the neighbor ranking distance is less than or equal to 30, the server will merge class A and class B into a new class, that is, class A The sample images in the and B categories all belong to this new category. The conditions to be satisfied can be expressed as follows:

MN(C _i ,C _j )<=3||d(C _i ,C _j )＞t||RO(A,B)<=30;

Wherein, t may be set according to actual needs, which is not specifically limited in this embodiment of the present application. In some embodiments, the server sets t so that the accuracy rate of sample classification reaches 97%.

Optionally, on the basis of the above embodiment corresponding to FIG. 3 , in an optional embodiment of the embodiment of the present invention, after generating the sample classification results corresponding to the first set to be processed and the second set to be processed, the method further includes :

Obtaining a target set, where the target set includes a first to-be-processed set and a second to-be-processed set;

determining a single eigenvector according to the first image eigenvector and the second image eigenvector;

Calculate the similarity between target sets based on a single feature vector;

The target set whose similarity is less than the set threshold is sent to the terminal device, so that the terminal device displays the target set.

In the embodiment of the present application, the server may obtain the target set, and the target set is formed by combining the first set to be processed and the second set to be processed. It is understandable that the server can form multiple target sets. For example, the server first obtains categories A, B, C, and D. Assume that categories A and B are combined into one target set, and categories C and D are Combined into one target set, the server gets two target sets. By analogy, the server can obtain several target sets in practical applications.

In some embodiments, after the server acquires the target set, and the target set includes the first sample image and the second sample image, the server may determine a single feature vector according to the first image feature vector and the second image feature vector. In some embodiments, the single feature vector is the average of the first image feature vector and the second image feature vector. Exemplarily, the first image feature vector is [1,2,3], the second image feature vector is [7,8,9], then the single feature vector is [(1+7)/2,(2+8) /2,(3+9)/2]=[4,5,6]. In some embodiments, the average of all image feature vectors in the target set can be used to obtain a single feature vector corresponding to the target set.

Then, the server can calculate the similarity between target sets according to the single feature vector. In some embodiments, the server may calculate the cosine similarity between the target sets according to the single feature vector, and the calculation method is not specifically limited here.

In some embodiments, the server sends the target set whose similarity is less than the set threshold to the terminal device for display. In some other embodiments, the server sorts the similarities, and displays the top 5 groups of targets with the highest similarities. In some embodiments, the way for the server to display may be that the server sends the content to be displayed to a terminal device with a display screen, so that the terminal device displays the content to be displayed on the display screen. The displayed content may be a vector showing the target set, or a sample image in the target set.

In some embodiments, the server selects a sample image in the target set through a pose optimization algorithm and sends it to a terminal device with a display screen for display. The posture optimization algorithm is not specifically limited in this embodiment of the application. FIG. 6 is a diagram showing an example of a target set in the embodiment of the present application. The server may send the selected sample image to a terminal device with a display screen, so that the terminal device displays the sample image on the display screen. It can be seen from Fig. 6 that the interface includes a title bar, a function block and a main interface, wherein the title bar is used to display the title of the program, and the function block is used to display optional functions. In the main interface, multiple sets of sample images are displayed, and each sample image can represent a target set. Due to margin limitations, only three sets of sample images are shown in FIG. 6 . In practical applications, there is no limit to the number of displayed sample images, which is not specifically limited in this embodiment of the present application.

Optionally, on the basis of the various embodiments corresponding to FIG. 3 above, in an optional embodiment of the embodiment of the present invention, after displaying target sets whose similarities are less than the set threshold, the method further includes:

Obtain annotation information, which has an association relationship with the target set;

Determine the secondary sample classification result corresponding to the target set according to the label information.

In the embodiment of the present application, the server may obtain annotation information, and the annotation information has an association relationship with the target set. Then, the server can combine several target sets associated with the label information into a new category to obtain a secondary sample classification result. Exemplarily, as shown in Figure 6, the first group of sample images in the main interface are all grandmothers, and the staff observes that the pedestrians in the two sample images are actually the same person, then the staff can click on the corresponding position on the main interface The "Yes" virtual button. In response to the click operation, the terminal device corresponding to the display screen may generate annotation information, where the annotation information includes an identifier of a target set corresponding to the first group of sample images. Then, the terminal device can send the marking information to the server, and the server can obtain the marking information.

Then, the server may combine the corresponding target sets into a new category according to the target set identifier in the annotation information. Exemplarily, as shown in FIG. 6 , the server can merge the target set corresponding to the grandmother into a new category according to the target set identifier in the tagging information, the category includes the two target sets, and the samples in the category The images are all sample images corresponding to the grandmother.

Fig. 7 is a schematic flowchart of an optional embodiment in the embodiments of the present application. The process can be described as:

701. Obtain pedestrian images;

In this embodiment, the server can acquire images of pedestrians (hereinafter referred to as pedestrian images). In some embodiments, the server may acquire images of pedestrians from camera equipment. In other embodiments, the server can read from the database. For details, reference may be made to the description of the aforementioned step 301 , which will not be repeated here.

702. Extract features;

In the embodiment of the present application, the server may extract features related to the pedestrian from the pedestrian image, or extract features of the entire image from the pedestrian image, which is not specifically limited in the embodiment of the present application. For details, reference may be made to the description of the foregoing step 302 , which will not be repeated here.

703. MN&RO fine-grained clustering;

In the embodiment of the present application, the server may cluster the images of pedestrians according to the extracted features. For details, please refer to the descriptions of the aforementioned steps 303 and 304 , which will not be repeated here. It can be understood that, in the embodiment of the present application, one pedestrian image may be clustered as one class.

704. Multiple features within a class are fused into a single feature;

In the embodiment of the present application, the server clusters a plurality of pedestrian images through the clustering in the above step 703 to obtain N categories, and each category includes n _i pedestrian images. Among them, each pedestrian image has corresponding features, and there are n _i features in each class. Then, the server can fuse the _ni features within each class into a single feature. In some embodiments, the server can perform fusion through a feature fusion algorithm. In some other embodiments, the server may take the average feature of multiple features of each class as the feature representation of the class. The n _i pedestrian images of each class are fused into one feature to form a new feature set F={f ₁ , f ₂ , f ₃ ,...f _N }. Wherein, N is an integer greater than or equal to 1, and n _i is an integer greater than or equal to 1.

705. Optimal posture;

In the embodiment of the present application, each category obtained by the server includes n _i pedestrian images. Therefore, the server can select a pedestrian image from n _i pedestrian images to represent this class. In some embodiments, the server makes the selection via a pose-optimization algorithm. In other embodiments, the server may be manually selected, which is not specifically limited here.

706. Self-retrieval;

In the embodiment of the present application, the server may calculate the similarity between N classes according to the new feature set F={f ₁ , f ₂ , f ₃ , . . . f _N }. Exemplarily, the server may calculate the cosine similarity between f ₁ and f ₂ to obtain the similarity between class 1 and class 2. By analogy, the server can obtain the similarity between all classes, and then the server sorts these similarities, and selects the top 5 groups of classes with the highest similarity. Exemplarily, the server calculates that the cosine similarity between f ₁ and f ₂ ranks first, then the server can select the class corresponding to f ₁ , and f ₂ (that is, class 1 and class 2).

707. Send bids;

In the embodiment of the present application, the server may send the representative image corresponding to the class selected in step 706 to the terminal device with a display screen, so that the terminal device displays the representative image, as shown in FIG. 6 . In some other embodiments, the server may also directly send the class selected in step 706 to the terminal device with a display screen, so that the terminal device selects a representative image in the class through an algorithm similar to step 705, and then performs the representative image exhibit. This embodiment of the present application does not specifically limit it.

708. Obtain a sample.

In this embodiment of the application, after submitting bids in step 707, the staff selects the same pedestrian category in the interface as shown in Figure 6, so that the terminal device generates labeling information. For details, refer to the description of the foregoing embodiments, which will not be repeated here. repeat. Then, the server obtains the label information, and further merges the N classes in the server according to the label information. After the merging is completed, the server can obtain the classified samples.

Fig. 8 is a schematic flow diagram of another optional embodiment in the embodiment of the present application, the flow includes:

801. Obtain pedestrian images;

Step 801 is similar to the aforementioned step 701 and will not be repeated here.

802. Extract features;

Step 802 is similar to the aforementioned step 702 and will not be repeated here.

803, MN&RO fine-grained clustering;

Step 803 is similar to the aforementioned step 703, and will not be repeated here. It should be noted that, in this embodiment of the application, the server obtains N classes through steps such as step 803 .

804. Multiple features within a class are fused into a single feature;

Step 804 is similar to the aforementioned step 704 and will not be repeated here.

805, MN&RO coarse-grained clustering;

In the embodiment of the present application, the server may perform further clustering on the aforementioned N categories. Step 805 is similar to the aforementioned step 803, but the setting conditions therein may be different. Exemplarily, the setting conditions in step 803 are:

MN(C _i ,C _j )<=3||d(C _i ,C _j )>t ₁ ||RO(A,B)<=30;

The setting condition in step 805 can be:

MN(C _i ,C _j )<=3||d(C _i ,C _j )>t ₂ ||RO(A,B)<=30;

Among them, t ₁ and t ₂ can be modified according to actual needs, so that for the P classes obtained in step 805, P is an integer greater than or equal to 0 and less than N. Exemplarily, in step 803, the server clusters to obtain 10,000 sub-categories, and in step 805, the server clusters to obtain 5,000 major categories, and each major category may include at least one sub-category.

806. Only one sub-category is reserved for each major category;

In the embodiment of the present application, the server may select a subcategory from each category to be reserved (the server may delete the subcategory that is not reserved). In some embodiments, the server sends the representative images corresponding to the subcategories to the terminal device with a display screen, allowing the staff to select one of the subcategories as a reservation. In some other embodiments, the server selects the sub-category with the largest number of pedestrian images for retention by counting the number of pedestrian images in each sub-category. In some other embodiments, the server may firstly count the number of pedestrian images in each sub-category, and select the top 5 sub-categories in which the number of pedestrian images is ranked from most to least. Then the server can select a subclass with the largest variance from the five subclasses to keep. The embodiment of the present application does not specifically limit the method for selecting and retaining subcategories.

807. Obtain a sample.

In the embodiment of this application, after the server obtains the P classes according to step 805, it can remove redundant small classes in the P classes to obtain the simplified P classes {c ₁ , c ₂ , c ₃ ,...c _P }. Each class includes several pedestrian images.

The pedestrian images obtained by the server through the embodiments of the present application have corresponding classifications, so when the server performs pedestrian recognition, it can select one of the categories for pedestrian recognition. Exemplarily, the images of pedestrians in category c ₁ are most likely to be images of grandmas, so the server can have a good recognition rate when identifying pedestrian images in category c ₁ .

FIG. 9 is an example diagram of a device for classifying samples in an embodiment of the present application. The device 900 for classifying samples includes:

An acquiring unit 901, configured to acquire a first set to be processed and a second set to be processed, the first set to be processed includes at least one first sample image, and the second set to be processed includes at least one second sample image;

The acquiring unit 901 is further configured to acquire a first image feature vector and a second image feature vector, the first image feature vector has a first corresponding relationship with the first sample image, and the second image feature vector has a second image feature vector with the second sample image Correspondence;

The processing unit 902 is configured to obtain the distance between the first set to be processed and the second set to be processed according to the first image feature vector and the second image feature vector, and the distance includes an absolute distance, a mutual neighbor distance and a neighbor ranking distance;

The processing unit 902 is further configured to generate the first set to be processed and the second set to be processed if the absolute distance satisfies the first set condition, the mutual neighbor distance satisfies the second set condition, and the neighbor ranking distance satisfies the third set condition The corresponding sample classification results.

Optionally, on the basis of the above embodiments corresponding to FIG. 9 , in an optional embodiment of the embodiment of the present invention, the processing unit 902 is further configured to:

determining a cosine distance between the first sample image and the second sample image according to the first image feature vector and the second image feature vector;

The absolute distance between the first set to be processed and the second set to be processed is determined according to the cosine distance, and the absolute distance is the minimum value of the cosine distance.

Optionally, on the basis of the above embodiments corresponding to FIG. 9 , in an optional embodiment of the embodiment of the present invention, the processing unit 902 is further configured to:

Obtain the first absolute distance corresponding to the first set to be processed;

Sorting according to the size of the first absolute distance to obtain the first nearest neighbor sequence corresponding to the first set to be processed;

Obtain a second absolute distance corresponding to the second set to be processed;

Sorting according to the size of the second absolute distance to obtain the second nearest neighbor sequence corresponding to the second set to be processed;

The mutual neighbor distance between the first set to be processed and the second set to be processed is determined according to the sequence number of the first set to be processed in the second nearest neighbor sequence and the sequence number of the second set to be processed in the first nearest neighbor sequence.

Optionally, on the basis of the above embodiments corresponding to FIG. 9 , in an optional embodiment of the embodiment of the present invention, the processing unit 902 is further configured to:

The neighbor ranking distance between the first set to be processed and the second set to be processed is determined according to the first nearest neighbor sequence and the second nearest neighbor sequence.

Optionally, on the basis of the above embodiments corresponding to FIG. 9 , in an optional embodiment of the embodiment of the present invention, the processing unit 902 is further configured to:

Obtaining a target set, where the target set includes a first to-be-processed set and a second to-be-processed set;

determining a single eigenvector according to the first image eigenvector and the second image eigenvector;

Calculate the similarity between target sets based on a single feature vector;

The target set whose similarity is less than the set threshold is sent to the terminal device, so that the terminal device displays the target set.

Optionally, on the basis of the above embodiments corresponding to FIG. 9 , in an optional embodiment of the embodiment of the present invention, the processing unit 902 is further configured to:

Select a sample image in the target collection;

The sample image is sent to the terminal device, so that the terminal device displays the sample image.

Optionally, on the basis of the above embodiments corresponding to FIG. 9 , in an optional embodiment of the embodiment of the present invention, the processing unit 902 is further configured to:

Obtain annotation information, which has an association relationship with the target set;

Determine the secondary sample classification result corresponding to the target set according to the label information.

FIG. 10 is a schematic structural diagram of a server provided by an embodiment of the present application. The server 1000 may have relatively large differences due to different configurations or performances, and may include one or more central processing units (central processing units, CPU) 1022 (for example, one or more processors) and memory 1032, one or more storage media 1030 (such as one or more mass storage devices) for storing application programs 1042 or data 1044. Wherein, the memory 1032 and the storage medium 1030 may be temporary storage or persistent storage. The program stored in the storage medium 1030 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations on the server. Furthermore, the central processing unit 1022 may be configured to communicate with the storage medium 1030 , and execute a series of instruction operations in the storage medium 1030 on the server 1000 .

The server 1000 can also include one or more power supplies 1026, one or more wired or wireless network interfaces 1050, one or more input and output interfaces 1058, and/or, one or more operating systems 1041, such as Windows Server™, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM, etc.

The steps performed by the server in the foregoing embodiments may be based on the server structure shown in FIG. 10 .

In the embodiment of this application, CPU1022 is specifically used to perform the following steps:

Acquiring a first set to be processed and a second set to be processed, the first set to be processed includes at least one first sample image, and the second set to be processed includes at least one second sample image;

Obtaining a first image feature vector and a second image feature vector, the first image feature vector has a first correspondence with the first sample image, and the second image feature vector has a second correspondence with the second sample image;

Obtaining the distance between the first set to be processed and the second set to be processed according to the first image feature vector and the second image feature vector, the distance includes an absolute distance, a mutual neighbor distance and a neighbor sorting distance;

If the absolute distance satisfies the first set condition, the mutual neighbor distance satisfies the second set condition, and the neighbor sorting distance satisfies the third set condition, then the sample classification results corresponding to the first set to be processed and the second set to be processed are generated.

In the embodiment of this application, CPU1022 is also used to perform the following steps:

determining a cosine distance between the first sample image and the second sample image according to the first image feature vector and the second image feature vector;

The absolute distance between the first set to be processed and the second set to be processed is determined according to the cosine distance, and the absolute distance is the minimum value of the cosine distance.

In the embodiment of this application, CPU1022 is also used to perform the following steps:

Determine the first nearest neighbor sequence corresponding to the first set to be processed and the second nearest neighbor sequence corresponding to the second set to be processed according to the absolute distance, the first nearest neighbor sequence is the sorting queue of the second sample image according to the cosine distance, The second nearest neighbor sequence is the sorting queue of the first sample image according to the cosine distance;

The mutual neighbor distance between the first set to be processed and the second set to be processed is determined according to the sequence number of the first absolute sample image in the second nearest neighbor sequence and the sequence number of the first absolute sample image in the second nearest neighbor sequence.

In the embodiment of this application, CPU1022 is also used to perform the following steps:

The neighbor ranking distance between the first set to be processed and the second set to be processed is determined according to the first nearest neighbor sequence and the second nearest neighbor sequence.

In the embodiment of this application, CPU1022 is also used to perform the following steps:

Obtaining a target set, where the target set includes a first to-be-processed set and a second to-be-processed set;

determining a single eigenvector according to the first image eigenvector and the second image eigenvector;

Calculate the similarity between target sets based on a single feature vector;

The target set whose similarity is less than the set threshold is sent to the terminal device, so that the terminal device displays the target set.

In the embodiment of this application, CPU1022 is also used to perform the following steps:

Select a sample image in the target collection;

The sample image is sent to the terminal device, so that the terminal device displays the sample image.

In the embodiment of this application, CPU1022 is also used to perform the following steps:

Obtain annotation information, which has an association relationship with the target set;

Determine the secondary sample classification result corresponding to the target set according to the label information.

Those skilled in the art can clearly understand that for the convenience and brevity of the description, the specific working process of the above-described system, device and unit can refer to the corresponding process in the foregoing method embodiment, which will not be repeated here.

In the several embodiments provided in this application, it should be understood that the disclosed system, device and method can be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or May be integrated into another system, or some features may be ignored, or not implemented. In another point, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces, and the indirect coupling or communication connection of devices or units may be in electrical, mechanical or other forms.

The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, each unit may exist separately physically, or two or more units may be integrated into one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

If the integrated unit is realized in the form of a software function unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium , including several instructions to make a computer device (which may be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk, and other media that can store program codes.

Claims (10)

1. A method of sample classification, comprising:

acquiring a first to-be-processed set and a second to-be-processed set, wherein the first to-be-processed set comprises at least one first sample image, and the second to-be-processed set comprises at least one second sample image;

Acquiring a first image feature vector and a second image feature vector, wherein the first image feature vector has a first corresponding relation with the first sample image, and the second image feature vector has a second corresponding relation with the second sample image;

Acquiring the distance between the first to-be-processed set and the second to-be-processed set according to the first image feature vector and the second image feature vector, wherein the distance comprises an absolute distance, a mutual neighbor distance and a neighbor sorting distance;

and if the absolute distance meets a first set condition, the mutual neighbor distance meets a second set condition and the neighbor sorting distance meets a third set condition, generating sample classification results corresponding to the first to-be-processed set and the second to-be-processed set.

2. The method according to claim 1, wherein said obtaining the distance between the first to-be-processed set and the second to-be-processed set according to the first image feature vector and the second image feature vector comprises:

determining a cosine distance between the first sample image and the second sample image according to the first image feature vector and the second image feature vector;

And determining an absolute distance between the first to-be-processed set and the second to-be-processed set according to the cosine distance, wherein the absolute distance is the minimum value of the cosine distance.

3. the method according to claim 2, wherein the absolute distance is determined from a first absolute sample image and a second absolute sample image, and wherein the obtaining the distance between the first to-be-processed set and the second to-be-processed set from the first image feature vector and the second image feature vector comprises:

Acquiring a first absolute distance corresponding to the first set to be processed;

Sequencing according to the magnitude of the first absolute distance to obtain a first nearest neighbor sequence corresponding to the first to-be-processed set;

acquiring a second absolute distance corresponding to the second to-be-processed set;

Obtaining a second nearest neighbor sequence corresponding to the second to-be-processed set according to the magnitude sorting of the second absolute distance;

And determining the mutual neighbor distance between the first to-be-processed set and the second to-be-processed set according to the sequence numbers of the first to-be-processed set in the second nearest neighbor sequence and the sequence numbers of the second to-be-processed set in the first nearest neighbor sequence.

4. The method according to claim 3, wherein said obtaining the distance between the first to-be-processed set and the second to-be-processed set according to the first image feature vector and the second image feature vector comprises:

determining a neighbor ordering distance between the first to-be-processed set and the second to-be-processed set according to the first nearest neighbor sequence and the second nearest neighbor sequence.

5. The method according to claim 1, wherein after the generating of the sample classification results corresponding to the first to-be-processed set and the second to-be-processed set, the method further comprises:

Acquiring a target set, wherein the target set comprises the first to-be-processed set and the second to-be-processed set;

Determining a single feature vector according to the first image feature vector and the second image feature vector;

calculating the similarity between the target sets according to the single feature vectors;

And sending the target set with the similarity smaller than a set threshold value to a terminal device, so that the terminal device displays the target set.

6. The method of claim 5, wherein the sending the target set with the similarity smaller than a set threshold to a terminal device, so that the terminal device presents the target set comprises:

selecting a sample image in the target set;

And sending the sample image to a terminal device, so that the terminal device displays the sample image.

7. the method of claim 5, wherein after the presenting the target set with the similarity less than a set threshold, the method further comprises:

Obtaining marking information, wherein the marking information has an incidence relation with the target set;

And determining a secondary sample classification result corresponding to the target set according to the labeling information.

8. an apparatus for sample classification, comprising:

An obtaining unit, configured to obtain a first to-be-processed set and a second to-be-processed set, where the first to-be-processed set includes at least one first sample image, and the second to-be-processed set includes at least one second sample image;

the acquiring unit is further used for acquiring a first image feature vector and a second image feature vector, wherein the first image feature vector has a first corresponding relation with the first sample image, and the second image feature vector has a second corresponding relation with the second sample image;

A processing unit, configured to obtain, according to the first image feature vector and the second image feature vector, a distance between the first to-be-processed set and the second to-be-processed set, where the distance includes an absolute distance, a mutual neighbor distance, and a neighbor sorting distance;

the processing unit is further configured to generate sample classification results corresponding to the first to-be-processed set and the second to-be-processed set if the absolute distance satisfies a first set condition, the mutual neighbor distance satisfies a second set condition, and the neighbor sorting distance satisfies a third set condition.

9. a server, comprising:

one or more than one central processing unit, a memory, an input/output interface, a wired or wireless network interface and a power supply;

the memory is a transient memory or a persistent memory;

the central processor is configured to communicate with the memory, the instructions in the memory being executable on the server to perform the method of any one of claims 1 to 7.

10. a computer-readable storage medium having stored therein instructions which, when run on a computer, cause the computer to perform the method of any one of claims 1 to 7.