School sports meeting photo automatic matching method based on participating license plate recognition and face recognition

Abstract

A school sports meeting photo automatic matching method based on competition number plate recognition and face recognition is characterized by comprising the following steps: 1) preprocessing a photo set of a school sports meeting; 2) for each photo in the photo set, executing the following operations: 2.1) classifying the photos in the photo set according to the number of the photos by adopting face recognition; 2.2) if the photo is an unmanned photo, the photo does not need to be processed; 2.3) if the photo is a single-person photo, carrying out face recognition and competition number plate recognition, and if the photo is the same person, putting the photo into a folder Pi; 2.4) if the photo is a multi-person photo, performing face recognition to give a candidate set; performing competition number plate identification to find corresponding competition participant sets, and if the two sets are the same, putting the photos into a plurality of corresponding folders; and 2.5) if not, waiting for manual determination.

Description

Technical Field

[0001] The present invention relates to the technical fields of image processing and face recognition, and in particular to an automatic matching method for school sports meeting photos that combines face recognition and participant number plate recognition for a large number of photos taken at school sports meetings. Background Art

[0002] Schools hold sports meetings every year and also take a large number of photos for publicity and distribute them to the corresponding students. Previously, manual methods were usually used for photo classification. However, due to the large number of people involved, the photo classification efficiency was low and could not meet the efficiency requirements.

[0003] When using face recognition technology alone, the recognition rate is not high due to reasons such as photo lighting, shooting angle, face occlusion, and facial expressions. Considering the uniqueness of the participant number plate and the relatively fixed numbers on it, the recognition rate is relatively high. Combining the two recognition technologies of face recognition and participant number plate can effectively improve the automatic matching efficiency and accuracy of photos. Summary of the Invention

[0004] In order to solve the deficiencies of slow, inefficient, and inaccurate manual matching of existing school sports meeting photos, the present invention provides an automatic matching method for school sports meeting photos based on participant number plate recognition and face recognition, which has high recognition accuracy and strong robustness.

[0005] The technical solution adopted by the present invention to solve its technical problems is:

[0006] An automatic matching method for school sports meeting photos based on participant number plate recognition and face recognition, characterized in that the method includes the following steps:

[0007] Step A: Preprocess the photo set of the school sports meeting;

[0008] Step B: For each photo in the photo set, perform the following operations:

[0009] 1) Classify the photos in the photo set according to the number of people in the photo using face recognition, into: photos without people, single-person photos, multi-person photos, and undetermined;

[0010] 2) If it is a photo without people, no processing is required;

[0011] 3) If it is a single-person photo, perform face recognition to give candidates and confidence levels (Pi, ci); use the participant number plate recognition method to give candidate participant number plates and confidence levels (Nj, tj); query the corresponding table of participants and participant number plates according to the candidate participant number plate Nj to find the corresponding participant Pk. If Pk and Pi are the same person, put the photo into the folder Pi; otherwise, put it into the folder Punkown to be determined manually;

[0012] 4) If it is a group photo, perform face recognition to give a set of candidates and confidence levels {(Pi, ci),...}; perform identification of the competition number plates to give a set of candidate competition number plates and confidence levels {(Nj, tj),...}; query the corresponding table of participants and competition number plates according to the set of candidate competition number plates {Nj,...}, and find the corresponding participants {Pk,...}. If the sets {Pk,...} and {Pi,...} are the same, put the photo into the folder {Pi,...} at the same time, otherwise put it into the folder Punkown to be determined manually;

[0013] 5) If it cannot be determined, put the photo into the folder Punkown to be determined manually;

[0014] Furthermore, the method for identifying the competition number plates in step B includes:

[0015] Step C: Obtain the area of the competition number plate in the image;

[0016] Step D: According to the naming rules of the competition number plates for school sports meetings, preset area parameters in the area of the competition number plate to determine the recognition areas of each character in the area of the competition number plate;

[0017] Step E: Perform character recognition on the recognition areas of each character to obtain the character recognition results; the character recognition results include: each recognized character and the confidence level corresponding to each character;

[0018] Step F: Calculate the sum of the matching degrees corresponding to each character in the area of the competition number plate;

[0019] Step G: From all the calculated sums of the matching degrees corresponding to each character in the area of the number plate, obtain the recognition result with the highest sum of the matching degrees corresponding to each character in the area of the number plate as the preferred recognition result;

[0020] Step H: Output the preferred recognition result as the number plate recognition result;

[0021] Still further, the face recognition in step B includes:

[0022] Step I: Collection of face data of students participating in the sports meeting: When students sign up for the sports meeting, in addition to entering the regular grade, class, and name, they are also required to upload a recent photo and have their faces collected;

[0023] Step J: Feature extraction: After the face image data is transmitted to the server, the server uses deep learning algorithms to perform feature extraction and training and modeling on the face data;

[0024] Step K: Perform grayscale conversion, draw a face rectangle, and compare face features based on the given photo, and give the face recognition result; the face recognition result includes: each recognized face and the confidence level corresponding to each face. Description of the Drawings

[0025] Figure 1 is the processing flow chart in the precise matching mode of the present invention.

[0026] Figure 2 is the flow chart of the method for recognizing the competition number plate.

[0027] Figure 3 is the face recognition flow chart.

[0028] Figure 4 is the processing flow chart in the fast matching mode of the present invention. Detailed Embodiments

[0029] The present invention will be further described below with reference to the drawings. It should be understood that the examples described herein are only for explaining and illustrating the present invention and are not used to limit the present invention.

[0030] Refer to Figure 1 , an automatic matching method for school sports meeting photos based on competition number plate recognition and face recognition includes the following steps:

[0031] Step A: Preprocess the photo set of the school sports meeting;

[0032] Step B: For each photo in the photo set, perform the following operations:

[0033] 1) Classify the photos in the photo set according to the number of people in the photo by using face recognition, and divide them into: photos without people, single-person photos, multi-person photos, and undetermined;

[0034] 2) If it is a photo without people, no processing is required;

[0035] 3) If it is a single-person photo, perform face recognition to give candidates and confidence levels (Pi, ci); use the method for recognizing the competition number plate to give candidate competition number plates and confidence levels (Nj, tj); query the corresponding table of participants and competition number plates according to the candidate competition number plate Nj to find the corresponding participant Pk. If Pk and Pi are the same person, put the photo into the folder Pi; otherwise, put it into the folder Punkown to be determined manually;

[0036] 4) If it is a group photo, perform face recognition to give a set of candidates and confidence levels {(Pi, ci),...}; perform recognition of the competition number plates to give a set of candidate competition number plates and confidence levels {(Nj, tj),...}; query the correspondence table between the participants and the competition number plates according to the set of candidate competition number plates {Nj,...} to find the corresponding participants {Pk,...}. If the sets {Pk,...} and {Pi,...} are the same, put the photo into the folder {Pi,...} at the same time; otherwise, put it into the folder Punkown to be determined manually.

[0037] 5) If it cannot be determined, put the photo into the folder Punkown to be determined manually.

[0038] Refer to Figure 2 , for the method of recognizing the competition number plates, it includes:

[0039] Step C: Obtain the area of the competition number plate in the image;

[0040] Step D: According to the naming rules of the competition number plates for school sports meets, preset area parameters in the area of the competition number plate to determine the recognition areas of each character in the area of the competition number plate;

[0041] Step E: Perform character recognition on the recognition areas of each character to obtain the character recognition results; the character recognition results include: each recognized character and the confidence level corresponding to each character;

[0042] Step F: Calculate the sum of the matching degrees corresponding to each character in the area of the competition number plate;

[0043] Step G: From all the calculated sums of the matching degrees corresponding to each character in the area of the number plate, obtain the recognition result with the highest sum of the matching degrees corresponding to each character in the area of the number plate as the preferred recognition result;

[0044] Step H: Output the preferred recognition result as the result of number plate recognition;

[0045] Refer to Figure 3 , the face recognition specifically includes:

[0046] Step I: Collection of face data of students participating in the sports meet: When students sign up for the sports meet, in addition to entering the regular grade, class, and name, they are also required to upload a recent photo and have their faces collected;

[0047] Step J: Feature extraction: After the face image data is transmitted to the server, the server uses deep learning algorithms to perform feature extraction and training and modeling on the face data;

[0048] Step K: Perform grayscale conversion, draw a face rectangle, and perform face feature comparison based on the given photo, and give the face recognition result; the face recognition result includes: each recognized face and the confidence level corresponding to each face.

[0049] The above embodiments are the processing flow steps in the precise matching mode.

[0050] Refer to Figure 4 , in the fast matching mode, the processing flow steps of the photo fast matching mode of the school sports meeting photo automatic matching system based on face recognition and participant number plate recognition are as follows:

[0051] Step A: Preprocess the photo set of the school sports meeting;

[0052] Step B: For each photo in the photo set, perform the following operations:

[0053] 1) Classify the photos in the photo set according to the number of people in the photo using face recognition, and divide them into: photos without people, photos with people, and undetermined;

[0054] 2) If it is a photo without people, no processing is required;

[0055] 3) If it is a photo with people:

[0056] a. Perform face recognition on the photo to give all candidates and the confidence level set {(Pi, ci),...}, and put the photo into the corresponding folder {Pi,...};

[0057] b. Perform participant number plate recognition on the photo to give the candidate participant number plates and the confidence level set {(Nj, tj),...}, query the corresponding table of participants and participant number plates according to the candidate participant number plates {Nj,...}, find the corresponding participants {Pk,...}, and put the photo into the folder {Pk,...};

[0058] 4) If it is undetermined, put the photo into the folder Punkown to be determined manually.

Claims

1. An automatic matching method for school sports meeting photos based on participant number plate recognition and face recognition, characterized in that The method includes the following steps: Step A: Preprocess the photo collection of the school sports meeting; Step B: For each photo in the photo collection, perform the following operations: 1) Classify the photos in the photo collection according to the number of people in the photo using face recognition, into: photos without people, single-person photos, multi-person photos, and undetermined; 2) If it is a photo without people, no processing is required; 3) If it is a single-person photo, perform face recognition to give candidates and confidence levels (Pi, ci); use the method of identifying competition number plates to give candidate competition number plates and confidence levels (Nj, tj); query the corresponding table of participants and competition number plates according to the candidate competition number plate Nj to find the corresponding participant Pk. If Pk and Pi are the same person, put the photo into the folder Pi; Otherwise, put it into the folder Punkown to be determined manually; 4) If it is a multi-person photo, perform face recognition to give a set of candidates and confidence levels {(Pi, ci),...}; perform competition number plate recognition to give a set of candidate competition number plates and confidence levels {(Nj, tj),...}; query the corresponding table of participants and competition number plates according to the set of candidate competition number plates {Nj,...} to find the corresponding participants {Pk,...}. If the set {Pk,...} is the same as the set {Pi,...}, put the photo into the folders {Pi,...} at the same time, otherwise put it into the folder Punkown to be determined manually; 5) If it cannot be determined, put the photo into the folder Punkown to be determined manually.

2. The automatic matching method for school sports meeting photos based on participant number plate recognition and face recognition according to claim 1, wherein: The method of identifying competition number plates in Step B includes: Step C: Obtain the area of the competition number plate in the image; Step D: According to the naming rules of the competition number plates for the school sports meeting, preset area parameters in the area of the competition number plate to determine the recognition areas of each character in the area of the competition number plate; Step E: Perform character recognition on the recognition areas of each character to obtain the character recognition result; the character recognition result includes: each recognized character and the confidence level corresponding to each character; Step F: Calculate the sum of the matching degrees corresponding to each character in the area of the competition number plate; Step G: From all the calculated sums of the matching degrees corresponding to each character in the area of the number plate, obtain the recognition result with the highest sum of the matching degrees corresponding to each character in the area of the number plate as the preferred recognition result; Step H: Output the preferred recognition result as the competition number plate recognition result.

3. The automatic matching method for school sports meeting photos based on competition number plate recognition and face recognition according to claim 1, wherein: The face recognition in Step B includes: Step I: Collect face data of the participating students in the sports meeting: When the students sign up for the sports meeting, in addition to entering the regular grade, class, and name, they are also required to upload a recent photo and have their faces collected; Step J: Feature extraction: After the face image data is transmitted to the server, the server uses deep learning algorithms to perform feature extraction and training modeling on the face data; Step K: According to the given photo, perform grayscale conversion, draw a face rectangle frame, and compare face features to give the face recognition result; the face recognition result includes: each recognized face and the confidence level corresponding to each face.

Images