preface
In the modern Internet of Things, artificial intelligence is developing faster and faster with the improvement of information technology, and the verification methods of personal information are constantly updated, which also puts forward higher requirements for the security and practicality of personal information.
Therefore, the automatic recognition technology of face expression image is proposed, which effectively analyzes the information collection and recognition accuracy of face recognition.
In order to solve the problems existing in the above methods, an automatic recognition system for facial expression images based on improved genetic algorithms is proposed. Before face expression recognition, the contour area of the face image is obtained for the face features, and the light compensation and Gaussian smoothing methods are used to process the face expression image to improve the image preprocessing effect and enhance the ability of automatic recognition of face expression image.
First, the overall structure of the system
The overall design framework of the face expression image automatic recognition system is shown in Figure 1.
Figure 1: Face recognition system framework
Acquisition image module. This module mainly obtains picture information, real-time acquisition of images through two ways of picture library and camera, and this module is designed to be completed independently in the interface.
Gets the face area. According to the key features and skin color changes of the face, the outline and details of the face can be outlined, and a mathematical model can be constructed through the collected information, and finally a face model with a high degree of similarity to the face can be formed.
Image preprocessing module. The light compensation and Gaussian smoothing method are used to process the clarity of the image to be processed to ensure that the image obtained is sharper than the original image. The main flow of system design is shown in Figure 2.
Figure 2: System design flowchart
2. User login module design
According to the above system design flow, the system modules need to be designed, as shown in Figure 3.
The user should log in to the system when entering the system, so the user's name and password are designed in this module to prevent the intrusion of malware, and the number of password entries is limited, and the example diagram of this module is shown in Figure 4.
Figure 3 System functional structure diagram
Figure 4: Example diagram of the user login module
Image acquisition module design: Image acquisition is one of the main steps in automatic recognition of facial expression images, and the way to obtain images is camera or database.
When the user queries the image, he only needs to click the image acquisition button in this module to automatically find it in the computer file, and the design can be widely used in the system.
Image preprocessing module design images need to be processed if they are stored in the system database. Select the image to be trained, and the selected image will become the code to realize the training of a single image.
This module is mainly divided into four parts to complete the recognition of face expression images, and the specific face recognition training function is shown in Figure 5.
Figure 5: Image preprocessing module
One-person operation design interface design. This function first collects a single image, determines whether it has been trained after acquisition, and determines whether it is operated using the image training core code later.
Multi-person operation design interface design. This function is similar to the operation process, except that multiple people are captured. Image training core part. This part mainly realizes the training of faces by writing system code.
The face recognition module is designed according to the user's perspective, and the module design is shown in Figure 6.
Figure 6: Face recognition module
This module first deletes or adds to image resources. Remove worthless images from the database, saving system space. And import the newly added images directly into the image library. The face library trains images, obtains model data from them, and provides important information through face information recognition.
Recognize facial expression images and obtain face information from them. The database will save the user login information, login time and the image to be recognized in real time, and the SQL2008 database is selected for design considering the system security.
Database entity diagram design. The main entity diagram of the design system database is shown in Figure 7.
Figure 7: User entity diagram
Figure 7 contains attributes such as user name, login password, permissions, login IP address, and login time.
Face image solid diagram design. As can be seen in Figure 8, the design mainly includes attributes such as image number, image person's name and address, and ID card.
Figure 8: Face image entity diagram
Eye, mouth, nose information entity diagram design. In the automatic recognition of face expression images, eyes, mouth, and nose are important information for recognition, so it is necessary to design the entity diagram of these information, as shown in Figures 9, 10, and 11.
Through the accurate design of the eyes, mouth and nose of the face, the accuracy of automatic recognition of facial expression images is improved.
Figure 9: Eye information entity diagram
Figure 10: Solid diagram of mouth information
Figure 11: Nose information entity diagram
The overall framework of the server side is designed by the framework design of the system and its functional modules to realize the overall design of the face expression image automatic recognition system, so the face expression image automatic recognition system software should be designed.
In the overall face expression image automatic recognition system, the server side of the system is the key to the system, including monitoring and acquisition, processing image files, face detection and recognition, etc., as shown in Figure 12.
Figure 12: System server-side framework
Suppose a high-definition camera is set up directly in front of the channel to capture faces in the passing crowd, so as to achieve the purpose of automatic recognition of face expression images. Set the left and right deflection less than 30°, the upper and lower deflection less than 15°, the height h=3m, and the camera pitch angle of α=15° to capture the image of the face.
After acquisition, the image is denoised and normalized, and the AdaBoost method is used to detect whether there is a face in the image.
If there is a face, then obtain the positioning of both eyes from it, and use the improved genetic algorithm to recognize the face, and its main process is shown in Figure 13.
Figure 13: Improved genetic algorithm flowchart
The improved genetic algorithm is an adaptive search algorithm with strong optimization ability, and it can be seen from Figure 13 that the steps of the face recognition method are. First, the original image is divided, mainly divided into two types: sample training set and sample test set, and then the features in the training image are extracted and normalized to obtain data. The population size and chromosome length are parameterized, and initialization is performed, and the weight and threshold are encoded by the BP network.
The fitness values of the individual are calculated while the optimal individual is sent to the next generation. Update the population. The individual is decoded, mainly divided into weights and thresholds, and the fitness values are obtained from which the threshold is calculated.
Judge the adaptation value, observe whether the accuracy value is met, if not, return to step 3 to retest. The training of BP neural network is carried out through the sample training set step to achieve the predetermined index, and it is sent to the BP neural network after training, so as to achieve the accuracy of face recognition and finally realize the recognition rule.
Through the design of the system function module, the collection of face expression image is realized, and the system software is designed on the basis of the system framework, and the improved genetic algorithm is used to identify the collected face image, from which the optimal solution for recognition is obtained, and the design method of face expression image automatic recognition system based on the improved genetic algorithm is realized.
In order to verify the overall effectiveness of the design method of face expression image automatic recognition system based on improved genetic algorithm, experimental tests are carried out on the proposed method. This experiment was conducted on the Windows 1064-bit operating system.
The CPU is InterCorei 7,8GB of RAM. A frontal facial expression image of a female actor was randomly selected in the portrait database, and the automatic facial expression image recognition system was effectively tested by using the design method of automatic facial expression image recognition system based on improved genetic algorithm (Method 1), the proposed face recognition method based on convolutional neural network (Method 2), and the proposed face recognition method of improved genetic algorithm (Method 3).
The forehead, eyes, nose, mouth and ears in the original image are identified.
Figure 14: Comparative test of face recognition effect of different methods
As shown in Figure 14, when method 1 is used to recognize the original image, it can be seen that the proposed method can accurately identify all the parts that need to be identified, and it can be seen that method 1 has a very high recognition efficiency.
When method 2 recognizes the original image, it is found that method 2 only recognizes 3 parts when it is necessary to identify 6 parts, and the recognition efficiency of method 2 is low and the recognition effect is poor compared with method 1.
When method 3 recognizes the original image, although four images are recognized, it can be observed that method 2 only correctly identifies this part of the forehead, and the rest are all wrong, which indicates that the image recognition effect of method 3 is worse than that of method 1 and method 2.
Through these analyses, it can be seen that the face expression image recognition effect of Method 1 has the highest effect and great accuracy.
In order to verify the superior performance of the proposed method in face recognition response time, 10 frontal facial expression images of faces are randomly selected in the portrait database, and the response time test of system face recognition is carried out by method 1, method 2 and method 3 to verify the recognition response time of different methods. The experimental results are shown in Figure 15.
Figure 15: System face recognition response time test
From the above figure, it can be seen that methods 2 and 3, as the number of users increases, the response time of the system according to methods 2 and 3 also grows, and it is getting higher and higher.
The response time of method 2 is 2.3s~6.5s, and the response time of method 3 is 3s~7s, the response time of these two methods is too long, so it can be judged that methods 2 and 3 are defective in terms of system response. In method 1, even if the user continues to increase, it does not affect the speed of system response time, in the whole test, only method 1 maintains the system response time of 1.5s, it can be seen that the system response speed of method 1 is higher and more stable, and the system work efficiency is better.
In summary, the system response time of method 1 is faster than that of method 2 and method 3, because method 1 uses the light compensation method and Gaussian smoothing method to preprocess the face expression image, so that the processed image is clearer than the original image, which is convenient for the system to recognize the face expression image, enhances the recognition ability of the system, and then improves the response speed of the face recognition image of the system.
In order to verify the success rate of face recognition in this method, the number of image recognition is experimented. Set the system recognition time to 20s, and select 10 face expression images for system recognition within a certain period of time.
Method 1, Method 2 and Method 3 were used to test the number of system face recognitions to check the number of recognitions of different methods, and the more recognitions, the higher the success rate of system recognition. The experimental results are shown in Figure 16.
According to the data in Figure 16, as the recognition time increases, the number of recognized images increases as the recognition time increases.
In general, the number of image recognition in Method 1 is more than that of Method 2 and Method 3, which indicates that Method 1 can recognize face images in real time, and the success rate of recognition is higher than that of Methods 2 and 3.
Figure 16: Number of system image recognitions
summary
Automatic recognition technology of facial expression images is an important place in the field of information.
However, due to the importance of personal information, people have high requirements for the accuracy of face recognition, and it is necessary to design the automatic recognition system of face expression images.
After investigation, it can be seen that the traditional automatic recognition system of face expression image has the problems of poor face recognition effect, slow response time of system face recognition and small number of system image recognition.
In view of the above problems, a design method of face expression image automatic recognition system based on improved genetic algorithm is proposed, and the overall framework of the system is first designed, from which the design of the functional module is completed, and then the system software is designed to realize the recognition and acquisition of face, so as to realize the system method.
The system improves the practicability and accuracy of automatic recognition of facial expression images and solves the problems existing in the current method, but the system still has some defects, and the system will continue to be upgraded and studied in the future