CN109815814B - Face detection method based on convolutional neural network - Google Patents

Abstract

The invention discloses a face detection method based on a convolutional neural network, which comprises the following steps: step (1), establishing a database; step (2), the images in the database are processed; pre-treating; step (3), training a deeply built learning network; and (4) testing the training result, wherein the detection accuracy of the human faces with shielding, different angles and side faces in the picture and the small and fuzzy human faces in the picture is high, the network structure is simple, the iteration parameters are less, and the training time is short.

Description

A Face Detection Method Based on Convolutional Neural Network

technical field

The invention belongs to the fields of computer vision and artificial intelligence, and in particular relates to a face detection method based on a convolutional neural network.

Background technique

Face detection refers to the process of determining the position and size of a face in an image of a human face. It is an important part in the field of computer vision and a key step in preprocessing for face recognition. Its detection accuracy is very high. To a large extent, it also determines the accuracy of face recognition and has a great impact on the follow-up work. Therefore, the research on face detection has great significance and practical value.

Face detection has a wide range of applications in real life: such as identity authentication and security protection, in media and entertainment about faces, mobile phones, digital cameras and other electronic products, as well as image retrieval and so on. Face detection methods can be roughly divided into traditional detection methods (including detection methods based on matching templates, detection methods based on distance, etc.) and detection methods based on deep learning.

In recent years, deep learning has been continuously improved and developed, and it has been widely used in both classification and regression tasks. The face detection method based on deep learning is also constantly developing, but for the current method, taking the most commonly used MTCNN method as an example, its recognition speed is not fast enough, and the recognition accuracy is not high enough, especially for images and videos. Faces that are occluded, or different angles, sides, and smaller in the frame are less likely to be detected. As a preprocessing step in the face recognition process, the accuracy of face detection also greatly affects the accuracy of subsequent recognition work, so it is very important to solve these problems.

Contents of the invention

Based on the prior art, the present invention proposes a face detection method based on a convolutional neural network, especially related to the detection of faces in the picture due to illumination, occlusion or being in a side state and very small in the picture, By establishing a new database, building a convolutional neural network, and continuously iteratively training the network by adjusting hyperparameters, a better detection effect can be obtained, thereby effectively detecting faces.

The present invention proposes a kind of face detection method based on convolutional neural network, and this method comprises the following steps:

A kind of face detection method based on convolutional neural network, this method comprises the following steps:

Step 1. Establish a database to obtain image data for preprocessing to construct a convolutional neural network;

Step 2, performing four iterative operations on the preprocessed data through the image feature analysis module in the convolutional neural network to generate image feature parameters;

Step 3, generate a one-dimensional image vector by operating the image feature parameters through the fully connected layer in the convolutional neural network;

Step 4. Classify and regress the image one-dimensional vector through the classification layer in the convolutional neural network to obtain the position coordinates of the face image.

Described step 2 image characteristic analysis module comprises the following steps to preprocessing data process:

The convolution layer of the image feature analysis module described in step 2.1 extracts image features by convolving the weights and parameters of the preprocessed data;

Step 2.2, the activation function layer of the image feature analysis module uses the ReLu function to perform nonlinear operations on image features to obtain nonlinear feature map parameters;

Step 2.3, the maximum pooling layer of the image feature analysis module reduces the parameters of the nonlinear feature map.

In the step 4, the classification layer performs the classification and regression process on the image one-dimensional vector: the following steps are included.

Step 4.1, iteratively weights the one-dimensional vector of the image through the optimization method of the stochastic gradient descent method, so as to continuously adjust the loss function, and continuously adjust the hyperparameters during training to obtain the best training results. The hyperparameters include: iteration Number of times, batches, maximum number of iterations, learning rate;

Step 4.2, the loss function selected in the classification process is the combination of the center loss function and the softmax loss function

method, the specific expression is:

Among them, L _S is the softmax loss function, L _c is the center loss function, and λ is a coefficient, which means that the weight of the two is here λ=0.1. In the formula, Wx+b is the output of the fully connected layer, after log, it represents the probability that x _i belongs to category y _i , and C represents the feature center of the category;

Step 4.3, the loss function used in the regression process is: Euclidean distance loss function, the specific expression is as follows:

y _i ∈ R ⁴

是网络预测的输出结果，y为标记的真实标签，即68个人脸关键点的坐标。步骤4.4，将最优权值条件下输出的68个人脸关键点的坐标与数据库中带标签的的人脸关键点坐标及人脸作比对，从而算出此卷积神经网络用于人脸检测的准确率。in,

is the output result of the network prediction, and y is the real label of the mark, that is, the coordinates of the 68 key points of the face. Step 4.4, compare the coordinates of the 68 key points of the face output under the condition of the optimal weight with the coordinates of the key points of the face with labels in the database and the face, so as to calculate the use of the convolutional neural network for face detection the accuracy rate.

Beneficial effect

Compared with the existing technology, the present invention provides a face detection method based on a convolutional neural network, which can detect occluded, different angles, and side faces in the picture as well as smaller and blurred faces in the picture. The detection accuracy of the face is higher, and the network structure is simple, the iteration parameters are less, and the training time is shorter.

Description of drawings

Fig. 1 is a kind of flowchart of face detection method based on convolutional neural network;

Fig. 2 is the connection mode of the used convolutional neural network of a kind of face recognition method based on convolutional neural network proposed by the present invention, which comprises four convolutional layers, four ReLu activation function layers, four maximum pooling layers , two fully connected layers, where the last fully connected layer is a softmax classification layer;

Detailed ways

Below in conjunction with accompanying drawing, the present invention is described in further detail:

As shown in Figure 1, it is a flowchart of a face detection method based on a convolutional neural network.

A kind of face detection method based on convolutional neural network, this method comprises the following steps:

Step 1 (110) establishes a database and obtains image data for preprocessing to construct a convolutional neural network;

In this step, establish a database to obtain image data, that is, include pictures with the following requirements in the established database: the picture contains at least one human face, and the position of the human face is not required, preferably people who are not in the center and far away face; and the background of the face is complex and diverse, including various indoor and outdoor scenes; the position of the face in the image is marked with a rectangular frame, and the face includes eyebrows, eyes, nose, mouth, and facial contours 68 key points included. Image clarity is not required. The established database contains a total of 6,000 images containing human faces and labeled.

In this step, the images in the database are pre-processed, and the images in the established database are firstly subjected to a spatial pyramid pooling operation. This operation can obtain multiple images of different pixels and different scales from one image, which is convenient to obtain from Extract fixed-size feature vectors from multi-scale features; randomly mirror all the pictures generated in the above steps, including up-down, left-right mirroring operations; use 4/5 of the database images processed in the above steps as training databases, 1 /5 is used as a test database;

Step 2, (210) performing four iterative operations on the preprocessed data to generate image feature parameters through the image feature analysis module in the convolutional neural network;

The convolution layer of the image feature analysis module described in step 2.1 extracts image features by convolving the weights and parameters of the preprocessed data;

2.2, the activation function layer of the image feature analysis module uses the ReLu function to perform nonlinear operations on image features to obtain nonlinear feature map parameters;

Step 2.3, the maximum pooling layer of the image feature analysis module reduces the parameters of the nonlinear feature map.

The present invention sends the preprocessed test database image into the trained neural network, the test picture passes through the trained neural network weight matrix, extracts features and passes through a classifier to output the results of classification and regression, and the results of classification In the form of probability, if the probability of judging as a human face is greater than the probability of judging as a non-human face, it is judged as a human face, and the part judged as a human face is marked with a rectangular frame; the result of the regression is the 68 key points of the human face in the picture is marked with a keypoint and returns the marked coordinates.

Step 3, (310) Generate a one-dimensional vector of the image by operating the image feature parameters through the fully connected layer in the convolutional neural network.

Step 4, (410) Classify and regress the image one-dimensional vector through the classification layer in the convolutional neural network to obtain the position coordinates of the face image. The classification layer in the step 4 classifies and returns the image one-dimensional vector: including the following steps:

Step 4.1, iteratively weights the one-dimensional vector of the image through the optimization method of the stochastic gradient descent method, so as to continuously adjust the loss function, and continuously adjust the hyperparameters during training to obtain the best training results. The hyperparameters include: iteration Number of times, batches, maximum number of iterations, learning rate;

In step 4.2, the loss function selected in the classification process is the method of combining the center loss function and the softmax loss function, and the specific expression is:

Among them, L _S is the softmax loss function, L _c is the center loss function, and λ is a coefficient, which means that the weight of the two is here λ=0.1. In the formula, Wx+b is the output of the fully connected layer, after log, it represents the probability that x _i belongs to category y _i , and C represents the feature center of the category;

Step 4.3, the loss function used in the regression process is: Euclidean distance loss function, the specific expression is as follows:

y _i ∈ R ⁴

是网络预测的输出结果，y为标记的真实标签，即68个人脸关键点的坐标。步骤4.4，将最优权值条件下输出的68个人脸关键点的坐标与数据库中带标签的的人脸关键点坐标及人脸作比对，从而算出此卷积神经网络用于人脸检测的准确率。in,

is the output result of the network prediction, and y is the real label of the mark, that is, the coordinates of the 68 key points of the face. Step 4.4, compare the coordinates of the 68 key points of the face output under the condition of the optimal weight with the coordinates of the key points of the face with labels in the database and the face, so as to calculate the use of the convolutional neural network for face detection the accuracy rate.

The whole training task of the present invention is divided into two parts: classification and regression. Classification refers to the problem of face detection as a binary classification problem of faces and non-faces; regression refers to the process of returning the coordinates of the frame of the face and the coordinates of the 68 key points of the face after neural network training , so as to complete the purpose of face detection. After iteratively updating the weights in the network to reduce the loss function, the optimal weights are finally obtained; the recognition results output under the conditions of the optimal weights and the coordinates of the key points of the face with labels in the database and the face For comparison, the accuracy of this convolutional neural network for face detection is calculated.

As shown in Figure 2, the convolutional neural network used in the face recognition method based on the convolutional neural network proposed by the present invention includes four convolutional layers, four ReLu activation function layers, four maximum pooling layers, two fully connected layers, the last fully connected layer is a softmax classification layer. Its functions are respectively: the convolutional layer is used to extract the features of the image by convolving the convolutional layer weights and parameters; the activation function layer is to increase the nonlinear capability of the network, wherein the ReLu function refers to y=max (0, x) this function; the maximum pooling layer is to reduce the output size and reduce the amount of parameters; the fully connected layer is to map the extracted features into a one-dimensional vector; the classification layer is to extract from the aforementioned network In the feature, two parts of the face and non-face are classified, and the coordinates of 68 key points of the face are returned. The whole training process is as follows: First, the parameters in the convolutional layer and the fully connected layer are randomly initialized. After the image in the established database is sent to the network, after four convolutional, activation, and pooling layers, the face is obtained. features, and then through the fully connected layer to obtain a fixed-size feature vector, and finally through the classification layer to obtain the coordinates of the location of the face. The classification layer is to classify the two parts of human face and non-human face in the features extracted by the aforementioned network, and to return the coordinates of 68 key points of the human face. In the process of network training, the data propagates forward along the network, and the error obtained by the loss function propagates backward along the network, so that the parameters in the convolutional layer and the fully connected layer are continuously optimized. Through continuous training and fine-tuning of various parameters, finally Get a good training effect.

The whole training process of the present invention is as follows: first, the parameters in the convolutional layer and the fully connected layer are randomly initialized, and after the images in the established database are sent to the network, after four convolutional, activation, and pooling layers, Get the features of the face, and then get the feature vector of fixed size through the fully connected layer, and finally get the coordinates of the location of the face through the classification layer. In the process of network training, the data propagates forward along the network, and the error obtained by the loss function propagates backward along the network, so that the parameters in the convolutional layer and the fully connected layer are continuously optimized. Through continuous training and fine-tuning various parameters, finally Get a good training effect. This step is to obtain the optimal parameters by training the database. During the whole training process, the loss function is used to represent the error between the actual label and the predicted result, that is, by minimizing the loss function, iterative training is carried out continuously, and the optimal parameters are obtained when the loss function reaches the minimum. Among them, the parameters to be trained include the convolution kernel and bias of the convolutional layer, and the neuron parameters in the fully connected layer. During the entire training process, the data is propagated forward, and the error calculated by the loss function is propagated backwards. Through the gradient descent method, the network can find the global optimal point in the process of continuous iteration, and the optimal parameters are obtained at this time. After the training is over, that is, after the optimal network parameters are obtained, the optimal parameters are substituted into the entire network. At this time, the entire network has the ability of face detection, and face detection can be performed. Afterwards, the accuracy of this neural network for face detection can be obtained through testing.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present invention are included in the protection of the present invention. within range.

Claims (2)

1. a face detection method based on convolutional neural network, is characterized in that, the method comprises the following steps: Step 1. Establish a database to obtain image data for preprocessing to construct a convolutional neural network; Step 2, performing four iterative operations on the preprocessed data through the image feature analysis module in the convolutional neural network to generate image feature parameters; Step 3, generate a one-dimensional image vector by operating the image feature parameters through the fully connected layer in the convolutional neural network; Step 4, classify and return the image one-dimensional vector by the classification layer in the convolutional neural network to obtain the position coordinates of the face image; wherein: the classification layer in the step 4 classifies and returns the image one-dimensional vector, including Follow the steps below: Step 4.1, through the optimization method of stochastic gradient descent method, iteratively weights the one-dimensional vector of the image to continuously adjust the loss function, and continuously adjust the hyperparameters during training to obtain the best training results. The hyperparameters include: iteration Number of times, batches, maximum number of iterations, learning rate; Step 4.2, the loss function selected in the classification process is the method of combining the center loss function and the softmax loss function, and the specific expression is:

Among them, L _S is the softmax loss function, L _c is the center loss function, and λ is the coefficient, indicating that the weight of the two is taken here as λ=0.1, where Wx+b is the output of the fully connected layer, and after log, it means that x _i belongs to the category The probability of y _i , C represents the feature center of the category; Step 4.3, the loss function used in the regression process is: Euclidean distance loss function, the specific expression is as follows:

y _i ∈ R ⁴ in,

is the output result of the network prediction, y is the real label of the mark, that is, the coordinates of the 68 key points of the face; Step 4.4, compare the coordinates of the 68 facial key points output under the optimal weight condition with the labeled human face key point coordinates and faces in the database, so as to calculate this convolutional neural network for face detection the accuracy rate. 2. the face detection method based on convolutional neural network as claimed in claim 1, is characterized in that, in described step 2, image characteristic analysis module comprises the steps to preprocessing data process: Step 2.1, the convolution layer of the image feature analysis module extracts image features by convolving the weights and parameters of the preprocessed data; Step 2.2, the activation function layer of the image feature analysis module uses the ReLu function to perform nonlinear operations on image features to obtain nonlinear feature map parameters; Step 2.3, the maximum pooling layer of the image feature analysis module reduces the parameters of the nonlinear feature map.

Images