Face recognition technology flow analysis algorithm and important technology development

Face Recognition (FR) is a biometric recognition technology based on human facial feature information for identification. A series of related techniques for capturing an image or video stream containing a face with a camera or a camera, and automatically detecting and tracking the face in the image, and then performing face recognition on the detected face, usually called portrait recognition, face recognition .
The research of face recognition system began in the 1960s. After the 1980s, with the development of computer technology and optical imaging technology, the application stage that really entered the primary stage was in the late 90s. In recent years, with deep learning With the advancement of artificial intelligence technology, face recognition technology has developed rapidly. "Face recognition system" integrates various professional technologies such as artificial intelligence, machine recognition, machine learning, model theory, expert system, video image processing, etc. It is a comprehensive and systematic system engineering technology.
Face recognition process
The face recognition system usually includes several processes: face image acquisition and detection, key point extraction, face regularization (image processing), face feature extraction and face recognition comparison.
Face image acquisition. Different face images can be captured by the camera lens, such as still images, moving images, different positions, different expressions, etc., which can be well collected. When the user is within the shooting range of the acquisition device, the acquisition device automatically searches for and captures the user's face image.
Face Detection. In practice, face detection is mainly used for pre-processing of face recognition, that is, the position and size of the face are accurately calibrated in the image.
Key point extraction (feature extraction). The features that can be used by the face recognition system are generally classified into visual features, pixel statistical features, face image transform coefficient features, face image algebra features, and the like. Face feature extraction is performed on certain features of the face. Face feature extraction, also known as face representation, is a process of character modeling a face. The methods of face feature extraction are summarized into two categories: one is based on knowledge representation methods; the other is based on algebraic features or statistical learning.
Face regularization (pre-processing). The image preprocessing for the face is based on the result of the face detection, processing the image and ultimately serving the process of feature extraction. The original image acquired by the system is often not directly used due to various conditions and random interference. It must be pre-processed with grayscale correction and noise filtering in the early stage of image processing. For face images, the preprocessing process mainly includes ray compensation, gradation transformation, histogram equalization, normalization, geometric correction, filtering and sharpening of face images.
Face recognition alignment (matching and recognition). The feature data of the extracted face image is searched and matched with the feature template stored in the database. By setting a threshold, when the similarity exceeds the threshold, the result of the matching is output. Face recognition is to compare the face features to be recognized with the obtained face feature templates, and judge the identity information of the faces according to the degree of similarity. Can be divided into 1:1, 1:N, attribute recognition. 1:1 is to compare the eigenvalue vectors corresponding to two faces, and 1:N is to compare the eigenvalue vector of one face photo with the eigenvalue vector corresponding to the other N faces, and the output is similar. A person with a high X or similarity ranking the top X.
Face feature analysis algorithm
A widely used regional feature analysis algorithm in face recognition technology, which combines computer image processing technology and biostatistics principle, uses computer image processing technology to extract portrait feature points from video, and analyzes by using the principle of biostatistics. Create a mathematical model, a face feature template. Using the established face feature template and the face image of the person being tested, the feature analysis is performed, and a similar value is given according to the analysis result. Use this value to determine if it is the same person.
There are many methods for face recognition. The main face recognition methods are:
(1) Face recognition method of geometric features: The geometric features may be the shapes of the eyes, nose, mouth, etc. and the geometric relationship between them (such as the distance between each other). These algorithms recognize fast, require less memory, but have a lower recognition rate.
(2) Face recognition method based on feature face (PCA): The feature face method is a face recognition method based on KL transform, and KL transform is an orthogonal transform of image compression. The high-dimensional image space is KL-transformed to obtain a new set of orthogonal bases, and the important orthogonal bases are preserved. These bases can be expanded into low-dimensional linear spaces. If the projection of the face in these low-dimensional linear spaces is assumed to be separable, these projections can be used as the identified feature vectors, which is the basic idea of ​​the feature face method. These methods require more training samples and are based entirely on the statistical properties of the image grayscale. There are currently some improved feature face methods.
(3) Face recognition method of neural network: The input of the neural network may be a face image with reduced resolution, an autocorrelation function of a local region, a second moment of local texture, and the like. This type of method also requires more samples for training, and in many applications, the number of samples is very limited.
(4) Ellipse matching method for face recognition: The elasticity map matching method defines a distance in the two-dimensional space that has a certain invariance to the normal face deformation, and uses the attribute topology map to represent the human face. Any vertex of the topology map contains a feature vector that records the information of the face near the vertex position. This method combines the gray-scale characteristics and geometric factors, and allows the image to be elastically deformed when compared. It has a good effect in overcoming the influence of expression changes on recognition, and it does not require multiple samples for a single person. training.
(5) Facial recognition method for line segment Hausdorff distance (LHD): Psychological research has shown that humans are no better at identifying the speed and accuracy of contour maps (such as comics) than identifying grayscale images. LHD is based on the line segment extracted from the face gray image, which defines the distance between the two line segment sets. What is different is that LHD does not establish a one-to-one correspondence between line segments between different line segment sets. Relationship, so it is better able to adapt to small changes between line segments. The experimental results show that LHD has excellent performance under different lighting conditions and different postures, but it has a poor recognition effect in the case of large expressions.
(6) Support vector machine (SVM) face recognition method: Support vector machine is a new hotspot in the field of statistical pattern recognition. It tries to make the learning machine reach a compromise in experience risk and generalization ability, thus improving learning. Machine performance. The SVM mainly solves a 2-class problem. Its basic idea is to try to transform a low-dimensional linear indivisible problem into a high-dimensional linearly separable problem. The usual experimental results show that SVM has a good recognition rate, but it requires a large number of training samples (300 per class), which is often unrealistic in practical applications. Moreover, the support vector machine has a long training time and the method is complicated to implement. There is no unified theory for the function.
Key technology
Deep learning based on big data
Some feature extraction and classification algorithms are mentioned in the section on face recognition principles, which can be understood as a shallow learning model. Shallow learning can exert strong expression ability under a certain scale of data set, but when the amount of data increases, these models will be in an under-fitting state. The popular point is that the amount of data is too large, the model is not complicated enough, and all data cannot be covered. Therefore, deep learning is a particularly hot research topic in recent years.
Deep learning based on big data will be one of the main trends of face recognition technology. Deep learning often involves deeper hierarchies. The lower the level, the simpler the feature, the higher the level, the more abstract the feature, but the closer it is to the intended expression. For example, from words to words, to sentences, to semantics, is a process of deepening layers. This is a typical deep structure. Going back to the scope of image analysis, for a picture, the low-level features are pixels, which are matrices from 0 to 255. Through the pixels, we can't understand what the target is in the picture, but we can find the edge features from the pixels, then combine the edge features into different parts, and finally form different kinds of targets. This is what we want to achieve. of.
The facial features extracted by deep learning can express the correlation between faces more effectively than the traditional techniques, and the effective classification method can significantly improve the recognition rate of the algorithm. Deep learning relies heavily on big data, which is why this technology has made breakthroughs in recent years. More and more sample data is added to the training model, which means that the algorithm model will be more general and closer to the real world model. On the other hand, the theoretical nature of deep learning needs to be strengthened, and the model needs to be optimized. This point is believed that deep learning will be more successful with the efforts of many academic and industrial colleagues. At that time, the face recognition application may be able to penetrate into our lives like the current license plate recognition technology.
3D face recognition technology
3D face recognition technology is an important development discovery of face recognition. Most of the current face recognition applications are limited to 2D images. The face is essentially a three-dimensional model, and 2D face recognition is easily affected by factors such as posture, illumination, and expression, because the 2D image itself has a defect and cannot express the depth information well. If deep learning is to understand face recognition from the perspective of human cognition, then 3D technology is to reflect face recognition from the real model.
At present, the algorithm research on 3D face recognition direction is not as rich and deep as 2D face recognition technology, and many factors limit the development of this technology. First, 3D face recognition often requires a specific acquisition device, such as a 3D camera or a binocular camera. At present, such collection devices are relatively expensive and are mainly used in specific scenarios. Secondly, the 3D modeling process requires a large amount of computation, high hardware requirements, and limits the current application. Third, the 3D face recognition database is relatively rare. The researchers lack training samples and test samples, and cannot conduct more in-depth theoretical research. It is believed that with the development of chip technology and sensors in the future, 3D face recognition will achieve important breakthroughs when computing power is no longer constrained and the cost of 3D acquisition equipment is greatly reduced.

1.5 Ton Electric Straddle Stacker

Zowell XEK15 1.5 ton electric straddle stacker is standard with adjustable fork. It is a ideal forklift for both single and double faced pallets. It is more easy to operate and widely used in the food, medicine, tobacco, textile and car manufacturing industry. It is the best choice for the customer who needs the short length and high lift capacity. Combined with the powerful components and first class spare parts make sure the electric reach stacker have best performance. As an ideal choice to operate in narrow space of warehouse, it can be widely used in petroleum, chemical, pharmaceutical, textile, pigment, coal and other industries, as well as ports, railways, freight yards, warehouses.


Compared with the traditional manual stacking, Zowell full Electric Stacker maintains great advantages in safety, efficiency and cost. In addition to durability and accuracy both outside and inside, Zowell XE electric stackers are also equipped with mast with clear view, double speed and lateral vertical moving switch, which are valuable when working in limited space.

If you need bigger capacity or other customized options such as rod, low mast, grabbling frame, please contact us so that we can provide you best solution with professional knowledge.

Xek15 1


Xe Electric Stakcer

Electric Straddle Stacker, Wide fork leg stacker, Straddle Stacker, Electric Straddle Forklift

Suzhou Pioneer Material Handling Equipment & Technology Co., Ltd , https://www.xianfengforklift.com