<title>Versatile architecture for image recognition applications</title>

With the rapid development of urban rail transit in my country, the safety inspection of subway tunnel infrastructure is in urgent need of automated and intelligent theory and technology research and development. The subway tunnel lining environment is bad, there are a lot of noise and uneven illumination, and the texture features are complex. The traditional manual and image detection and recognition methods have gradually been replaced by more intelligent and effective deep learning algorithms. Therefore, this paper proposes a subway tunnel lining image recognition and detection algorithm that combines image processing and deep convolutional networks, which can quickly and accurately distinguish between cracked images and non-cracked images, and can realize intelligent target search and category labeling in complex images. In this experiment, a total of 4000 original images of subway tunnels were selected, and pixel-level shallow processing operations such as batch preprocessing, hierarchical feature analysis, and extraction of the circumscribed rectangle of the connected area were carried out. For the training sample library of crack images, pixel-level image processing experiments and classification and recognition experiments are carried out for the training sample library to obtain the optimal database processing results, and then the image recognition algorithm for complex cracks in subway tunnels is studied. Experiments show that there are more point noises and block noises in the results of the subway tunnel crack images, and the segmentation results of the crack images have relatively less noise residues. Due to the wall material, some cement materials are a little off, resulting in noise retention. This shows that the algorithm in this paper has a certain application value in the intelligent identification of actual subway tunnel diseases, and provides technical support and research ideas for the application of subsequent deep learning in subway tunnel disease detection.

Remote sensing includes all types of non-contact. surveys that are carried out from various measuring platforms. The tasks in this area are the following: inventory of agricultural land, control of the state of crops, forecasting yields. The aim of the work is to classify 6 types of crop images (wheat, rice, sugarcane, corn, cotton and jute) with greater accuracy. The paper considers an algorithm for primary processing and recognition of images of agricultural crops and algorithms for constructing a neural network for initial processing and recognition of images to solve problems of noise elimination, minimization, smoothing, normalization, segmentation and image recognition.

Traditional image recognition algorithms often face problems such as low recognition accuracy and insufficient robustness when facing complex scenes and multi-class image data. To this end, a hierarchical discriminant analysis (HDA)-based image recognition algorithm was proposed, which effectively improves image recognition performance by constructing a multi-scale feature extraction module, principal component analysis (PCA) dimensionality reduction, attention mechanism, and dynamic hierarchical adjustment strategy combined with a hierarchical feature extraction and discrimination model. The experiment was conducted on three public datasets: CIFAR-10, ImageNet subset (selecting 100 categories with a total of 150000 images, based on covering common object categories and moderate data volume for fair validation of algorithm performance), and MNIST. The performance was compared with models such as VGG16, ResNet50, SVM, KNN, Hierarchical CNN, EfficientNet, GoogLeNet, etc. The results indicated that the proposed method had higher recognition accuracy than other comparative algorithms on different datasets, with accuracies exceeding 90%. The proposed method performed better in terms of mean absolute error and root mean squared error. The F1 value curve of the proposed method was located at the top of the coordinate axis, reaching a maximum value of 92.39%, which was 14.56% higher than the lowest value of 78.24% in the EfficientNet model. This algorithm has better recognition accuracy than traditional algorithms on multiple public datasets, and has strong anti-interference ability and robustness, which can provide reference for optimizing the accuracy of image recognition.

The paper is focused on the analysis of the relation between the stability of the flotation process and the efficiency of Machine Learning (ML) algorithms based on the flotation froth images. An ML process should enable researchers to construct Artificial Intelligence (AI) algorithms for flotation process control. The image of the flotation froth includes information characterizing the flotation process. The information can be extracted with the aid of the Image Recognition (IR) algorithms based on the ML. This enables construction of a flotation process control system in the mineral processing plant, which is based on the recognition of images of the flotation froth. The IR algorithms do not provide stable image recognition results and are not efficient in the situation where the parameters of the flotation process are highly unstable. The classification results were equal to 75.11% and 69.62% for a stable and unstable process, respectively. The experimental data collected at the Polish Pb/Zn mineral processing plant provided better insight to the relationships between the flotation process parameters and ML efficiency. These relationships were analyzed, and guidelines for the construction of the ML process for flotation process control have been formulated.

Image recognition is a feature extraction and pattern matching technique to classify various objectives in computer science. In theory, it mainly relies on the images with distinguishable features as a good starting point. Each image has its own unique characteristics, and image features can be categorized into color features, texture features, and shape features etc. Therefore, using modern recognition techniques, we extract the image features through various algorithms to search for images with high similarities. There are many image recognition algorithms, and some of them are with high recognition rate while others are robust. But not all algorithms can be unconditionally implemented without adjusting to the real situation. In this paper, we introduce the classification algorithm based on Support Vector Machine (SVM) and the feature extraction method based on Principal Components Analysis (PCA). We employ the feature extraction algorithm to characterize facial features and recognize faces by comparing them to those stored in the training data set. Finally, we show the applications of feature extraction and classification algorithms in character recognition. The real characters printed on a cord are preprocessed by PCA, and then classified and identified by SVM with a good recognition rate if properly processed.

Aggregating multiple types of complex data in stock market prediction: A model-independent framework

In portable precision visual measurement, processing result of the measurement image has an important influence on final measurement accuracy in large field working space. However, the images taken in large field working space always have the following characteristics: containing a large amount of information, having a high redundancy and including a lot of background noises. In order to solve these problems, precision image processing and recognition algorithm is researched. Firstly, portable precision visual measurement method is introduced. Then, the design method of global control points including noncoding points and coding points is displayed. Finally, the image processing and recognition algorithm is investigated in detail, including the largest variance threshold method used to partly binary the image, the detection algorithm of global control points based on the constraint of distance and acreage and coding value recognition method through correcting the image feature. A large amount of experiments show that the image processing and recognition algorithm is accurate and effective to processing the image taken in the working space.

Natural history collections play a vital role in biodiversity research and conservation by providing a window to the past. The usefulness of the vast amount of historical data depends on their quality, with correct taxonomic identifications being the most critical. The identification of many of the objects of natural history collections, however, is wanting, doubtful or outdated. Providing correct identifications is difficult given the sheer number of objects and the scarcity of expertise. Here we outline the construction of an ecosystem for the collaborative development and exchange of image recognition algorithms designed to support the identification of objects. Such an ecosystem will facilitate sharing taxonomic expertise among institutions by offering image datasets that are correctly identified by their in-house taxonomic experts. Together with openly accessible machine learning algorithms and easy to use workbenches, this will allow other institutes to train image recognition algorithms and thereby compensate for the lacking expertise.

Feature selection and extraction are often combined for learning effective representation in image recognition. In this paper, we propose a serial combination method for feature representation in face image recognition. The algorithm first selects a subset in high-dimensional low level feature spaces with sparse learning and then feature extraction is further performed with manifold learning. Both feature selection and extraction are performed evolving class-specific information to learn task-specific feature representation. The serial combination serves to form the final lower dimensional representation in a multi-layer structure. Experimental analysis for multiple face image recognition tasks prove the effectiveness of the proposed serial combination of feature learning.

In order to get a good image recognition model, a lot of high-quality data must be trained. However, some special fields (such as battlefield, rare medical images, etc.) have the characteristics of difficult to obtain data and difficult to label. In order to get a small batch image recognition model with good performance under the condition of insufficient data, a Synchronous Deep Reinforcement Learning algorithm for small-batch image recognition is proposed, and the algorithm is used to recognize image data such as combat elements. The algorithm is based on reinforcement learning and convolutional neural network and so on, it is obtained by optimizing it. Through experimental analysis, the algorithm can quickly realize small batch image classification, and the accuracy rate is higher than other similar algorithms, which proves the practicability and feasibility of this algorithm.

The authors suggest a new technique for improvement of recognition reliability. The procedure is based on analysis of an image at several degrees of detail produced in the course of discrete wavelet transformation of the image. The results are presented by efficiency analysis of the algorithm for recognition of images by their distinctive points.

With the rapid development of technology, artificial intelligence is gradually applied in the field of image processing. Traditional image processing and recognition takes human extraction of features as a means to describe and characterize low-level edge information in images, but a large number of features extraction causes certain troubles to image recognition, and it is difficult to extract deep-level information. In recent years, deep learning, as an important emerging technology in the field of image recognition, has broad application prospects, and the research on image recognition technology has important theoretical value and practical significance to promote the development of computer vision and artificial intelligence. It opens up a whole new direction in image processing research, thus meeting the demand for efficient image processing. This paper reviews the applications of deep learning in image recognition, analyzes the classical convolutional neural network model, and summarizes the applications of deep learning in the fields of license plate recognition, face recognition, and medical image recognition.

Image processing and recognition methods are useful in many fields. According to situation, different techniques are used. For many years, methods based on optical Fourier transformation were very popular. Image recognition was performed generally by using optical correlators. Correlation techniques were strongly developed especially for military applications, but in many cases (industrial, biological and biomedical applications) these techniques suffer from a number of limitations. For these reasons, methods based on extraction and statistical processing of image features are more useful. Set of features can be extracted directly from an image (features based on image morphology, image moments etc.) or from image transforms (Fourier, Radon, Hough, Sine, Cosine etc.). The Fourier transformation is one of the most important in image processing. It can be simply performed by using an optical diffractometer. It allows to build image descriptors independent on image translation and after processing independent on image rotation. Diffractometers are very convenient in industrial and medical applications. Digital image processing and recognition were strongly developed on powerful workstations, however these procedures can also be implemented in PCs with DSP microprocessor cards or in situations where digital transforms used for image processing can be simply implemented and do not consume a lot of time. The example of biomedical image recognition performed in an optical way, by using a diffractometer, and in a digital system with a CCD camera will be described here.

An algorithm for image recognition and retrieval of image from image collection is developed. Basis of the algorithm is the progressive wavelet correlation. The recognition consists of three incremental steps, each of them quadruples the number of correlation points. The process can be aborted at any stage if the intermediate results indicate that the correlation will not result in a match. The final result is the recognition and retrieval of the required image, if exists in the image collection. Instructions for the choice of correlation threshold value for obtaining desired results are defined. We perform a series of image search experiments that cover the following scenarios: the given image is present in the database; copies of the given image are present but with different names; similar (but not identical) images are present; and the given image is not present. Experiments are performed with data bases up to 1000 images, using the Oracle database and the Matlab component Database Toolbox for operations with databases.

A series of events generates multiple types of time series data, such as numeric and text data over time, and the variations of the data types capture the events from different angles. This paper aims to integrate the analyses on such numerical and text time-series data influenced by common events with a single model to better understand the events. Specifically, we present a topic model, called a supervised dynamic topic model (sDTM), which finds topics guided by a numerical time series. We applied sDTM to financial indexes and financial news articles. First, sDTM identifies topics associated with the characteristics of time-series data from the multiple types of data. Second, sDTM predicts numerical time-series data with a higher level of accuracy than does the iterative model, which is supported by lower mean squared errors.