International Journal of Pattern Recognition and Artificial Intelligence

Infrared small target detection is one of the key techniques in infrared imaging guidance system. The technology of infrared small target detection still needs to be further studied to improve the detection performance. This paper combines the high-pass filtering characteristics of morphological top-hat transform with SUSAN algorithm, and proposes a small infrared target detection method based on morphology and SUSAN algorithm. This method uses top-hat transform to detect the high-frequency region in infrared image, and filters out the low-frequency region in the image to implement the preliminary background suppression of infrared image. Then the SUSAN algorithm is used to detect small targets in the image after background suppression. The proposed method is applied to the single infrared image which is acquired by the infrared guidance system in the process of detecting and tracking the target under specific conditions. The experimental results show that the method is effective and can detect infrared small targets under different background.

With the accumulation of people’s wealth and the improvement of purchasing power, more and more people are buying cars as a means of travel. Walking and cycling of the past have now become a car trip. License plate recognition technology is especially important in intelligent transportation systems. It has been widely used in large shopping malls or supermarket parking lots, highway toll stations, speeding violation supervision and other fields. However, the accuracy and efficiency of license plate image recognition are insufficient. To solve the above problems, we propose a license plate character recognition method based on local HOG and layered LBP feature fusion from the perspectives of image pre-processing, license location, characters’ segmentation and recognition. First, pre-processing the license image area by highlighting the license plate image; then, the license plate is positioned based on wavelet decomposition and brightness moment; next the tilted license plate image is corrected, the license plate frame is adjusted, and characterization is performed by using the improved projection method based on the fact that the projection of the character is a single peak or a double peak. Finally, the local HOG and hierarchical LBP feature fusion methods are used to identify the license characters. The results show that the license plate’s character recognition rate of the proposed method reaches 99.71%, and the time taken is small. This not only improves the character recognition rate, but also saves recognition time. The results show that the method has important practical significance in license plates’ recognition.

With the growth of massive data in the current mobile Internet, network recruitment is gradually growing into a new recruitment channel. How to effectively mine available information in the massive network recruitment data has become the technical bottleneck of current education and social supply and demand development. The renewal of talent demand information is carried out every day, which produces a large amount of text data. How to manage these talents’ demand information reasonably becomes more and more important. Artificial classification is time-consuming and laborious, which is unrealistic naturally. Therefore, using automatic text categorization technology to classify and manage this information becomes particularly important. To break through the bottleneck of this technology, a heuristic KNN text categorization algorithm based on ABC (artificial bee colony) is proposed to adjust the weight of features, and the similarity between test observation and training observation is measured by using the method of fuzzy distance measurement. Firstly, the recruitment information is segmented and feature selection and noise data elimination are carried out by using term frequency-inverse document frequency (TF-IDF) algorithm and AP (affinity propagation) clustering algorithm. Finally, the text information is classified by using KNN algorithm combined with heuristic search and fuzzy distance measurement. The experimental results show that this method effectively solves the problem of poor stability and low classification accuracy of traditional KNN algorithm in text categorization method for talent demand.

With the improvement of our country’s economic level and quality of life, the numbers and scales of highway networks and motor vehicles are constantly expanding, which makes the current road traffic burden more and more serious. As an important means of traffic automation management, license plate recognition (LPR) technology plays an important role in traffic surveillance and control. However, the recognition rate and accuracy of the traditional license plate recognition methods still need to be improved. In the case of poor surrounding environment, it is prone to localization failure, vehicle license plate recognition errors or unrecognizable phenomena. Wavelet transform, as another landmark signal processing method after Fourier transform, has been widely used in the field of image processing. In China, the number of horizontal lines is usually larger than that of vertical lines. If the two vertical boundaries of the license plate can be detected successfully, the four angles of the license plate can be determined efficiently to complete the license plate positioning. In view of the advantages of wavelet transform technology and the characteristics of vehicle license plate, in this paper, a vehicle license plate recognition algorithm based on wavelet transform and vertical edge matching is proposed. The edge of the license plate is detected by wavelet transform technology, and then the license plate is located by vertical edge matching technology. After the location is realized, the characters are segmented by vertical projection method and the characters are recognized by improved BP neural network algorithm. The experimental results show that the proposed vehicle license plate recognition algorithm based on wavelet transform and vertical edge matching performs well in algorithm performance, which provides a good reference for the development of vehicle license plate recognition system.

In the past, most of the digital camouflage used textural features to extract the configuration features of spots in gray images, unable to effectively utilize the position relationship between color information. In order to overcome this shortcoming, a new digital camouflage pattern design model was proposed based on the model of adversarial autoencoder network. Firstly, the complexity and performance of several main color extraction algorithms were analyzed and compared, and combined with AFK-MC² algorithm and color similarity coefficient, a fast camouflage main color clustering method was proposed. Then a deep convolution adversarial autoencoder network was designed to extract and describe the configuration features of the spots in background pattern. In order to diffuse pixel spot and achieve the effect of spatial color blending, a morphological processing algorithm was proposed to process the generated camouflage patterns. Finally, two sets of grassland and woodland datasets were established, respectively. The influence of the number of latent variables of network on the training process was tested on the dataset, and the number of camouflage feature descriptions was determined to be greater than or equal to 10. In order to verify the effectiveness of the generated camouflage, the spots in background region and target region were randomly selected, and the Euclidean distance between the feature parameters of these spots was calculated. Both the visual and experimental results demonstrate that the generated spots have high fusion with the background.

This paper presents a novel Echo State Network (ESN) model for chaotic time series prediction, which consists of three steps including input reconstruction, dimensionality reduction and regression. First, phase-space reconstruction is used to reconstruct the original ‘attractor’ of the input time series. Then, Independent Component Analysis (ICA) is used to identify independent components, reduce dimensionality and overcome multicollinearity problem of the reconstructed input matrix. Finally, Bayesian Ridge Regression provides accurate predictions thanks to its regularization effect to avoid over-fitting and its robustness to noise owing to its probabilistic strategy. Our experimental results show that our model significantly outperforms other ESN models in predicting both artificial and real-world chaotic time series.

At present, image recognition processing technology has been playing a decisive role in the field of pattern recognition, of which automatic recognition of bank notes is an important research topic. Due to the limitation of the size of bill layout and printing method, many invoice layouts are not clear, skewed or distorted, and even there are irregular handwritten signature contents, which lead to the problem of recognition of digital characters on bill surface. In this regard, this paper proposes a data acquisition and recognition algorithm based on improved BP neural network for ticket number identification, which is based on the theory of image processing and recognition, combined with improved bill information recognition technology. First, in the pre-processing stage of bill image, denoising and graying of bill image are processed. After binarization of bill image, the tilt detection method based on Bresenham integer algorithm is used to correct the tilted bill image. Secondly, character localization and feature extraction are carried out for par characters, and the target background is separated from the interference background in order to extract the desired target characters. Finally, the improved BP neural network-based bill digit data acquisition and recognition algorithm is used to realize the classification and recognition of bill characters. The experimental results show that the improved method has better classification and recognition effect than other data acquisition and recognition algorithms.

As an important branch of artificial intelligence, computer vision plays a huge role in the rapid development of artificial intelligence. From a biological point of view, in the acquisition and processing of information, vision is much more important than hearing, touch, etc., because 70% of the human cerebral cortex is processing visual information. Therefore, advances in computer vision technology are critical to the development of artificial intelligence that is designed to allow machines to think and handle things like humans. The acquisition and processing of visual information has always been the focus of computer vision research, and it is also difficult. The main problem of traditional computer vision technology in the processing of visual information is that the extracted image features are less discriminative, the generalization ability of image features in complex background scenes is insufficient, and the recognition ability on object recognition is poor. In response to these problems, based on the visual neural mechanism, this paper establishes an appropriate computer model for the neuronal cells in the human primary visual cortex, models the recognition response mechanism of the visual ventral system, and performs image feature extraction on the training samples. And object recognition. The results show that compared with the traditional methods, the proposed method effectively improves the discrimination of image features, and the image features extracted under complex background scenes have good generalization ability. On this basis, the training samples can be effectively recognized. The results show that the model based on the visual neural mechanism, the recognition of the edge, orientation and contour of the training sample show the advantages of the biological vision mechanism in object recognition.

The detection of internal damage characteristics of concrete is an important aspect of damage evolution mechanism in concrete meso-structure. In this paper, the improved Faster R-CNN is used to detect the porosity and cracks in concrete CT images. Based on the Faster R-CNN, ResNet-101 and ResNet-50 are used as the main framework. Feature pyramid network (FPN) and ROI Align are introduced to improve the performance of the model. FPN can generate high quality feature maps. ROI Align solves the region mismatch caused by the quantization operation. Experiments show that the detection accuracy of ResNet-101+FPN+ROI Align reaches 87.08%, which is 4.74 higher than that of ResNet-101. The detection accuracy of ResNet-50 + FPN + ROI Align reached 81.36%, which is 3.12% points higher than ResNet-50. These two improved algorithms are slower than the original algorithm for the detection time of a single picture. An effective method is provided to analyze concrete meso-damage evolution through the research.

The accuracy of measuring the target object displacement is greatly influenced by the offset of central coordinate value in a laser speckle contour (CCVLSC) due to the defects obtained in background or on measuring surface, as a measuring combination of both monocular vision and laser speckle is used. In this paper, the theoretical principle of displacement measurement is first presented by a combination of monocular vision and laser speckle. Then, a model between object displacement and CCVLSC (particularly, $Y$ coordinate value) is derived. Finally, a denoising algorithm with competitive protection of contour effective points is proposed, on the basis of effects of noises coming from background and contour edge on CCVLSC. The algorithm includes ellipse fitting to laser speckle contour, calculating offsets between all contour points and the fitted eclipse, eliminating noise points with higher deviation (generally about 5% of all contour points) by using competitive strategy, ellipse refitting, and recalculating and re-eliminating until the deviation is below a specified threshold. It is shown that the algorithm can not only eliminate the fixed noise points in each round but also protect the number of effective points to the greatest extent. Finally, the feasibility of the algorithm is verified by two ways. One is an ideal data validation. It proves that the algorithm can guarantee the convergence towards the ideal center coordinate value. Another is an experimental verification. An experimental system is built up based on the relationship between object displacement and Y coordinate value of CCVLSC for obtaining relevant dada. It is shown by the comparison between predictions and experimental data that the algorithm has a better robustness and a higher accuracy of distance measurement than other typical algorithms.

Speech recognition in multilingual scenario is not trivial in the case when multiple languages are used in one conversation. Language must be identified before we process speech recognition as such tools are language-dependent. We present a language identification system (or AI tool) to distinguish top-seven world languages namely Chinese, Spanish, English, Hindi, Arabic, Bangla and Portuguese [G. F. Simons and C. D. Fennig (eds.), Ethnologue: Laguage of the Americas and the Pacific, Twentieth Edn. (SIL Internatinal, 2017)]. The system uses linear predictive coefficients-based feature, i.e. the line spectral pair–grade ratio (LSP–GR) feature, and ensemble learning for classification. Experiments were performed on more than 200h of real-world YouTube data and the highest possible accuracy of 96.95% was received. The results can be compared with other machine learning classifiers.

With the advent of the era of big data, people are eager to extract valuable knowledge from the rapidly expanding data, so that they can more effectively use these massive storage data. The traditional data processing technology can only achieve basic functions such as data query and statistics, and cannot achieve the goal of extracting the knowledge existing in the data to predict the future trend. Therefore, along with the rapid development of database technology and the rapid improvement of computer’s computing power, data mining (DM) came into existence. Research on DM algorithms includes knowledge of various fields such as database, statistics, pattern recognition and artificial intelligence. Pattern recognition mainly extracts features of known data samples. The DM algorithm using pattern recognition technology is a better method to obtain effective information from massive data, thus providing decision support, and has a good application prospect. Support vector machine (SVM) is a new pattern recognition algorithm proposed in recent years, which avoids dimension disaster by dimensioning and linearization. Based on this, this paper studies the DM algorithm based on pattern recognition, and proposes a DM algorithm based on SVM. The algorithm divides the vector of the SV set into two different types and iterates through multiple iterations to obtain a classifier that converges to the final result. Finally, through the cross-validation simulation experiment, the results show that the DM algorithm based on pattern recognition can effectively reduce the training time and solve the mining problem of massive data. The results show that the algorithm has certain rationality and feasibility.

With the development of industry, the fault diagnosis requirements for rolling bearings are getting higher and higher. This paper aims to develop low-complexity solutions for bearing fault diagnosis. In this paper, we use wavelet decomposition to obtain gesture Monitoring Index Vector (MIVs), after this, an improved Hidden Markov Model (HMM) algorithm was proposed for bearing fault diagnosis, in which we apply the Genetic Algorithm (GA) to avoid the convergence to local optimum, thus improving the recognition performance. The experimental results on 11 groups of test datasets demonstrate that the proposed algorithm (GAHMM) can achieve a higher average recognition rate of 93%, 87%, 87%, 93%, 93%, 97%, 100%, 97%, 97%, 100%, 97%.

In this paper, the effect of water-cement ratio and chloride ions on the concrete meso-structure was studied. Three kinds of concrete cubes with different water-cement ratios were immersed in fresh water and salt water, respectively. Then, the Electrochemical Impedance Spectroscopy (EIS) analysis of various test cubes were carried out by using electrochemical workstation. The results show that the salt water can improve electric double layer capacitance in the test cubes with the same water-cement ratio, but it can reduce some other parameters such as resistance of pore solution, resistance to transfer the hydrated electron, coefficient of diffusion impedance of concreter, which shows that the chloride ions diffused into the concrete in salt water and increase the ionic concentration in pore solution and C-S-H gel. However, the phase angle index is constant whether in fresh water or salt water, which shows chloride ions cannot affect the concrete meso-structure even though they can improve the ion concentration of pore structure. For the concrete test cubes which has different water-cement ratio in salt water, with the reduction of water-cement ratio, the electric double-layer capacitance of concrete remains unchanged, which indicates when the water-cement ratio becomes smaller, the porosity becomes lower, and the internal structure of concrete becomes denser.

The thorniest difficulties for multi-objective evolutionary algorithms (MOEAs) handling many-objective optimization problems (MaOPs) are the inefficiency of selection operators and high computational cost. To alleviate such difficulties and simplify the MaOPs, objective reduction algorithms have been proposed to remove the redundant objectives during the search process. However, those algorithms can only be applicable to specific problems with redundant objectives. Worse still, the Pareto solutions obtained by reduced objective set may not be the Pareto solutions of the original MaOPs. In this paper, we present a novel objective grouping evolutionary algorithm (OGEA) for general MaOPs. First, by dividing original objective set into several overlapping lower-dimensional subsets in terms of interdependence correlation information, we aim to separate the MaOPs into a number of sub-problems so that each of them can be able to preserve as much dominance structure in the original objective set as possible. Subsequently, we employ the nondominated sorting genetic algorithm II (NSGA-II) to generate Pareto solutions. Besides, instead of nondominated sorting on the whole population, a novel dual selection mechanism is proposed to choose individuals either having high ranks in subspaces or locating sparse region in the objective space for better proximity and diversity. Finally, we compare the proposed strategy with the other two classical space partition methods on benchmark DTLZ5 (I, M), DTLZ2 and a practical engineering problem. Numerical results show the proposed objective grouping algorithm can preserve more dominance structure in original objective set and achieve better quality of Pareto solutions.