Image-based Pattern Recognition Research Articles

Versatility Evaluation of Landslide Risk with Window Sizes and Sampling Techniques Based on Deep Learning

Across the globe, landslides cause significant loss of life, injuries, and widespread damage to homes and infrastructure. Therefore, assessing and analyzing landslide hazards is crucial to human, environmental, cultural, economic, and social sustainability. This study utilizes ArcGIS 10.8 and Python 3.9 to create landslide databases for Niigata Prefecture (NIG), Iwate and Miyagi Prefectures (IWT-MYG), and Hokkaido (HKD), drawing on data obtained from the National Research Institute for Earth Science and Disaster Resilience, Japan. A distinguishing feature of this study is the application of a Convolutional Neural Network (CNN), which significantly outperforms traditional machine learning models in image-based pattern recognition by extracting contextual information from surrounding areas, a distinct advantage in image and pattern recognition tasks. Unlike conventional methods that often require manual feature selection and engineering, CNNs automate feature extraction, enabling a more nuanced understanding of complex patterns. By experimenting with CNN input window sizes ranging from 3 × 3 to 27 × 27 pixels and employing diverse sampling techniques, we demonstrate that larger windows enhance the model’s predictive accuracy by capturing a wider range of environmental interactions critical for effective landslide modeling. CNN models with 19 × 19 pixel windows typically yield the best overall performance, with CNN-19 achieving an AUC of 0.950, 0.982 and 0.969 for NIG, HKD, and IWT-MYG, respectively. Furthermore, we improve prediction reliability using oversampling and a random window-moving method. For instance, in the NIG region, the AUC of the oversampling CNN-19 is 0.983, while the downsampling AUC is 0.950). These techniques, less commonly applied in traditional machine learning approaches to landslide detection, help address the issue of data imbalance often seen in landslide datasets, where instances of landslides are far outnumbered by non-landslide occurrences. While challenges remain in enhancing the model’s generalization, this research makes significant progress in developing more robust and adaptable tools for landslide prediction, which are vital for ensuring environmental and societal resilience.

A machine learning approach to the geomorphometric detection of ribbed moraines in Norway

Abstract. Machine learning is a powerful yet underutilised tool in geomorphology, commonly used for image-based pattern recognition. Analysing new high-resolution (1–10 m) elevation datasets, we investigate its usefulness for detecting discrete geomorphological features. This study develops a machine-learning-based method for identifying ribbed moraines in digital elevation data and progresses to test its performance versus time-consuming, manual methods. Ribbed moraines share geomorphometric characteristics with other glacial landforms, hence representing a valuable test of our new methodology in terms of differentiating between similar features, and for detecting landforms with similar characteristics. Furthermore, mapping ribbed moraines may provide valuable indications of their origin, a topic of debate within glacial geomorphology. To automatically detect ribbed moraines, we extract simple morphometrics from high-resolution digital elevation model data and mask regions where ribbed moraines are unlikely to form. We then test several machine learning algorithms before examining the best performer (K-means clustering) for three study areas of 15 km2 in Norway. Our results demonstrate a balanced accuracy of 65 %–75 % when validating versus ground-truthing. The performance depends on the availability of high-resolution elevation data in Norway that are needed to resolve the spatial scale of the target (10–100 m). We find the method effective at detecting both fields of ribbed moraines, as well as individual ribbed moraines. We propose pathways for the future implementation of this method on a large scale and for increasing the detail of information gained about detected landforms. In conclusion, we demonstrate K-means clustering as a promising method for detecting ribbed moraines, with great potential to reduce the time needed to produce landform maps.

Suitability of curvature as a feature for image-based pattern recognition: a case study on leaf image classification based on machine learning

Suitability of curvature as a feature for image-based pattern recognition: a case study on leaf image classification based on machine learning

Pattern recognition framework using asynchronous discrete binary data for condition and damage assessment in plate-like structures

Recent advances in energy harvesting technologies have led to the evolution of self-powered structural health monitoring techniques that are energy-efficient. Concurrent to the emergence of self-powered sensing has been the development of power-efficient data communication protocols. One such technology is an energy-aware pulse communication architecture that employs ultrasonic pulses through the material substrate for information forwarding. This results in limited discrete binary data that raises the need for new data analysis methods for structural health monitoring purposes. A pattern recognition framework that allows interpretation of the resulting asynchronous discrete binary data for condition and damage assessment in plate-like structures is presented in this article. The proposed pattern recognition framework is based on integration of image-based pattern recognition using anomaly detection, a pattern anomaly measure, a focal density concept, and the k-nearest neighbor algorithm. Using numerical simulations, damage indication parameters were determined from the strain response of dynamically loaded plates. Simulated test cases considering different levels of damage severity, single and multiple damage regions, loading conditions, and measurement noise were studied to evaluate the effectiveness and robustness of the strategy. Furthermore, the effect of sensor density on the proposed strategy was explored. Results demonstrate satisfactory performance and robustness of the proposed pattern recognition framework for localized damage detection in plate-like structures using limited and low-resolution discrete binary data.

Plant disorder precognition by image based pattern recognition

Abstract Proliferating population in the world has ushered a lot of burden on resources in agriculture. In order to sustain the population and the economy, it is very important to obtain maximum yield from crops. Diseases in plants are the chief cause of damage in plants which leads to production loss in agriculture. Earlier disease detection is manual, where farmers refer guide books or use their experiences to recognize the diseases. This process is time-consuming and requires some knowledge for selection of pesticides. Capturing images of leaves and processing them with the help of an automated system to identify the disease infected leaves could be a better solution for farmers. This paper discuss about how image processing and machine learning is effectively used in detection of diseases in plants. Analysis is done with various classification techniques and the comparison is made between Active contour and Sobel edge detection techniques. Experimental results indicate that Multiclass Support Vector Machines combined with Active contour edge detection gives better results compared to other methods for disease classification.

Integrating scattering feature maps with convolutional neural networks for Malayalam handwritten character recognition

Convolutional neural network (CNN)-based deep learning architectures are the state-of-the-art in image-based pattern recognition applications. The receptive filter fields in convolutional layers are learned from training data patterns automatically during classifier learning. There are number of well-defined, well-studied and proven filters in the literature that can extract informative content from the input patterns. This paper focuses on utilizing scattering transform-based wavelet filters as the first-layer convolutional filters in CNN architecture. The scattering networks are generated by a series of scattering transform operations. The scattering coefficients generated in first few layers are effective in capturing the dominant energy contained in the input data patterns. The present work aims at replacing the first-layer convolutional feature maps in CNN architecture with scattering feature maps. This architecture is equivalent to utilizing scattering wavelet filters as the first-layer receptive fields in CNN architecture. The proposed hybrid CNN architecture experiments the Malayalam handwritten character recognition which is one of the challenging multi-class classification problems. The initial studies confirm that the proposed hybrid CNN architecture based on scattering feature maps could perform better than the equivalent self-learning architecture of CNN on handwritten character recognition problems.

Structural damage identification using image-based pattern recognition on event-based binary data generated from self-powered sensor networks

A continuing challenge in structural health monitoring is power availability for sensors to collect and communicate data. A way to minimize the communication power demand is to transmit the minimum amount of information, namely, one bit. Event-based binary signals are generated at sensor nodes according to local rules based on physical measurements, but interpretation at the global level requires dealing with discrete binary data, which implies system information with reduced resolution. This study presents an investigation on approaches for the interpretation of event-based binary data provided by a self-powered sensor network using a pulse-communication protocol for use in structural assessment and damage identification. Pattern recognition (PR) methods based on image data analysis techniques were adapted for such purpose. The methods used were classifiers based on anomaly detection, a Bayesian method, and a nearest neighbor method. To improve the performance of the approach, 2-dimensional principal component analysis and 2-dimensional linear discriminant analysis were used as feature extraction techniques along with a nearest neighbor classifier. The PR methods and the performance of the interpretation algorithms were evaluated by using virtual data from finite element simulations and real data from experiments on plates. The ability of the PR methods to identify service demands, load variations, and localized material degradation was examined. Results indicate that image-based PR methods can be effectively used for structural damage identification in plate-like structures using event-based binary data sets in novel wireless self-powered sensor networks.

An image-based pattern recognition approach to condition monitoring of reciprocating compressor valves

This work presents the development of a vibration-based condition monitoring method for early detection and classification of valve wear within industrial reciprocating compressors through the combined use of time-frequency analysis with image-based pattern recognition techniques. Two common valve related fault conditions are spring fatigue and valve seat wear and are seeded on the crank-side discharge valves of a Dresser-Rand ESH-1 industrial compressor. Operational data including vibration, cylinder pressure, and crank shaft position are collected and processed using a transformed time-frequency domain approach. The results are processed as images with features extracted using 1st and 2nd order image texture statistics and binary shape properties. Feature reduction is accomplished by principal component analysis and a Bayesian classification strategy is employed with accuracy rates greater than 90%.

Information Based Universal Feature Extraction in Shallow Networks

In many real-world image based pattern recognition tasks, the extraction and usage of task-relevant features are the most crucial part of the diagnosis. In the standard approach, either the features are given by common sense like edges or corners in image analysis, or they are directly determined by expertise. They mostly remain task-specific, although human may learn the life time, and use different features too, although same features do help in recognition. It seems that a universal feature set exists, but it is not yet systematically found. In our contribution, we try to find such a universal image feature set that is valuable for most image related tasks. We trained a shallow neural network for recognition of natural and non-natural object images before different backgrounds, including pure texture and handwritten digits, using a Shannon information-based algorithm and learning constraints. In this context, the goal was to extract those features that give the most valuable information for classification of the visual objects, hand-written digits and texture datasets by a one layer network and then classify them by a second layer. This will give a good start and performance for all other image learning tasks, implementing a transfer learning approach. As result, in our case we found that we could indeed extract unique features which are valid in all three different kinds of tasks. They give classification results that are about as good as the results reported by the corresponding literature for the specialized systems, or even better ones.

Reciprocating compressor valve condition monitoring using image-based pattern recognition

This work presents the development of a vibration-based condition monitoring method for early detection and classification of valve wear within industrial reciprocating compressors through the combined use of time-frequency analysis with image-based pattern recognition techniques. A common valve related fault condition is valve seat wear that is caused by repeated impact and accentuated by chatter. Seeded faults consistent with valve seat wear are introduced on the crankside discharge valves of a Dresser-Rand ESH-1 industrial compressor. A variety of operational data including vibration, cylinder pressure, and crank shaft position are collected and processed using a time-frequency domain approach. The resulting diagrams are processed as images with features extracted using 1st and 2nd order image statistics. A Bayesian classification strategy is employed with accuracy rates greater than 90% achieved using two and three-dimensional features spaces.

Image Color Registration for Illumination-Robust Pattern Recognition

Image-based pattern recognition techniques have attracted much attention in vision-based applications. Color-based methods have shown several benefits. However, due to color variations resulting from illumination changes, many color-based techniques have yet to demonstrate stable performance. For illumination-robust pattern recognition, we propose an image color registration method based on an image acquisition model. Since the image acquisition model is created using the variables related to an illumination condition, camera characteristics, and an object's surface reflectance, the proposed method normalizes the image's color by taking into account both the illumination and camera characteristics. To evaluate the performance of the proposed method in terms of illumination-robust pattern recognition, we perform both an image similarity test and a feature similarity test between images acquired under different illumination conditions. Through the experiments, the superiority and the usefulness of the proposed method was validated.

Efficient update of the covariance matrix inverse in iterated linear discriminant analysis

For fast classification under real-time constraints, as required in many image-based pattern recognition applications, linear discriminant functions are a good choice. Linear discriminant analysis (LDA) computes such discriminant functions in a space spanned by real-valued features extracted from the input. The accuracy of the trained classifier crucially depends on these features, its time complexity on their number. As the number of available features is immense in most real-world problems, it becomes essential to use meta-heuristics for feature selection and/or feature optimization. These methods typically involve iterated training of a classifier after substitutions or modifications of features. Therefore, we derive an efficient incremental update formula for LDA discriminant functions for the substitution of features. It scales linearly in the number of altered features and quadratically in the overall number of features, while completely retraining scales cubically in the number of features. The update rule allows for efficient feature selection and optimization with any meta-heuristic that is based on iteratively modifying existing solutions. The proposed method was tested on an artificial benchmark problem as well as on a real-world problem. Results show that significant time savings during training are achieved while numerical stability is maintained.