Fast Anomaly Research Articles

Upper mantle structure beneath the Mongolian region from multimode surface waves: Implications for the western margin of Amurian plate

Multimode phase speeds of surface waves are used to build a new radially anisotropic S wave model in the eastern Eurasian and Mongolian regions. Our dataset includes seismic waveforms of over 1655 teleseismic events (Mw≥5.8) from 2009 to 2021, recorded at permanent and temporary stations in and around Mongolia. The multimode dispersion curves of Love and Rayleigh waves were extracted using the nonlinear waveform fitting method for individual seismograms. Then, we retrieved phase speed maps for each mode and frequency, incorporating finite-frequency effects. Finally, localized multimode dispersion curves extracted from the phase speed maps were inverted for local 1-D SV and SH wave profiles, which are combined into a radially anisotropic 3-D shear wave model. Our new model exhibits significant lateral variations of S wave speeds at 70–100 km depth beneath Mongolia, i.e., slow anomalies in the tectonically active western Mongolia in contrast to fast anomalies in stable eastern Mongolia. In the radial anisotropy model, SH waves are faster than SV waves in most areas of the Mongolian lithosphere above 100 km depth, except for the northeast of the Altay Mountains. The Hangay Dome region is characterized by significantly slower velocities that may relate to its uplifting. A large-scale low velocity beneath the northeast of the Hangay Dome with a slower SV wave speed than SH may indicate the existence of partially molten layers. This study also reveals distinct lateral variations of S wave speeds across the boundary between the Amurian and Eurasian plates, characterized by the fast anomaly in eastern Mongolia, corresponding to the lithosphere in the western Amurian plate.

Fast Anomaly Detection for IoT Services Based on Multisource Log Fusion

Fast Anomaly Detection for IoT Services Based on Multisource Log Fusion

Ultra-Lightweight Fast Anomaly Detectors for Industrial Applications.

Quality inspection in the pharmaceutical and food industry is crucial to ensure that products are safe for the customers. Among the properties that are controlled in the production process are chemical composition, the content of the active substances, and visual appearance. Although the latter may not influence the product's properties, it lowers customers' confidence in drugs or food and affects brand perception. The visual appearance of the consumer goods is typically inspected during the packaging process using machine vision quality inspection systems. In line with the current trends, the processing of the images is often supported with deep neural networks, which increases the accuracy of detection and classification of faults. Solutions based on AI are best suited to production lines with a limited number of formats or highly repeatable production. In the case where formats differ significantly from each other and are often being changed, a quality inspection system has to enable fast training. In this paper, we present a fast method for image anomaly detection that is used in high-speed production lines. The proposed method meets these requirements: It is easy and fast to train, even on devices with limited computing power. The inference time for each production sample is sufficient for real-time scenarios. Additionally, the ultra-lightweight algorithm can be easily adapted to different products and different market segments. In this work, we present the results of our algorithm on three different real production data gathered from food and pharmaceutical industries.

A Feature Memory Rearrangement Network for Visual Inspection of Textured Surface Defects Toward Edge Intelligent Manufacturing

Recent advances in the industrial inspection of textured surfaces-in the form of visual inspection-have made such inspections possible for efficient, flexible manufacturing systems. We propose an unsupervised feature memory rearrangement network (FMR-Net) to accurately detect various textural defects simultaneously. Consistent with mainstream methods, we adopt the idea of background reconstruction; however, we innovatively utilize artificial synthetic defects to enable the model to recognize anomalies, while traditional wisdom relies only on defect-free samples. First, we employ an encoding module to obtain multiscale features of the textured surface. Subsequently, a contrastive-learning-based memory feature module (CMFM) is proposed to obtain discriminative representations and construct a normal feature memory bank in the latent space, which can be employed as a substitute for defects and fast anomaly scores at the patch level. Next, a novel global feature rearrangement module (GFRM) is proposed to further suppress the reconstruction of residual defects. Finally, a decoding module utilizes the restored features to reconstruct the normal texture background. In addition, to improve inspection performance, a two-phase training strategy is utilized for accurate defect restoration refinement, and we exploit a multimodal inspection method to achieve noise-robust defect localization. We verify our method through extensive experiments and test its practical deployment in collaborative edge--cloud intelligent manufacturing scenarios by means of a multilevel detection method, demonstrating that FMR-Net exhibits state-of-the-art inspection accuracy and shows great potential for use in edge-computing-enabled smart industries.

[formula omitted] wave teleseismic tomography of the subducted Chile rise

We investigate the crustal and upper mantle structure of the Aysén region of Chile from 44° to 49° S, a place where the diverging oceanic Nazca and Antarctic plates subduct beneath the South American continent. The Seismic Experiment in the Aysén Region of Chile (SEARCH) project operated a network of up to 60 land-based seismometers in this region between 2004 and 2006, centered over a 6 Myr old subducted spreading center. The data are used to examine the P wave velocity structure beneath the region using relative-arrival teleseismic travel time tomography, using 2534 P wave residuals from 173 teleseismic earthquakes. It is possible to image the velocity structure beneath the seismic network from ∼30 down to ∼300 km depth. The model can resolve structures at a spatial scale between ∼60 and ∼200 km and shows a large difference between the northern and southern parts of the region. To the north, a ∼100 km thick fast anomaly exists which dips away from the subduction trench; likely to be related to the subducting Nazca plate. Going to the south, as the age of this plate at the subduction trench decreases and arc volcanism shuts off, the fast anomaly migrates further from the trench, suggesting that the Nazca plate subducts at a reduced angle over a larger distance before the subduction angle steepens. The distinct sections of the fast anomaly suggest that slab tears exist across the fracture zones between subducted plate segments. Where the 6 Myr old subducted ridge segment is predicted to lie, there is a region of low velocity between ∼100 and ∼300 km depth, and no fast region associated with a subducting slab is present, indicating the presence of an asthenospheric window between the subducted Nazca and Antarctic plates.

Federated Graph Neural Network for Fast Anomaly Detection in Controller Area Networks

Due to the lack of CAN frame encryption and authentication, CAN bus is vulnerable to various attacks, which can in general be divided into message injection, suspension, and falsification. Existing CAN bus anomaly detection mechanisms either can only detect one or two of these attacks, or require numerous CAN messages during predictions, which can hardly realize real-time performance. In this paper, we propose a CAN bus anomaly detection system that can detect all these attacks simultaneously in as short as 3 milliseconds (ms) based on Graph Neural Network (GNN). This work generates directed attributed graphs based on CAN message streams in given message intervals. Node attributes denote data contents in CAN messages while each edge attribute represents the frequency of a typical CAN ID pair in the given interval. Afterwards, a GNN is trained based on generated CAN message graphs. Considering highly imbalanced training data, a two-stage classifier cascade is developed in this paper, which is composed of a one-class classifier for anomaly detection and a multi-class classifier for attack classification. An openmax layer is further introduced to the multi-class classifier to tackle new anomalies from unknown classes. To take advantage of crowdsourcing while protecting user data privacy, we adopt federated learning to train a universal model that covers different driving scenarios and vehicle states. Extensive experiment results show the effectiveness and efficiency of our methodology.

A Fast Anomaly Diagnosis Approach Based on Modified CNN and Multisensor Data Fusion

In this article, a modified convolutional neural network (CNN) algorithm, namely 1D-GAPCNN-SVM, is proposed to address the early anomaly diagnosis problem. Considering the fact that traditional 2D-CNN based approaches contain too many model parameters and are not suitable for fast diagnosis applications using multisensor 1-D time-series measurements, 1D-CNN is introduced to deal with this problem. To reduce the number of parameters, a 1-D global average pooling layer is designed to substitute the fully connected layer with two or three layers. In order to further improve the diagnosis accuracy, a nonlinear multiclass support vector machine (SVM) is adopted to replace the traditional Softmax classifier as the final discriminator. Raw multisensor 1-D time-series data are directly fed into the diagnosis model, then the diagnosis result can be automatically generated. Two experiments, which are rolling bearing and GPS anomaly detection, have been conducted to demonstrate the effectiveness and the superior performance of the proposed method compared to the conventional SVM, K-nearest neighbor, deep neural network (DNN), and traditional 2D-CNN.

Lithosphere Structure of the Southern Dinarides and Continuity of the Adriatic Lithosphere Slab Beneath the Northern Dinarides Unravelled by Seismic Modelling

Currently recognised lithospheric models hypothesise the non-existence of a lithospheric slab (a so-called slab gap) in the area of the Northern Dinarides, and the possible existence of a shallow slab in the Southern Dinarides. These geological models are mostly based on previous regional and global 3D velocity models obtained from teleseismic tomography. Recent local tomographic models providing a good resolution have regularly shown the existence of a fast anomaly underneath the entire Dinarides, directly indicating the existence of a lithospheric slab. To avoid interpretation pitfalls and increase reliability, forward seismic modelling based on new tomographic models was performed. Seismic modelling indicates a continuous lithospheric slab along the entire Dinarides in the shallow mantle, but it is not continuous vertically. In the Northern Dinarides, the shallow lithospheric slab extends at least to a depth of 150 km. In the Southern and Central Dinarides, there is a deep fast anomaly that can be interpreted in two ways due to the weak vertical resolution of teleseismic tomography. The first model suggests a steeply dipping continuous Adriatic lithospheric slab whereas the second model shows that the slab consists of two separate blocks, meaning that the deeper block was formed by delamination of the Adriatic lithospheric slab. Due to a similar correlation between the inverse velocity models for the synthetic model and the observed data, preference is not given to any model. The second model could indicate two independent Dinaridic subduction phases, as opposed to viewing subduction as a single long process during the geological past.

Generative Adversarial Networks for anomaly detection in aerial images

Generative Adversarial Networks (GANs) are commonly used as a system able to perform unsupervised learning. We propose and demonstrate the use of a GAN architecture, known as the fast Anomaly Generative Adversarial Network (f-AnoGAN), to solve the problem of anomaly detection from aerial images. This architecture was previously applied to medical images and, in this work, we adapt it for use on satellite or aerial photographs. To test the effectiveness of this approach, we implemented anomaly detection schemes based on the Bi-directional Generative Adversarial Network (BiGAN), the image-z-image mapping (izi), the z-image-z (ziz) mapping, and a deep convolutional autoencoder (AE). The results show that the f-AnoGAN outperformed others, achieving AUC (area under the curve) values of 0.99 and 0.92 for urban and rural spaces image sets, respectively.

Fast anomaly detection with locality-sensitive hashing and hyperparameter autotuning

This paper presents LSHAD, an anomaly detection (AD) method based on Locality Sensitive Hashing (LSH), capable of dealing with large-scale datasets. The resulting algorithm is highly parallelizable and its implementation in Apache Spark further increases its ability to handle very large datasets. Moreover, the algorithm incorporates an automatic hyperparameter tuning mechanism so that users do not have to implement costly manual tuning. Our LSHAD method is novel as both hyperparameter automation and distributed properties are not usual in AD techniques. Our results for experiments with LSHAD across a variety of datasets point to state-of-the-art AD performance while handling much larger datasets than state-of-the-art alternatives. In addition, evaluation results for the tradeoff between AD performance and scalability show that our method offers significant advantages over competing methods.

Application of Fast P2P Traffic Recognition Technology Based on Decision Tree in the Detection of Network Traffic Data

With the rapid development of large-scale enterprise informatization construction, the network scale has become huge and complex, and the data traffic carried by the network is increasing. Accurate network traffic identification is the basis of network management and is of great significance to enterprise informatization construction and operation and maintenance. In response to the network operation and maintenance requirements of large enterprises, this paper analyzes the network architecture and network traffic distribution of large enterprise groups from the perspective of enterprise network operators and introduces the current operation and maintenance process of enterprise network performance failures. Maintenance process optimization and reengineering are carried out to plan and find out the shortcomings of the current process links and put forward corresponding solutions. Based on the research of traffic identification in recent years, a fast P2P network traffic anomaly identification algorithm based on decision tree model is proposed, which improves the efficiency and accuracy of network traffic application identification and network traffic anomaly data identification.

Low-Power Hyperspectral Anomaly Detector Implementation in Cost-Optimized FPGA Devices

Onboard data processing for on-the-fly decision-making applications has recently gained momentum in the field of remote sensing. In this context, hyperspectral anomaly detection has received special attention since its main purpose lies in the identification of abnormal events in an unsupervised manner. Nevertheless, onboard real-time hyperspectral image processing still poses several challenges before becoming a reality. This is why there is an emerging trend toward the development of hardware-friendly algorithmic solutions embedded in reconfigurable devices. In this context, this work contributes to a hardware architecture that ensures a progressive line processing in time-sensitive applications limited by the scarcity of hardware resources. In this sense, we have implemented the state-of-the-art hardware-friendly line-by-line fast anomaly detector for hyperspectral imagery (HW-LbL-FAD) detector on a reconfigurable hardware for a real-time performance. Specifically, we have selected a cost-optimized field-programmable gate array (ZC7Z020-CLG484) to implement our solution whose results draw up a good tradeoff between the following three features: time performance, energy consumption, and cost. The experimental results indicate that our hardware component is able to process hyperspectral images of 825x1024 pixels and 160 bands in 0.51 s with a power budget of 1.3 W and costs around 150€.Regarding detection performance, the HW-LbL-FAD algorithm outperforms other state-of-the-art algorithms.

Imaging structure and geometry of slabs in the greater Alpine area – a P-wave travel-time tomography using AlpArray Seismic Network data

Abstract. We perform a teleseismic P-wave travel-time tomography to examine the geometry and structure of subducted lithosphere in the upper mantle beneath the Alpine orogen. The tomography is based on waveforms recorded at over 600 temporary and permanent broadband stations of the dense AlpArray Seismic Network deployed by 24 different European institutions in the greater Alpine region, reaching from the Massif Central to the Pannonian Basin and from the Po Plain to the river Main. Teleseismic travel times and travel-time residuals of direct teleseismic P waves from 331 teleseismic events of magnitude 5.5 and higher recorded between 2015 and 2019 by the AlpArray Seismic Network are extracted from the recorded waveforms using a combination of automatic picking, beamforming and cross-correlation. The resulting database contains over 162 000 highly accurate absolute P-wave travel times and travel-time residuals. For tomographic inversion, we define a model domain encompassing the entire Alpine region down to a depth of 600 km. Predictions of travel times are computed in a hybrid way applying a fast TauP method outside the model domain and continuing the wave fronts into the model domain using a fast marching method. We iteratively invert demeaned travel-time residuals for P-wave velocities in the model domain using a regular discretization with an average lateral spacing of about 25 km and a vertical spacing of 15 km. The inversion is regularized towards an initial model constructed from a 3D a priori model of the crust and uppermost mantle and a 1D standard earth model beneath. The resulting model provides a detailed image of slab configuration beneath the Alpine and Apenninic orogens. Major features are a partly overturned Adriatic slab beneath the Apennines reaching down to 400 km depth still attached in its northern part to the crust but exhibiting detachment towards the southeast. A fast anomaly beneath the western Alps indicates a short western Alpine slab whose easternmost end is located at about 100 km depth beneath the Penninic front. Further to the east and following the arcuate shape of the western Periadriatic Fault System, a deep-reaching coherent fast anomaly with complex internal structure generally dipping to the SE down to about 400 km suggests a slab of European origin limited to the east by the Giudicarie fault in the upper 200 km but extending beyond this fault at greater depths. In its eastern part it is detached from overlying lithosphere. Further to the east, well-separated in the upper 200 km from the slab beneath the central Alps but merging with it below, another deep-reaching, nearly vertically dipping high-velocity anomaly suggests the existence of a slab beneath the eastern Alps of presumably the same origin which is completely detached from the orogenic root. Our image of this slab does not require a polarity switch because of its nearly vertical dip and full detachment from the overlying lithosphere. Fast anomalies beneath the Dinarides are weak and concentrated to the northernmost part and shallow depths. Low-velocity regions surrounding the fast anomalies beneath the Alps to the west and northwest follow the same dipping trend as the overlying fast ones, indicating a kinematically coherent thick subducting lithosphere in this region. Alternatively, these regions may signify the presence of seismic anisotropy with a horizontal fast axis parallel to the Alpine belt due to asthenospheric flow around the Alpine slabs. In contrast, low-velocity anomalies to the east suggest asthenospheric upwelling presumably driven by retreat of the Carpathian slab and extrusion of eastern Alpine lithosphere towards the east while low velocities to the south are presumably evidence of asthenospheric upwelling and mantle hydration due to their position above the European slab.

A Fast Hyperspectral Anomaly Detection Algorithm Based on Greedy Bilateral Smoothing and Extended Multi-Attribute Profile

To address the difficulty of separating background materials from similar materials associated with the use of “single-spectral information” for hyperspectral anomaly detection, a fast hyperspectral anomaly detection algorithm based on what we term the “greedy bilateral smoothing and extended multi-attribute profile” (GBSAED) method is proposed to improve detection precision and operation efficiency. This method utilizes “greedy bilateral smoothing” to decompose the low-rank part of a hyperspectral image (HSI) dataset and calculate spectral anomalies. This process improves the operational efficiency. Then, the extended multi-attribute profile is used to extract spatial anomalies and restrict the shape of anomalies. Finally, the two components are combined to limit false alarms and obtain appropriate detection results. This new method considers both spectral and spatial information with an improved structure that ensures operational efficiency. Using five real HSI datasets, this study demonstrates that the GBSAED method is more robust than eight representative algorithms under diverse application scenarios and greatly improves detection precision and operational efficiency.

A Fast Recursive LRX Algorithm with Extended Morphology Profile for Hyperspectral Anomaly Detection

In this paper, a fast anomaly target detection method based on morphological profile and improved Reed-Xiaoli (RX) is proposed. First, an extended morphological profile (EMP) containing spatial information is extracted from the original hyperspectral images by means of mathematical morphological transformations. Moreover, a novel fast local RX (FLRX) algorithm is also proposed. This algorithm iteratively updates the inverse matrix of covariance using matrix inversion lemma, thereby reducing the computational complexity of the Mahalanobis distance and improving the algorithm's calculation speed. Finally, a combined EMP and FLRX detector, named the EMP-FLRX method is constructed; as this approach can effectively utilize the spectral information and spatial information of hyperspectral images, it greatly improves detection accuracy and reduces the running time. We compare the proposed method with some classical and recently proposed approaches on six real datasets. the area under the curve (AUC) value of EMP-FLRX on six datasets is 0.9978, 0.9822, 0.9780, 0.9492, 0.9999 and 0.9852 respectively, while the running time is 9.4070, 14.4330, 6.2478, 9.0242, 19.9820, and 1.9060 s respectively. Experimental results clearly demonstrate that the performance of our proposed method is quite competitive in terms of detection accuracy and running time.

Local Earthquake Tomography in the Tjörnes Fracture Zone (North Iceland)

AbstractLocal earthquake tomography has been carried out in the Tjörnes Fracture Zone. This transform region connects the Mid‐Atlantic Ridge to the Northern Volcanic Zone in Iceland in a mostly offshore area. The challenge to record seismic information in this area was the motivation for the North ICeland Experiment (NICE). Fourteen ocean‐bottom seismometers and eleven on‐land stations were installed in the project and operated simultaneously with the permanent Icelandic seismic network (SIL) during summer 2004. Data from the experiment were used to estimate P‐ and S‐wave crustal velocities. Also, the gravity anomaly was derived for comparison with the tomographic results. Upper‐crustal velocities are found to be relatively low in the offshore region. In particular, low velocities are mapped along the Húsavík‐Flatey Fault, where a more confined negative gravity anomaly and a sedimentary basin are found. Low velocities are also mapped along the Grímsey Oblique Rift and in a zone connecting these two main lineaments. The northern half of the aseismic Grímsey Shoal appears as a fast anomaly. Furthermore, localized high‐velocity anomalies are found beneath northern Tröllaskagi and Flateyjarskagi Peninsulas, where bedrock dates from Upper and Middle Miocene (10–15 Ma). Regions of low Vp/Vs ratios are mapped at depth along the main lineaments. Low velocities along the lineaments are interpreted as due to fracturing extending into the middle crust, while high velocities in the upper crust beneath Tertiary formations are associated with relic volcanoes. Low Vp/Vs variations along the lineaments are interpreted as due to the presence of supercritical fluids.

Seismic evidence for craton chiseling and displacement of lithospheric mantle by the Tintina fault in the northern Canadian Cordillera

AbstractThe northern Canadian Cordillera (NCC) of northwestern Canada is segmented by several margin-parallel, right-lateral, strike-slip faults that accumulated several hundred kilometers of displacement between the Late Cretaceous and the Eocene. The depth extent of these faults, notably the Tintina fault (TF), has important implications for the tectonic assemblage and evolution of NCC lithospheric mantle, but geophysical models and geochemical data remain inconclusive. Using a recent three-dimensional P-wave seismic velocity model, we resolved a series of sharp (∼10 km) P-wave velocity contrasts (∼4%) at uppermost mantle depths beneath the surface trace of the TF. Seismic anisotropy data that represent upper-mantle fabrics revealed similar changes in the orientation and magnitude of anisotropy in the vicinity of the TF. These data suggest that the TF is a lithospheric-scale shear zone. After restoration of 430 km of right-lateral displacement along the TF, fast P-wave anomalies align with the outline of the North American craton margin. We propose the fast anomaly structure currently located in eastern Alaska represents a fragment of the Mackenzie craton that was chiseled and displaced to the northwest by the TF between the Late Cretaceous and the Eocene. A second cratonic fragment currently located in the southern NCC may be associated with the Cassiar terrane at upper-mantle depth. These observations provide the first evidence that large lithospheric-scale shear zones cut through refractory mantle and produce major lateral displacement of cratonic mantle material within cordilleras worldwide.

A Line-by-Line Fast Anomaly Detector for Hyperspectral Imagery

In recent years, anomaly detection (AD) has enjoyed a growing interest in hyperspectral data analysis. However, most state-of-the-art detectors need to work with the entire hyperspectral cube, what prevents their use for applications under real-time constraints, especially when the hyperspectral data are collected by push-broom scanners that acquire the hyperspectral images (HSIs) in a line-by-line fashion. In this paper, a Line-by-Line Fast Anomaly Detector for Hyperspectral Imagery (LbL-FAD) is proposed, which is capable of processing each sensed line as soon as it is captured. The LbL-FAD works under the assumption that anomalous pixels cannot be well represented by the background distribution. It uses an orthogonal projection strategy for extracting a set of pixels from the first captured hyperspectral frames, i.e., lines of pixels, that are used for representing the background distribution. Using these pixels, the LbL-FAD proposes a hardware-friendly alternative to compute the orthogonal subspace to that spanned by the selected background samples, making the anomalous pixels better detectable. In addition, the LbL-FAD incorporates an automatic thresholding method which provides line-by-line and real-time binary maps where anomalous targets are segmented from the background. This novelty clearly differentiates the proposed LbL-FAD from the conventional anomaly detectors, which usually are not able to automatically discriminate anomalous pixels from background pixels until the entire image is processed. Several experiments have been carried out using real HSIs collected by different sensors. The obtained results clearly support the benefits of our proposal, both in terms of the accuracy of the detection performance and the computational complexity.

Seismic Evidence for a Paleoslab in the D" Layer Residing Adjacent to the Southeastern Edge of the Perm Anomaly

AbstractThe Perm Anomaly, which seismologically resembles the large low‐shear velocity provinces (LLSVPs) in the lowermost mantle under Africa and the South Pacific, but is considerably smaller, has been identified beneath Eurasia. Numerical simulations have indicated that the LLSVPs dynamically interact with subducted slabs in D", producing strong deformation near the edges of the LLSVPs. The existence of deformation has been corroborated by seismic anisotropic observations in the D" layer nearby the borders of the slow anomalies such as the Perm Anomaly (Long & Lynner, 2015, https://doi.org/10.1002/2015GL065506; Thomas & Kendall, 2002, https://doi.org/10.1046/j.1365-246X.2002.01760.x). The existence of paleoslabs, however, has rarely been reported. Our purpose of this study, therefore, is to detect fast anomalies associated with paleoslabs near the edges of the Perm Anomaly. Here we find evidence of the existence of the D" discontinuity at a height above the core‐mantle boundary of 250 km with a +3% shear wave velocity increase; it is located adjacent to the southeastern edge of the Perm Anomaly, which is most likely due to a phase transition from perovskite to postperovskite in conjunction with the presence of slab remnants, by analyzing the Scd (the lower mantle triplication) phases recorded at stations in southeastern Europe. The existence of fast anomaly in the D" region (likely due to accumulation of a paleoslab) is further confirmed by analyzing differential traveltime residuals of SKS versus SKKS and SKKKS measured at stations in North America. The subducted slab—probably attributed to the Central China slab—at the base of the mantle may play a role in influencing the shape and location of the Perm Anomaly.

Deep-cascade: Cascading 3D Deep Neural Networks for Fast Anomaly Detection and Localization in Crowded Scenes.

This paper proposes a fast and reliable method for anomaly detection and localization in video data showing crowded scenes. Time-efficient anomaly localization is an ongoing challenge and subject of this paper. We propose a cubicpatch- based method, characterised by a cascade of classifiers, which makes use of an advanced feature-learning approach. Our cascade of classifiers has two main stages. First, a light but deep 3D auto-encoder is used for early identification of "many" normal cubic patches. This deep network operates on small cubic patches as being the first stage, before carefully resizing remaining candidates of interest, and evaluating those at the second stage using a more complex and deeper 3D convolutional neural network (CNN). We divide the deep autoencoder and the CNN into multiple sub-stages which operate as cascaded classifiers. Shallow layers of the cascaded deep networks (designed as Gaussian classifiers, acting as weak single-class classifiers) detect "simple" normal patches such as background patches, and more complex normal patches are detected at deeper layers. It is shown that the proposed novel technique (a cascade of two cascaded classifiers) performs comparable to current top-performing detection and localization methods on standard benchmarks, but outperforms those in general with respect to required computation time.