This PDF file contains the front matter associated with 10203 Proceedings Volume 8806, including the Title Page, Copyright information, Table of Contents, and Conference Committee listing.

Large-scale feed-forward neural networks have seen intense application in many computer vision problems. However, these networks can get hefty and computationally intensive with increasing complexity of the task. Our work, for the first time in literature, introduces a Cellular Simultaneous Recurrent Network (CSRN) based hierarchical neural network for object detection. CSRN has shown to be more effective to solving complex tasks such as maze traversal and image processing when compared to generic feed forward networks. While deep neural networks (DNN) have exhibited excellent performance in object detection and recognition, such hierarchical structure has largely been absent in neural networks with recurrency. Further, our work introduces deep hierarchy in SRN for object recognition. The simultaneous recurrency results in an unfolding effect of the SRN through time, potentially enabling the design of an arbitrarily deep network. This paper shows experiments using face, facial expression and character recognition tasks using novel deep recurrent model and compares recognition performance with that of generic deep feed forward model. Finally, we demonstrate the flexibility of incorporating our proposed deep SRN based recognition framework in a humanoid robotic platform called NAO.

This article presents a neural network based multi-spectral image segmentation method. A neural network is trained on the selected features of both the objects and background in the longwave (LW) Infrared (IR) images. Multiple iterations of training are performed until the accuracy of the segmentation reaches satisfactory level. The segmentation boundary of the LW image is used to segment the midwave (MW) and shortwave (SW) IR images. A second neural network detects the local discontinuities and refines the accuracy of the local boundaries. This article compares the neural network based segmentation method to the Wavelet-threshold and Grab-Cut methods. Test results have shown increased accuracy and robustness of this segmentation scheme for multi-spectral IR images.

Zeltmann, et al. demonstrated that structural integrity and other quality damage to objects can be caused by changing its position on a 3D printer’s build plate. On some printers, for example, object surfaces and support members may be stronger when oriented parallel to the X or Y axis. The challenge presented by the need to assure 3D printed object orientation is that this can be altered in numerous places throughout the system. This paper considers attack scenarios and discusses where attacks that change printing orientation can occur in the process. An imaging-based solution to combat this problem is presented.

Correlation filters are a well established means for target recognition tasks. However, the unintentional effect of circular correlation has a negative influence on the performance of correlation filters as they are implemented in frequency domain. The effects of aliasing are minimized by introducing zero aliasing constraints in the template and test image. In this paper, the comparative analysis of logarithmic zero aliasing optimal trade off correlation filters has been carried out for different types of target distortions. The zero aliasing Maximum Average Correlation Height (MACH) filter has been identified as the best choice based on our research for achieving enhanced results in the presence of any type of variance which are discussed in results section. The reformulation of the MACH expressions with zero aliasing has been made to demonstrate the achievable enhancement to the logarithmic MACH filter in target detection applications.

Feature extraction is a process used to reduce data dimensions using various transforms while preserving the discriminant characteristics of the original data. Feature extraction has been an important issue in pattern recognition since it can reduce the computational complexity and provide a simplified classifier. In particular, linear feature extraction has been widely used. This method applies a linear transform to the original data to reduce the data dimensions. The decision boundary feature extraction method (DBFE) retains only informative directions for discriminating among the classes. DBFE has been applied to various parametric and non-parametric classifiers, which include the Gaussian maximum likelihood classifier (GML), the k-nearest neighbor classifier, support vector machines (SVM) and neural networks. In this paper, we apply DBFE to deep neural networks. This algorithm is based on the nonparametric version of DBFE, which was developed for neural networks. Experimental results with the UCI database show improved classification accuracy with reduced dimensionality.

A fully invariant system helps in resolving difficulties in object detection when camera or object orientation and position are unknown. In this paper, the proposed correlation filter based mechanism provides the capability to suppress noise, clutter and occlusion. Minimum Average Correlation Energy (MACE) filter yields sharp correlation peaks while considering the controlled correlation peak value. Difference of Gaussian (DOG) Wavelet has been added at the preprocessing stage in proposed filter design that facilitates target detection in orientation variant cluttered environment. Logarithmic transformation is combined with a DOG composite minimum average correlation energy filter (WMACE), capable of producing sharp correlation peaks despite any kind of geometric distortion of target object. The proposed filter has shown improved performance over some of the other variant correlation filters which are discussed in the result section.

A new wavelet-filtered-based Shifted- phase-encoded Joint Transform Correlation (WPJTC) technique has been proposed for efficient face recognition. The proposed technique uses discrete wavelet decomposition for preprocessing and can effectively accommodate various 3D facial distortions, effects of noise, and illumination variations. After analyzing different forms of wavelet basis functions, an optimal method has been proposed by considering the discrimination capability and processing speed as performance trade-offs. The proposed technique yields better correlation discrimination compared to alternate pattern recognition techniques such as phase-shifted phase-encoded fringe-adjusted joint transform correlator. The performance of the proposed WPJTC has been tested using the Yale facial database and extended Yale facial database under different environments such as illumination variation, noise, and 3D changes in facial expressions. Test results show that the proposed WPJTC yields better performance compared to alternate JTC based face recognition techniques.

A commercially available iris recognition system uses only a narrow band of the near infrared spectrum (700-900 nm) while iris images captured in the wide range of 405 nm to 1550 nm offer potential benefits to enhance recognition performance of an iris biometric system. The novelty of this research is that a group selection algorithm based on coalition game theory is explored to select the best patch subsets. In this algorithm, patches are divided into several groups based on their maximum contribution in different groups. Shapley values are used to evaluate the contribution of patches in different groups. Results show that this group selection based iris recognition

The interference theory is developed for of the phase conjugate Michelson interferometer in which its ordinary mirrors are replaced by a single externally pumped phase conjugate mirror. According to the theory, it was found that for an interferometer with two equal arms, the path length difference depends solely on the initial alignment of the two input beams, and the vertical alignment readout. Small vertical misalignments in the readout beam by mrad causes a huge change in the phase difference in the phase between the two interferometer arms beam. The phase difference is proportional to the interferometer arm lengths. The overlap between the phase conjugate beams is not affected by the interferometer beam alignment. The interferometer is proposed for nondestructive testing and the design all optical logic and associated fuzzy logic for ultrafast optical pattern recognition.

Multiquadrics (MQ) are radial basis spline function that can provide an efficient interpolation of data points located in a high dimensional space. MQ were developed by Hardy to approximate geographical surfaces and terrain modelling. In this paper we frame the task of interactive image segmentation as a semi-supervised interpolation where an interpolating function learned from the user provided seed points is used to predict the labels of unlabeled pixel and the spline function used in the semi-supervised interpolation is MQ. This semi-supervised interpolation framework has a nice closed form solution which along with the fact that MQ is a radial basis spline function lead to a very fast interactive image segmentation process. Quantitative and qualitative results on the standard datasets show that MQ outperforms other regression based methods, GEBS, Ridge Regression and Logistic Regression, and popular methods like Graph Cut,4 Random Walk and Random Forest.6

The Modified Forward Backward Linear Prediction, MFBLP, is powerful technique that enables an adaptive dimensionality reduction of the data through the estimation of the frequency domain representation of the data poles and the utilization of the ensuing transfer function for dimensionality reduction of the data. In this work, we isolate a data-region that is expected to encompass the statistical features of a given anomalous event relative to the statistical common data points. The isolated anomalous events are then compared with the adaptively extracted data using the MFBLP and the comparison is utilized to isolate the anomalous events of interest. The effects of different levels of noise are discussed in relation to dimensionality reduction using Eigen-features alone and by using Eigen-features accompanied by MFBLP.

A missile warning system can detect the incoming missile threat(s) and automatically cue the other Electronic Attack (EA) systems in the suit, such as Directed Infrared Counter Measure (DIRCM) system and/or Counter Measure Dispensing System (CMDS). Most missile warning systems are currently based on passive sensor technology operating in either Solar Blind Ultraviolet (SBUV) or Midwave Infrared (MWIR) bands on which there is an intensive emission from the exhaust plume of the threatening missile. Although passive missile warning systems have some clear advantages over pulse-Doppler radar (PDR) based active missile warning systems, they show poorer performance in terms of time-to-impact (TTI) estimation which is critical for optimizing the countermeasures and also “passive kill assessment”. In this paper, we consider this problem, namely, TTI estimation from passive measurements and present a TTI estimation scheme which can be used in passive missile warning systems. Our problem formulation is based on Extended Kalman Filter (EKF). The algorithm uses the area parameter of the threat plume which is derived from the used image frame.

This research investigates the features retained after image compression for automatic pattern recognition purposes. Many raw images with vehicles in them were collected for these experiments. These raw images were significantly compressed using open-source JPEG and JPEG2000 compression algorithms. The original and compressed images are processed with a Map Seeking Circuit (MSC) pattern recognition algorithm, as well as a Histogram of Oriented Gradient (HOG) with Support Vector Machine (SVM) pattern recognition program. Detection rates are given for these images that demonstrates the feature extraction capabilities as well as false alarm rates when the compression was increased. JPEG2000 compression results show preservation of the features needed for automatic pattern recognition which was better than the JPEG standard image compression results.

A new two dimensional (2-D) Discrete Orthogonal Stcokwell Transform (DOST) based Local Phase Pattern (LPP) technique has been proposed for efficient face recognition. The proposed technique uses 2-D DOST as preliminary preprocessing and local phase pattern to form robust feature signature which can effectively accommodate various 3D facial distortions and illumination variations. The S-transform, is an extension of the ideas of the continuous wavelet transform (CWT), is also known for its local spectral phase properties in time-frequency representation (TFR). It provides a frequency dependent resolution of the time-frequency space and absolutely referenced local phase information while maintaining a direct relationship with the Fourier spectrum which is unique in TFR. After utilizing 2-D Stransform as the preprocessing and build local phase pattern from extracted phase information yield fast and efficient technique for face recognition. The proposed technique shows better correlation discrimination compared to alternate pattern recognition techniques such as wavelet or Gabor based face recognition. The performance of the proposed method has been tested using the Yale and extended Yale facial database under different environments such as illumination variation and 3D changes in facial expressions. Test results show that the proposed technique yields better performance compared to alternate time-frequency representation (TFR) based face recognition techniques.

In imagery and pattern analysis domain a variety of descriptors have been proposed and employed for different computer vision applications like face detection and recognition. Many of them are affected under different conditions during the image acquisition process such as variations in illumination and presence of noise, because they totally rely on the image intensity values to encode the image information. To overcome these problems, a novel technique named Multi-Texture Local Ternary Pattern (MTLTP) is proposed in this paper. MTLTP combines the edges and corners based on the local ternary pattern strategy to extract the local texture features of the input image. Then returns a spatial histogram feature vector which is the descriptor for each image that we use to recognize a human being. Experimental results using a k-nearest neighbors classifier (k-NN) on two publicly available datasets justify our algorithm for efficient face recognition in the presence of extreme variations of illumination/lighting environments and slight variation of pose conditions.

In this paper, we propose architectures for the implementation 16 Boolean optical gates from two inputs using externally pumped phase- conjugate Michelson interferometer. Depending on the gate to be implemented, some require single stage interferometer and others require two stages interferometer. The proposed optical gates can be used in several applications in optical networks including, but not limited to, all-optical packet routers switching, and all-optical error detection. The optical logic gates can also be used in recognition of noiseless rotation and scale invariant objects such as finger prints for home land security applications.

Extreme learning machine (ELM), as a single hidden layer feedforward neural network, has shown very effective performance in pattern analysis and machine intelligence; however, there are some limitations that constrain the performance of ELM, such as data multicollinearity issues. The generalization capability of ELM could be significantly deteriorated when multicollinearity is present in the hidden layer output matrix which causes the matrix to become singular or ill-conditioning. To overcome such a problem, ridge regression can be utilized. The conventional way to avoid multicollinearity in ELM is achieved by precisely adjusting the ridge constant, which may not be a sophisticate solution to obtain the optimal value. In this paper, we present a solution for finding a satisfactory ridge constant by incorporating variance inflation factors (VIF) during calculating output weights in ELM, we termed this technique as ELM-VIF. Experimental results on handwritten digit recognition show that the proposed ELM-VIF, compared with the original ELM, has better stability and generalization performance.

In prior work, Zeltmann, et al. demonstrated the negative impact that can be created by defects of various sizes in 3D printed objects. These defects may make the object unsuitable for its application or even present a hazard, if the object is being used for a safety-critical application. With the uses of 3D printing proliferating and consumer access to printers increasing, the desire of a nefarious individual or group to subvert the desired printing quality and safety attributes of a printer or printed object must be considered. Several different approaches to subversion may exist. Attackers may physically impair the functionality of the printer or launch a cyber-attack. Detecting introduced defects, from either attack, is critical to maintaining public trust in 3D printed objects and the technology. This paper presents an alternate approach. It applies a quality assurance technology based on visible light sensing to this challenge and assesses its capability for detecting introduced defects of multiple sizes.

By this paper, the major goal is to investigate the Multi-CPU/FPGA SoC (System on Chip) design flow and to transfer a know-how and skills to rapidly design embedded real-time vision system. Our aim is to show how the use of these devices can be benefit for system level integration since they make possible simultaneous hardware and software development. We take the facial detection and pretreatments as case study since they have a great potential to be used in several applications such as video surveillance, building access control and criminal identification. The designed system use the Xilinx Zedboard platform. The last is the central element of the developed vision system. The video acquisition is performed using either standard webcam connected to the Zedboard via USB interface or several camera IP devices. The visualization of video content and intermediate results are possible with HDMI interface connected to HD display. The treatments embedded in the system are as follow: (i) pre-processing such as edge detection implemented in the ARM and in the reconfigurable logic, (ii) software implementation of motion detection and face detection using either ViolaJones or LBP (Local Binary Pattern), and (iii) application layer to select processing application and to display results in a web page. One uniquely interesting feature of the proposed system is that two functions have been developed to transmit data from and to the VDMA port. With the proposed optimization, the hardware implementation of the Sobel filter takes 27 ms and 76 ms for 640x480, and 720p resolutions, respectively. Hence, with the FPGA implementation, an acceleration of 5 times is obtained which allow the processing of 37 fps and 13 fps for 640x480, and 720p resolutions, respectively.

Optical/Digital pattern recognition and tracking based on optical/digital correlation are a well-known techniques to detect, identify and localize a target object in a scene. Despite the limited number of treatments required by the correlation scheme, computational time and resources are relatively high. The most computational intensive treatment required by the correlation is the transformation from spatial to spectral domain and then from spectral to spatial domain. Furthermore, these transformations are used on optical/digital encryption schemes like the double random phase encryption (DRPE). In this paper, we present a VLSI architecture for the correlation scheme based on the fast Fourier transform (FFT). One interesting feature of the proposed scheme is its ability to stream image processing in order to perform correlation for video sequences. A trade-off between the hardware consumption and the robustness of the correlation can be made in order to understand the limitations of the correlation implementation in reconfigurable and portable platforms. Experimental results obtained from HDL simulations and FPGA prototype have demonstrated the advantages of the proposed scheme.

This paper addresses the problem of detecting and tracking weapon discharge event in an Infrared Imagery collection. While most of the prior work in related domains exploits the vast amount of complementary in- formation available from both visible-band (EO) and Infrared (IR) image (or video sequences), we handle the problem of recognizing human pose and activity detection exclusively in thermal (IR) images or videos. The task is primarily two-fold: 1) locating the individual in the scene from IR imagery, and 2) identifying the correct pose of the human individual (i.e. presence or absence of weapon discharge activity or intent). An efficient graph-based shortlisting strategy for identifying candidate regions of interest in the IR image utilizes both image saliency and mutual similarities from the initial list of the top scored proposals of a given query frame, which ensures an improved performance for both detection and recognition simultaneously and reduced false alarms. The proposed search strategy offers an efficient feature extraction scheme that can capture the maximum amount of object structural information by defining a region- based deep shape descriptor representing each object of interest present in the scene. Therefore, our solution is capable of handling the fundamental incompleteness of the IR imageries for which the conventional deep features optimized on the natural color images in Imagenet are not quite suitable. Our preliminary experiments on the OSU weapon dataset demonstrates significant success in automated recognition of weapon discharge events from IR imagery.

An image encryption method combing chaotic map and Arnold transform in the gyrator transform domains was proposed. Firstly, the original secret image is XOR-ed with a random binary sequence generated by a logistic map. Then, the gyrator transform is performed. Finally, the amplitude and phase of the gyrator transform are permutated by Arnold transform. The decryption procedure is the inverse operation of encryption. The secret keys used in the proposed method include the control parameter and the initial value of the logistic map, the rotation angle of the gyrator transform, and the transform number of the Arnold transform. Therefore, the key space is large, while the key data volume is small. The numerical simulation was conducted to demonstrate the effectiveness of the proposed method and the security analysis was performed in terms of the histogram of the encrypted image, the sensitiveness to the secret keys, decryption upon ciphertext loss, and resistance to the chosen-plaintext attack.

Air Traffic Management (ATM) concepts are commonly tested in simulation to obtain preliminary results and validate the concepts before adoption. Recently, the researchers found that simulation is not enough because of complexity associated with ATM concepts. In other words, full-scale tests must eventually take place to provide compelling performance evidence before adopting full implementation. Testing using full-scale aircraft produces a high-cost approach that yields high-confidence results but simulation provides a low-risk/low-cost approach with reduced confidence on the results. One possible approach to increase the confidence of the results and simultaneously reduce the risk and the cost is using unmanned sub-scale aircraft in testing new concepts for ATM. This paper presents the simulation results of using unmanned sub-scale aircraft in implementing ATM concepts compared to the full scale aircraft. The results of simulation show that the performance of sub-scale is quite comparable to that of the full-scale which validates use of the sub-scale in testing new ATM concepts. Keywords: Unmanned